A method for modeling sandstone-type uranium deposits based on stochastic algorithms

Abstract

The present application relates to the technical field of uranium ore geology and geophysics, and particularly relates to a sandstone type uranium mine modeling method based on a random algorithm, comprising the following steps: step one, collecting known data; step two, counting data; step three, designing the four-to range of a grid, the model size and the growth factor, and establishing a model grid in a rectangular division manner; step four, determining the geometric parameters and the resistivity parameters of sand body strata, and randomly generating a plurality of sand body strata model combinations with different characteristics; step five, determining the top interface shape, and assigning values from the deep region of the top interface to the bottom of the grid region to obtain a crystallization basement model; step six, assigning background mudstone resistivity values to other grids; and step seven, converting the grid model into a png file as a label file of an intelligent system training library. The present application solves the problems of insufficient training library data sample quantity and poor diversity of training convolutional neural network models when using magnetotelluric results to interpret sandstone type uranium mines.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of uranium ore geology and geophysical exploration, and more particularly to a sandstone type uranium mine modeling method based on a random algorithm. BACKGROUND

[0002] In the exploration of sandstone type uranium mines, using magnetotelluric inversion data to divide the crystalline basement and the distribution range of the sand body is an important means to delineate favorable uranium mineralization areas. When using deep learning networks to supervise the training of the inversion results, a large number of resistivity numerical models need to be established to simulate the corresponding images and labeled results, so that the convolutional neural network has the ability to intelligently identify the crystalline basement and the sand body in the inversion results.

[0003] At present, the technology of using artificial intelligence to intelligently identify the basement and the sand body in the inversion results is still in the exploratory stage. The method of generating training images and labels relies on technical personnel to establish numerical models and then calculate them in the forward and inverse software. This method has great limitations in model diversity and complexity. At the same time, due to the low efficiency of manual modeling, it is not conducive to the requirement of deep learning networks for a large number of training data and labels. This greatly restricts the popularization and application of artificial intelligence technology in the field of magnetotelluric method. SUMMARY

[0004] The present application is to overcome the problem of insufficient number of training database data samples and poor diversity of convolutional neural network models when using magnetotelluric results for intelligent geological interpretation of sandstone type uranium mines, and aims to provide a sandstone type uranium mine modeling method based on a random algorithm.

[0005] A sandstone type uranium mine modeling method based on a random algorithm, comprising the following steps:

[0006] Step one, collect known data: collect geological data of the work area and its surrounding area;

[0007] Step two, statistical data: statistics of the spatial distribution of sand body and basement and the value range of electrical parameters, the spatial distribution parameters of sand body and basement include the depth range of sand body and basement top and bottom interface, the inclination angle of sand body stratum, and the electrical parameters include the resistivity value range of crystalline basement and sand body;

[0008] Step three, design model grid: according to the size and depth of the detection object, design the four-to-range, model size and growth factor of the grid, and establish the model grid by using rectangular partitioning;

[0009] Step four, obtain the sand body stratum model: determine the geometric parameters and resistivity parameters of the sand body stratum, and randomly generate a plurality of sand body stratum model combinations with different characteristics;

[0010] Step five, get the crystalline basement model: first determine the crystalline basement top interface shape, and then assign the crystalline basement top interface to the deep area to the bottom of the grid area;

[0011] Step six, assign the background mudstone resistivity value to other grids: assign the background part of the model to the resistivity value of the lithology mainly composed of mudstone in the working area, and the model grid generation is completed;

[0012] Step seven, convert the grid model into a png file as a label file for the intelligent system training library.

[0013] Further, in step one, collecting geological data, the geological data includes drilling core, resistivity logging data and rock resistivity statistical data.

[0014] Further, in step two, the statistical data, according to the collected working area data, the data cannot be counted, then according to the surrounding geological data experience given.

[0015] Further, in step three, the design of the model grid, the selection principle is as follows:

[0016] Horizontal direction: the horizontal grid range is equivalent to the length of the profile to be interpreted, or meets the complexity of the basement and sand body in model design, the minimum value of horizontal grid is selected as half of the point distance, and the horizontal grid growth coefficient on both sides of the profile is selected as 1.5;

[0017] Vertical direction: the vertical grid range extends from the surface to 10 times the maximum detection depth, the first row of vertical grid value is selected together with the vertical growth factor, and the value of the first row of vertical grid and the vertical growth factor is determined by the minimum target layer thickness and the burial depth.

