A method for obtaining random fields of geological models based on conditional random field decision trees

Through the method of combining conditional random field and decision tree, a parameterized geological model was constructed, which solved the problem of geological spatial variability in the geological model, and achieved the actual situation that accurately reflects the geological conditions and the certainty at the drilling point, satisfying the Various engineering application requirements.

CN115344985BActive Publication Date: 2024-11-01HUAZHONG UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210798250.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-06
Publication Date
2024-11-01
Estimated Expiration
2042-07-06

AI Technical Summary

Technical Problem

The prior art cannot effectively consider geological spatial variability, which leads to the inability to accurately reflect the certainty of soil at the drilled hole and the uncertainty of geological conditions at non-drilled holes in the construction of geological models.

Method used

The parameterized geological model acquisition method based on conditional random airfield and decision tree is adopted. Conditional random airfield is constructed through drilling data, combined with unconditional random airfield theory and Kriging method, and three-dimensional/two-dimensional conditional random airfields are generated to characterize the variation of geological space. and certainty at the drilling.

Benefits of technology

It realizes the actual situation of accurately reflecting the geological conditions, provides a parameterized geological model construction method that can efficiently and conveniently consider geological space variability, meets the needs of various engineering application scenarios, and can better serve the actual needs of engineers. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115344985B_ABST
    Figure CN115344985B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field related to geological survey, and discloses a method for obtaining a random field of a geological model based on a conditional random field decision tree. The method comprises the following steps: S1 selects a geotechnical analysis domain of a proposed project, drills holes in the analysis domain and records the drilling information; S2 constructs a decision tree based on a conditional random field according to the variation characteristics of the geological space, the availability of drilling data and the distribution of soil layers; S3 uses the drilling information of the drilling point recorded in step S1 as input, selects the corresponding parameterized geological modeling method using the decision tree based on the conditional random field constructed in step S2, and calculates and obtains a parameterized geological geometric model, a gridded geometric model and the random field values ​​corresponding to each grid node in the gridded geometric model; S4 characterizes the random field values ​​obtained in step S3, so as to realize the visualization of the random field values ​​of the geotechnical model to be analyzed. Through the present invention, the problem that the variability of geological space cannot be considered in the construction of geological models is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field related to geological survey, and more specifically, relates to a method for obtaining a random field of a geological model based on a conditional random field decision tree. Background Art

[0002] The deposition of soil has gone through a long period of time. It is affected by many factors during the formation process, and its random distribution is very complex. Although the properties of a point in the soil are objectively determined, they are random before the test. There is obvious random variability in the parameter changes between different points in the soil. Generally speaking, soil parameters have obvious uncertainty and spatial variability. Many domestic and foreign experts have conducted a large number of statistical analyses on soft soils in different regions.

[0003] In general, the uncertainty of soil parameters can be simply characterized based on the average value of the parameters and the corresponding coefficient of variation (COV). According to previous actual engineering data and scholars' research, the variability of soil physical indicators is often smaller than that of mechanical indicators. The coefficient of variation of gravity is usually less than 10%, and the average value is usually in the range of 13 to 20 kN / m 3 . The coefficients of variation of soil cohesion and internal friction angle are generally between 10% and 30% and 5% and 15%, respectively, and these two shear strength indicators usually obey normal distribution or log-normal distribution. The coefficient of variation of the modulus of soil usually varies between 10% and 20%, usually showing normal distribution or log-normal distribution. However, the mean and coefficient of variation of soil parameters are single-point statistical parameters and cannot fully describe the spatial structural characteristics of the soil. Therefore, in order to characterize the spatial structural characteristics of soil parameters, the researchers introduced random field theory to characterize the random variability of soil parameters through statistical indicators of soil parameters (such as mean, coefficient of variation and correlation distance, etc.) and related equations.

[0004] However, in the field of underground engineering, geological conditions are usually surveyed by drilling soil samples. The soil samples at the borehole can more accurately reflect the geological conditions at that location, but the geological conditions at non-drilling locations still have significant uncertainty. Although traditional random field theory can characterize the uncertainty of geological conditions, it cannot take into account the certainty of the soil sampled at the borehole, that is, the geological conditions at the borehole are still randomly generated under the framework of traditional random field theory.

[0005] On the other hand, with the continuous advancement of informatization in the engineering field, parametric representation of engineering objects has become a trend. By parametrically modeling engineering objects, the geometric and non-geometric properties of engineering objects can be more accurately represented, which facilitates engineering personnel to understand and control engineering projects. In particular, for three-dimensional geological space, the automatic construction and update of parametric geological models that consider the spatial variability of geotechnical materials are particularly important for understanding engineering difficulties and risk sources. The engineering field urgently needs a vivid, convenient and efficient parametric geological model construction method that can consider geological spatial variability and face a variety of actual engineering application scenarios. Summary of the invention

[0006] In view of the above defects or improvement needs of the prior art, the present invention provides a method for obtaining a random field of a geological model based on a conditional random field decision tree, which solves the problem that geological spatial variability cannot be considered in geological model construction.

[0007] To achieve the above object, according to the present invention, a method for obtaining a parameterized geological model based on a conditional random field and a decision tree is provided, the method comprising the following steps:

[0008] S1 selects a geotechnical analysis domain of the proposed project, selects multiple locations within the analysis domain for drilling, and records the drilling information of each drilling point;

[0009] S2 sets multiple discriminants of conditional random fields according to the variation characteristics of geological space, the availability of drilling data and the distribution of soil layers. Each discriminant corresponds to a corresponding parameterized geological model method, thereby forming a decision tree based on conditional random fields.

[0010] S3 takes the drilling information of the drilling point recorded in step S1 as input, selects the corresponding parameterized geological model method using the decision tree based on conditional random field constructed in step S2, and calculates and obtains the parameterized geological geometric model, the gridded geometric model, and the random field value corresponding to each grid node in the gridded geometric model;

[0011] S4 represents the random field value at each grid node in the gridded geometric model obtained in step S3, so as to realize the visualization of the random field value of the rock and soil to be analyzed.

[0012] Further preferably, in step S2, the conditional random field is formed by combining unconditional random field theory and Kriging method.

[0013] Further preferably, in step S1, the drilling information includes the coordinates of the drilling point, whether it is a corner point, the height of the rock and soil layer and the rock and soil properties of the measuring point selected at each drilling point.

[0014] Further preferably, the geotechnical properties are the friction angle, cohesion, modulus or gravity of the geotechnical layer where each measuring point is located.

[0015] Further preferably, in step S2, the discriminant of the conditional random field is: whether the soil parameters have clear expected values, whether the attributes of all measuring points at the drilling location can be obtained, and whether the soil stratification is obvious.

[0016] Further preferably, in step S2, the updating method includes SL-SK-CRF, SL-OK-CRF, BH-SK-CRF and BH-OK-CRF.

[0017] Further preferably, in step S4, the random field value is represented by a color in a color lookup table.

[0018] Further preferably, in step S3, the geometric model and the random field values ​​corresponding to each grid node in the geometric model constitute a parameterized geological model.

[0019] Further preferably, in step S3, the meshed geometric model is obtained by a Delaunay triangulated mesh generation algorithm.

[0020] In general, the above technical solutions conceived by the present invention have the following beneficial effects compared with the prior art:

[0021] 1. The present invention adopts a combination of conditional random fields and decision trees to construct and obtain a parameterized geological model. The geological model constrains the unconditional random field by sampling soil at the borehole as a condition point, thereby reflecting the variability of the geological space and accurately reflecting the actual situation of the geological conditions.