[0018] Further, in step four, the following steps are included:

[0019] S41, determine the number of sand body layers according to the statistical situation;

[0020] S42, randomly generate sand body burial depth: determine the sand body layer center burial depth value range according to the statistical sand body top and bottom layer burial depth, and then determine the sand body layer center burial depth as any random number in the value range;

[0021] S43, randomly segment the sand body: according to the stratigraphic variation in the working area, specify the maximum number of sand body segments p max , generate segmented monocline sand body, determine the stratigraphic segmentation number P as any random number between 1 and the maximum number of sand body segments p max , the value formula of stratigraphic segmentation number P is

[0022] P = random (1, p max )

[0023] wherein P is an integer, random(a, b) is a random function, and a random number is selected in the interval [a, b];

[0024] S44, randomly generating sand body thickness: Since it is difficult to identify thin sand bodies by magnetotelluric method, it is suggested that the general sand body thickness is not less than 50 meters and not greater than the statistical value determined in step two. In order to increase the complexity of the model, a random number of 20% is added to the sand body thickness value, so the value formula of the sand body thickness d is

[0025] d=random(d min , d max )+20%*random(d min , d max )

[0026] wherein the value of d min is 50 meters, and the value of d max is the maximum value of the sand body thickness statistically determined in step two;

[0027] S45, randomly selecting the dip angle of each segment of sand body: according to the geological data, the maximum stratigraphic dip angle range is determined, and the dip angle θ of each segment of sand body is randomly specified in the maximum stratigraphic dip angle range. The value formula of the dip angle θ of the segment of sand body is

[0028] θ=random(-θ max , θ max );

[0029] S46, randomly selecting the sand body resistivity value: the sand body resistivity value is randomly selected in the sand body resistivity value range statistically determined in step two as the sandstone layer resistivity value;

[0030] S47, based on the spatial geometric parameters randomly generated in S41-S45, the sand body stratigraphic area is obtained by using the segmented linear equation;

[0031] S48, the grid in the sand body stratigraphic range is assigned the random sand body resistivity value in S46 to form a complete sand body stratigraphic model.

[0032] Further, in step five, the following steps are included:

[0033] S51, randomly determining the number of inflection points of the crystallization base: referring to the sand body segmentation parameters, the number of inflection points of the fluctuation base top interface is determined by using a random function;

[0034] S52, randomly determine the position of the inflection point in space: the horizontal coordinates of each inflection point are uniformly distributed in the survey area, the vertical coordinate range of the inflection point is obtained according to the previous data or geological experience, and the vertical coordinate of the top interface node is generated in the vertical coordinate range of the inflection point through a random way;

[0035] S53, connecting each node by using quadratic curve, spline curve and other high-order interpolation methods to form the top interface of the crystalline basement;

[0036] S54, randomly select the crystalline basement resistivity value in the statistical range: according to the rock physical parameter test result, the crystalline basement resistivity value is randomly selected in the corresponding resistivity value range of the crystalline basement lithology;

[0037] S55, assign the crystalline basement resistivity value to the grid in the crystalline basement area: the top interface of the crystalline basement is assigned to the bottom of the grid area, and a complete crystalline basement model is formed.

[0038] Further, in step S43, when the maximum number of segments p of the designated sand body is max If the stratum changes quickly in the region, the value is appropriately increased, and if the stratum is relatively stable in the region, the value is relatively reduced.

[0039] Compared with the prior art, the technical scheme provided by the present application has the following beneficial effects:

[0040] The sandstone type uranium mine modeling method based on the random algorithm in the present application is based on the existing sand body stratum spatial distribution and electrical parameter, generates spatial geometric parameters and resistivity parameters based on the above multiple random conditions, and can obtain a model with good diversity and in line with the characteristics of the sand body in the working area through a linear equation; based on the existing crystalline basement spatial distribution and electrical parameter, in the exploration of the sandstone type uranium mine, the crystalline basement is generally only detected for its top interface, and the bottom interface is buried deeper than the profile detection depth, so in modeling, the top interface shape is first determined, and the grid below it belongs to the crystalline basement range; based on the above principle, only a few inflection points of the top interface of the basement are determined in the profile, and the spatial distribution of the top interface of the basement is obtained by interpolation, wherein the number, horizontal position and vertical position of the inflection points are generated by a random function, thereby obtaining a model with good diversity and in line with the characteristics of the crystalline basement in the working area. The sand body stratum model and the crystalline basement model are generated from the above two aspects, and the other parts of the model are valued, thereby completing the model grid, generating a png picture of the model grid, taking the png picture as a label file of an intelligent system training library, and solving the problem of insufficient number of training library data samples and poor diversity of the training convolutional neural network model in the intelligent geological interpretation of the sandstone type uranium mine using the magnetotelluric result. BRIEF DESCRIPTION OF DRAWINGS

[0041] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0042] Figure 1 is a method flow chart of the sandstone type uranium mine modeling method based on a random algorithm in the present application.