[0022] 2. The conditional random field in the present invention is formed by combining the unconditional random field theory and the Kriging method. In the above two methods, the unconditional random field theory is used to characterize the spatial variability and correlation of soil parameters, and the Kriging method is used to consider the certainty of the soil parameters at the borehole to constrain the unconditional random field. Through the combination of these two methods, a three-dimensional / two-dimensional conditional random field can be generated quickly, efficiently and accurately to characterize the parameterized geological model;

[0023] 3. In the present invention, whether the soil parameters have clear expected values, whether the attributes of all the measuring points at the borehole can be obtained, and whether the soil stratification is obvious are selected as discriminants for generating conditional random fields. These discriminants are summarized and summarized in combination with the key problems encountered in actual projects, taking into account the actual engineering experience of engineers (whether there are clear expectations for the soil), the availability of engineering data (whether the measuring points of the borehole can be directly obtained), and the boundary conditions of the soil when generating the conditional random field (whether the soil stratification is obvious). By considering these key engineering problems, the needs of actual projects can be more effectively met;

[0024] 4. In the present invention, SL-SK-CRF, SL-OK-CRF, BH-SK-CRF and BH-OK-CRF are adopted. The above four methods focus on solving various key problems encountered in actual engineering. Among them, SL and BH can be used to solve the boundary condition problem caused by soil stratification, and SK and OK can solve the problem of whether the soil parameters have clear expectations and the properties of the measuring points can be obtained. Different solutions are given for different key problems, which can better serve the engineering;

[0025] 5. The present invention provides an automatic construction method for a parameterized geological geometric model based on boreholes, and generates a three-dimensional / two-dimensional conditional random field that meets various practical engineering application scenarios through the parameterized geological geometric model. The parameterized geological geometric model and the conditional random field together constitute a parameterized geological model. Combined with a visualization method, it can efficiently, conveniently and intuitively characterize the spatial variability, correlation and certainty of the geology at the borehole. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A method for obtaining a random field of a geological model based on a conditional random field decision tree constructed according to a preferred embodiment of the present invention;

[0027] Figure 2 It is a visualization flow chart of a three-dimensional random field of a layered parameterized geological model constructed according to a preferred embodiment of the present invention;

[0028] Figure 3 It is a visualization flow chart of a two-dimensional random field of a layered parameterized geological model constructed according to a preferred embodiment of the present invention;

[0029] Figure 4 It is a visualization flow chart of three-dimensional and two-dimensional random fields of a drill hole parameterized geological model constructed according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0031] like Figure 1 As shown, a method for obtaining a geological model random field based on a conditional random field decision tree comprises the following steps:

[0032] S1. Parametric input of drilling information is achieved by defining a standardized drilling data structure for characterizing geological spatial variability;

[0033] S2, construct a conditional random field characterization decision tree based on geological spatial variability to select the appropriate automatic construction of parameterized geological models and conditional random field characterization methods;

[0034] S3. Calculate the random value of each grid node in the parameterized geological model, as follows:

[0035] S301 generates an unconditional random field of soil parameters, whose spatial correlation can be represented by the following variogram:

[0036]

[0037] Among them, γ is the variation function, μ is the Gaussian random field, x1 and x2 are two arbitrary points in u(x), and E is the expected value.

[0038] The corresponding covariance and correlation functions can be expressed as:

[0039]

[0040]

[0041] Among them, C is the covariance, ρ is the correlation coefficient, σ is the standard deviation, r is the distance between two arbitrary points, and l is the correlation distance.

[0042] S302: The soil parameter conditional random field can be realized by the Kriging method based on the unconditional random field. The Kriging method can calculate the soil parameter conditional random field of a given point x. i (i=1,2,…,n) fixed condition value z(x i ) to estimate the value z0 at the unknown point x0, thereby ensuring that the estimated value is the same as the mathematical expectation of the actual value and the variance is minimal. The best linear unbiased estimate of the value z0 at point x0 can be obtained by the following formula:

[0043]

[0044] Where n is the number of conditional points, ω i is the conditional point x i The relevant weighting factor.