[0043] Figure 2 is a flow chart of obtaining a sand body stratigraphic model in the present application.

[0044] Figure 3 is a flow chart of obtaining a crystalline basement model in the present application.

[0045] Figures 4-7 is a random model image generated in an embodiment of the present application.

[0046] Figures 4-7 In the middle: white is a sand body stratum, gray is a crystalline basement, and black is a stratum mainly composed of mudstone in the background. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the present application. DETAILED EMBODIMENT

[0048] The specific implementation process of the sandstone type uranium mine modeling method based on a random algorithm in the present application will be described below by taking the Ihegao'le region in the Erenhot Basin as an example. The region is a key investigation area of a sandstone type uranium mine, and a small amount of drilling, electrical exploration and geological investigation work has been implemented in the early stage, and a series of geological recognitions have also been formed.

[0049] As shown in Figure 1 , the sandstone type uranium mine modeling method based on a random algorithm in the present application includes the following steps:

[0050] Step one, collecting known data: collecting geological data of the work area and its surrounding region, and the work area mainly collects the results of drilling, resistivity logging and rock sample resistivity testing.

[0051] Step two, data statistics:

[0052] Through the results of the early electrical exploration, it is considered that the magnetotelluric survey can divide a set of thick sandstone layers in the Early Cretaceous Bayinhu Group in this area. Therefore, when using the drilling logging results to count the top and bottom interface depths of the sandstone formation, the thin sandstone and mud-sand interbedded layers are discarded and processed, and the top and bottom interface depths of the sandstone formation are obtained.

[0053] Through comprehensive analysis of the electrical exploration results and regional geological understanding, the regional strata change from sandstone and mudstone-based sedimentary strata to igneous and metamorphic rocks, so the Pre-Cretaceous strata are defined as the crystalline basement of the working area. Based on the known data, it is inferred that the top interface depth of the working area basement is distributed from the near-surface to the maximum detection depth (about 1000 meters).

[0054] Through the statistical results of rock physical property parameter testing, the resistivity value range of the sandstone formation and the crystalline basement is determined.

[0055] Step three, design the model grid: rectangular partitioning is adopted.

[0056] In the horizontal direction, the survey line area is evenly divided by 50-meter grid, the survey line length is 10000 meters, and the survey line outer area is gradually increased by 1.5 times increment, and the horizontal grid length is about 10 times the survey line length.

[0057] In the vertical direction, according to the minimum target layer thickness and depth, the first layer thickness is selected as 5 meters, the longitudinal grid thickness increment coefficient is 1.05, and the longitudinal grid length is about 10 times the exploration depth (1000 meters).

[0058] Step four, obtain the sandstone formation model: determine the spatial distribution and electrical parameters of the sandstone formation, and obtain a variety of sandstone combination models through random combination of parameters such as depth, thickness, formation dip angle, and resistivity value. As shown in Figure 2 , the specific implementation steps are as follows:

[0059] S41, determine the number of sandstone layers according to the statistical situation: according to the statistical situation of the working area, it is determined that the number of sandstone layers in this area is 1 layer.

[0060] S42, randomly generate sandstone depth: according to the statistical results of the top and bottom layers of the sandstone from the drilling logging, a random number in the range of 200-500 meters is determined for the center depth of the sandstone formation in each model.

[0061] S43, randomly segment the sandstone:

[0062] First, determine the maximum number of sandstone formation segments, according to the known geological data of the working area, the sandstone changes relatively quickly in the working area, so for a 10000-meter profile, the maximum number of sandstone formation segments p max is defined as 4.

[0063] Then generate the segmented monocline sand body formation, the maximum sand body formation segment number p max is 4, at this time the formation segment number P in each model is selected randomly from 1, 2, 3, and 4.

[0064] P = random (1, p max ) = random (1, 4)

[0065] That is, the formation segment number P in each model is randomly selected from any value of 1, 2, 3, and 4.