[0045] Specifically, the following four methods can be generated by combining unconditional random fields with Kriging:

[0046] (1) SL-SK-CRF method

[0047] According to the conditional random field characterization decision tree, if the geological survey provides clear expected values ​​of soil parameters E p (A1), the attributes of all measuring points in all boreholes (with known measuring point locations) cannot be obtained (B2) and the soil stratification is obvious (C1), then the soil layer-based (SL) parameterized geological model based on simple kriging (SK) conditional random field (CRF) is automatically constructed and updated (SL-SK-CRF method).

[0048] The implementation of the SL-SK-CRF method specifically includes the following steps:

[0049] (a) obtaining a representation of the stratum interface based on the borehole by normalizing the borehole data described in S1, and constructing a layered geological geometric model with the borehole information by the upper and lower limits of the stratum interface;

[0050] (b) specifying the boreholes involved in the conditional random field representation;

[0051] (c) If the specified boreholes are located on the same plane, the parameterized geological model can be characterized by a two-dimensional or three-dimensional conditional random field, otherwise the parameterized geological model can only be characterized by a three-dimensional conditional random field. Based on the frontier-advanced Delaunay triangulated mesh generation algorithm, the corresponding two-dimensional or three-dimensional geological model mesh is generated;

[0052] (d) The expected value E of the soil parameters (such as cohesion, friction angle and modulus) at each drilling point p,i,j and preset conditional random field configuration parameters as input, and the corresponding conditional random fields are generated by the conditional random field method based on the simple kriging method to characterize the spatial variability of the parameterized geological model.

[0053] (2) SL-OK-CRF method

[0054] According to the conditional random field characterization decision tree, if the geological survey does not provide a clear expected value E of the soil parameters p(A2), the attributes of all measuring points in all boreholes (known measuring point locations) can be obtained (B1) and the soil stratification is obvious (C1), then the soil layer-based (SL) parameterized geological model based on ordinary kriging (OK) conditional random field (CRF) is automatically constructed and updated (SL-OK-CRF method). The difference between this method and SL-SK-CRF is that the soil parameters (such as cohesion, friction angle and modulus, etc.) of the measuring point corresponding to each borehole point are measured values ​​E i,j , which is calculated based on the ordinary Kriging method.

[0055] (3) BH-SK-CRF

[0056] According to the conditional random field characterization decision tree, if the geological survey provides clear expected values ​​of soil parameters E p (A1), the attributes of all measuring points in all boreholes (with known measuring point locations) cannot be obtained (B2) and the soil stratification is not obvious (C2), then the borehole (BH) parameterized geological model automatic construction and update method based on simple kriging (SK) conditional random field (CRF) (BH-SK-CRF method) is adopted.

[0057] The implementation of the BH-SK-CRF method specifically includes the following steps:

[0058] (a) obtaining key drilling points of the outer contour of all drilling data (i.e., the envelope point set of all drillings) through the normalized drilling data described in S1, and constructing a geological geometric model with drilling information through the outer contour points;

[0059] Steps (b) to (d) of this method are the same as those of the SL-SK-CRF method.

[0060] (4) BH-OK-CRF method

[0061] According to the conditional random field characterization decision tree, if the geological survey does not provide a clear expected value E of the soil parameters p(A2), the attributes of all measuring points in all boreholes (the positions of measuring points are known) can be obtained (B1) and the soil stratification is not obvious (C2), then the borehole (BH) parameterized geological model automatic construction and update method based on ordinary kriging (OK) conditional random field (CRF) (BH-OK-CRF method) is used. The difference between this method and BH-SK-CRF is that the soil parameters (such as cohesion, friction angle and modulus, etc.) of the measuring point corresponding to each borehole point are measured values ​​E i,j , which is calculated based on the ordinary Kriging method.

[0062] S4. The two-dimensional conditional random field representation of the parameterized geological model is realized by assigning a corresponding random field value to each vertex of the triangular mesh, and the three-dimensional conditional random field representation of the parameterized geological model is realized by assigning a corresponding random field value to each vertex of the tetrahedral mesh. These triangular or tetrahedral meshes with vertex random field values ​​are finally visualized through a color lookup table (LUT). The correctness of the parameterized geological model represented by the two-dimensional / three-dimensional conditional random field can be reflected by the absolute difference of the conditional random field.