[0066] S44, randomly generate sand body thickness: the statistics of sand body thickness is 50-200 meters, so any random value in this interval is selected as the sand body thickness d, and a random number of 20% is added to this thickness to make the sand body formation have certain lateral heterogeneity. The value of the sand body thickness d is

[0067] d = random (d min , d max ) + 20% * random (d min , d max )

[0068] = random (50, 200) + 20% * random (50, 200)

[0069] S45, randomly select the dip angle of each segment of the sand body formation: according to the geological data, the occurrence of the strata in this area is generally not more than 4°, so the maximum strata dip angle range is determined to be -4°-4°. Different segments of the sand body in each model will select a random value in this interval as the sand body formation dip angle. The value of the segmented sand body formation dip angle θ is

[0070] θ = random (-θ max , θ max ) = random (-4°, 4°)

[0071] S46, randomly select the sand body resistivity value: according to the statistical results of rock physical property testing, the resistivity of sandstone above medium grain in this area is generally 30-60 ohm meters, and a random value in this interval is selected as the sand body resistivity in each model. The sand body resistivity value is randomly selected from the statistical sand body resistivity value range as the sandstone layer resistivity value.

[0072] S47, based on the spatial geometric parameters randomly generated in S41-S45, use the segmented linear equation to obtain the sand body formation area;

[0073] S48, assign the grid in the sand body formation range to the random sand body resistivity value in S46 to form a complete sand body formation model.

[0074] Step five, obtain the crystalline basement model: in the exploration of sandstone type uranium deposit, the crystalline basement is generally only detected for its top interface, and the bottom interface is buried deeper than the profile detection depth, so when modeling, only the top interface shape needs to be determined first, and the grid below it belongs to the range of the crystalline basement. Based on the above principle, only the inflection points of the top interface of the basement need to be determined in the profile, and the spatial distribution of the top interface of the basement can be obtained by interpolation. The number, horizontal position and vertical position of the inflection points are generated by random functions. As shown in the following figure, the specific implementation steps are as follows: Figure 3

[0075] S51, randomly determine the number of inflection points of the crystalline basement: refer to the sand body segmentation parameters, and determine the number of inflection points of the fluctuating top interface of the basement by using a random function; the number of inflection points selected in this work is 4.

[0076] S52, randomly determine the position of the inflection point in space: the horizontal coordinates of each inflection point are uniformly distributed in the survey area, and any random value within ±20% of the horizontally divided nodes of the grid area is adopted. According to the information obtained by comprehensive analysis, the vertical burial depth of the top interface node is 0-1000 meters, and the vertical coordinate of the top interface node is generated in this interval by a random function.

[0077] S53, connect each node by using high-order interpolation methods such as quadratic curve and spline curve to form the top interface of the crystalline basement;

[0078] S54, randomly select the resistivity value of the crystalline basement in the statistical range: according to the test results of rock physical parameters, the resistivity value corresponding to the lithology of the crystalline basement is in the range of 300-1000 ohm meters, and any value in the interval is randomly selected as the resistivity value of the basement for each model.

[0079] S55, assign the crystalline basement resistivity value to the grid in the crystalline basement area: assign the grid from the deep part of the top interface of the crystalline basement to the bottom grid of the model to the basement resistivity value, and form a complete crystalline basement model.

[0080] Step six, assign the background mudstone resistivity value to other grids: assign the background part of the model other than the sandstone formation model and the crystalline basement model grid to the lithology mainly composed of mudstone, and directly input the constant 10 ohm meters as the resistivity value of the model grid.

[0081] Step seven, generate png picture of the model grid as the label file of the intelligent system training library. As shown in the following figure: Figure 4 to the following figure: Figure 7 The random model image generated in this embodiment is shown in the following figure, where white represents sandstone formation, gray represents crystalline basement, and black represents the background mainly composed of mudstone.

[0082] ​The sandstone type uranium mine modeling method based on a random algorithm in the application, based on the existing sand body stratum spatial distribution and electrical parameters, based on the above multiple random conditions to generate spatial geometric parameters and resistivity parameters, and through linear equations, a model with good diversity and in line with the characteristics of the sand body in the working area can be obtained; based on the existing crystalline basement spatial distribution and electrical parameters, in the exploration of sandstone type uranium mines, the crystalline basement is generally only detected for its top interface, and the bottom interface is buried deeper than the profile detection depth, so when modeling, only the top interface shape needs to be determined, and the grid below it belongs to the crystalline basement range, based on the above principle, only the inflection points of the top interface of the basement need to be determined in the profile, and the spatial distribution of the top interface of the basement can be obtained by interpolation, wherein the number, horizontal position and vertical position of the inflection points are generated by a random function, so that a model with good diversity and in line with the characteristics of the crystalline basement in the working area is obtained. The sand body stratum model and the crystalline basement model are generated from the above two aspects, and the other parts of the model are valued, so as to complete the model grid, generate a png picture of the model grid, and use it as a label file of the intelligent system training library, solving the problem of insufficient number of training library data samples and poor diversity of the training convolutional neural network model when using the magnetotelluric results to interpret the sandstone type uranium mine.

[0083] Finally, it should be noted that: the above only describes the preferred embodiments of the application and is not used to limit the application, although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. A stochastic algorithm based modeling method for sandstone type uranium deposits, characterized in that, The method comprises the following steps: Step one, collecting known information: collecting geological information of the work area and its surrounding area; Step two, statistical data: statistics of the spatial distribution of sand bodies and basement and the value range of electrical parameters, the spatial distribution parameters of sand bodies and basement including the depth range of sand body and basement top and bottom interface, the angle of sand body stratigraphic dip, and the electrical parameters including the resistivity value range of crystalline basement and sand body; Step three, designing model grid: designing the four-to range, model size and growth factor of the grid according to the size and depth of the detection object, and establishing the model grid by using rectangular partitioning; Step four, obtaining sand body stratigraphic model: determining the geometric parameters and resistivity parameters of the sand body stratigraphy, and randomly generating a plurality of sand body stratigraphic model combinations with different characteristics, in step four, comprising the following steps: S41, determining the number of sand body layers according to the statistical situation; S42, randomly generating sand body depth: determining the value range of the center depth of each sand body layer according to the statistical top and bottom layer depth of the sand body, and then determining the center depth of the sand body layer as an arbitrary random number in the value range; S43, randomly segmenting the sand body: according to the stratigraphic variation in the working area, the maximum segment number p of the sand body is specified max , generating a segmented monocline sand body stratum, determining the stratigraphic segment number P as an arbitrary random number between 1 and the maximum segment number p of the sand body max , the value formula of the stratigraphic segment number P is P = random(1, p max ) Wherein P is an integer, and random (a, b) is a random function for selecting a random number in the interval [a, b]; S44, randomly generating sand body thickness: the value formula of sand body thickness d is d = random(d min , d max ) + 20% * random(d min , d max ) where d min has a value of 50 meters, d max has a value of the maximum sand thickness calculated in step two. S45, randomly selecting the dip angle of each section of sand body stratigraphy: according to the geological data, the maximum stratigraphic dip angle range is determined, and the dip angle θ of each section of sand body stratigraphy is randomly specified in the maximum stratigraphic dip angle range, and the value formula of the dip angle θ of the section of sand body stratigraphy is θ = random(-θ max , θ max ); S46, randomly selecting the sand body resistivity value: randomly selecting the sand body resistivity value in the sand body resistivity value range in step two; S47, obtaining the sand body stratigraphic area based on the spatial geometric parameters randomly generated in S41-S45 by using the piecewise linear equation; S48, assigning the grid in the sand body stratigraphic range with the random sand body resistivity value in S46 to form a complete sand body stratigraphic model; Step five, obtaining the crystalline basement model: firstly determining the top interface shape, and then assigning the area from the top interface to the bottom of the grid area, in step five, comprising the following steps: S51, randomly determining the number of crystalline basement inflection points: referring to the sand body section parameters, the number of inflection points of the top interface of the fluctuating basement is determined by using a random function; S52, randomly determining the position of the inflection point in space: the horizontal coordinates of each inflection point are uniformly distributed in the survey area, the coordinate range of the vertical depth of the inflection point is obtained according to the previous information or geological experience, and the vertical coordinate of the top interface node is generated in the coordinate range of the vertical depth of the inflection point by random method; S53, connecting each node by using high-order interpolation to form the top interface of the crystalline basement; S54, randomly selecting the crystalline basement resistivity value in the statistical range: according to the rock physical property parameter test result, randomly selecting the model basement resistivity value in the resistivity value range corresponding to the lithology of the crystalline basement; S55, assigning the grid in the crystalline basement area with the crystalline basement resistivity value: assigning the area from the top interface of the crystalline basement to the bottom of the grid area to the basement resistivity value to form a complete crystalline basement model; Step six, other grid background mudstone resistivity value: sand body formation model and crystalline basement model, model background part of the value of the working area to mudstone lithology resistivity value, model grid generation end; Step seven, the grid model is converted into a png file as a label file for the intelligent system training library.

2. The random algorithm based sandstone type uranium deposit modeling method according to claim 1, characterized in that, In step one, in the collection of geological data, the geological data includes drilling core, resistivity logging data and rock resistivity statistical data.

3. The random algorithm based sandstone type uranium deposit modeling method according to claim 1, characterized in that, In step two, in the statistical data, the data that cannot be counted according to the collected working area data is given according to the geological data experience of the surrounding area.

Images