[0063] like Figure 2 As shown in the figure, it is a flowchart of generating a hierarchical parameterized geological geometry model by normalizing the borehole data and then generating a three-dimensional conditional random field of the specified borehole. This hierarchical parameterized geological geometry model is suitable for conditional random field generation based on SL-SK-CRF and SL-OK-CRF methods.

[0064] like Figure 3 As shown in the figure, the layered parameterized geological model is triangulated, then the two-dimensional conditional random field is calculated, and finally the two-dimensional conditional random field is visualized. Among them, the cloud map of the absolute difference of the two-dimensional conditional random field at the location of the borehole of the two-dimensional conditional random field has the darkest color, and the shape of the cloud map is consistent with the shape and position of the borehole very well, while the cloud map at the non-drilling location is random, which meets the characteristics of the conditional random field, that is, authenticity and accuracy are guaranteed at the conditional constraint points, and spatial variability is ensured at the non-conditional points.

[0065] like Figure 4As shown in the figure, it is a flowchart of generating a borehole parametric geological geometric model by normalizing borehole data, then calculating a 3D / 2D conditional random field, and finally visualizing the 3D / 2D conditional random field. This borehole parametric geological model is suitable for conditional random field generation based on the BH-SK-CRF and BH-OK-CRF methods. Its main feature is that the geometric model is constructed by obtaining the outer contour key drilling points of all borehole data (i.e., the envelope point set of all boreholes) and the outer contour points, which is different from the construction method of the hierarchical geometric model.

[0066] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for obtaining a geological model random field based on a conditional random field decision tree, characterized in that: The method comprises the following steps: S1 selects a geotechnical analysis domain of the proposed project, selects multiple locations within the analysis domain for drilling, and records the drilling information of each drilling point; S2 sets multiple discriminants of conditional random fields according to the variation characteristics of geological space, the availability of drilling data and the distribution of soil layers. Each discriminant corresponds to a corresponding parameterized geological model method, thereby forming a decision tree based on conditional random fields. S3 takes the drilling information of the drilling point recorded in step S1 as input, selects the corresponding parameterized geological model method by using the decision tree based on conditional random field constructed in step S2, and calculates and obtains the parameterized geological geometric model, the gridded geometric model and the random field value corresponding to each grid node in the gridded geometric model; S4 represents the random field value at each grid node in the gridded geometric model obtained in step S3, so as to realize the visualization of the random field value of the rock and soil to be analyzed; In step S2, the conditional random field is formed by combining the unconditional random field theory and the Kriging method, and the method for forming the conditional random field is SL-SK-CRF, SL-OK-CRF, BH-SK-CRF or BH-OK-CRF; In step S2, the discriminant of the conditional random field is: whether the soil parameters have clear expected values, whether the attributes of all measuring points at the drilling location can be obtained, and whether the soil stratification is obvious.

2. The method for obtaining a geological model random field based on a conditional random field decision tree according to claim 1, characterized in that: In step S1, the drilling information includes the coordinates of the drilling point, whether it is a corner point, the height of the rock and soil layer and the rock and soil properties of the measuring point selected at each drilling point.

3. A method for obtaining a geological model random field based on a conditional random field decision tree as claimed in claim 2, characterized in that: The rock and soil properties are the friction angle, cohesion, modulus or density of the rock and soil layer where each measuring point is located.

4. A method for obtaining a geological model random field based on a conditional random field decision tree as claimed in claim 1 or 2, characterized in that: In step S4, the random field value is represented by a color in a color lookup table.

5. The method for obtaining a geological model random field based on a conditional random field decision tree according to claim 1, characterized in that: In step S3, the random field values ​​corresponding to each grid node in the gridded geometric model constitute a parameterized geological model.

6. The method for obtaining a geological model random field based on a conditional random field decision tree according to claim 1, characterized in that: In step S3, the meshed geometric model is obtained by using a Delaunay triangulated mesh generation algorithm.