Initial crustal stress field inversion method

Abstract

The invention relates to the field of crustal stress inversion, in particular to an initial crustal stress field inversion method, which comprises the following steps of: acquiring an actual measurement value of crustal stress of each field monitoring point obtained according to borehole stress measurement; screening the data of the crustal stress measured value to obtain the screened crustal stress measured value, and determining a to-be-inverted parameter; and establishing a BP neural network model according to the screened crustal stress measured values, and performing iterative optimization in combination with a genetic algorithm to obtain an initial crustal stress field of crustal stress inversion. The method is applied to the inversion of the initial crustal stress field of the rock mass, has the characteristics of high data utilization rate, stable iteration process, fast convergence and the like, can remarkably reduce the influence caused by a small amount of sample data and human subjectivity, improves the reliability of a result, and is suitable for large-scale popularization and application. Related engineering technicians can simply and conveniently obtain the accurate initial crustal stress field of the area where the engineering is located, and the method has important application value for underground engineering stability evaluation and underground engineering design and construction.

Description

Technical Field

[0001] The invention relates to the technical field of geostress inversion, in particular to an initial geostress field inversion method. Background Art

[0002] It is an inevitable trend of the development of earth science to advance into the deep earth, explore the mysteries of the earth, and tap deep resources. The natural stress field generated in the rock mass by its own weight and other effects is called the geostress field. The distribution of the geostress field is closely related to the stability of rock mass engineering. The study of geostress is the basis for evaluating the stability, designing and constructing underground engineering. Geostress not only affects the stability of engineering rock mass in geological areas, but also has direct or indirect effects on the design and construction of various underground engineering projects.

[0003] Among the existing geostress inversion technologies, the stress measurement method directly measures underground stress through special instruments, and the data reliability is high, but the equipment is expensive, the installation and maintenance costs are high, and the measurement can only be carried out at specific locations, and the spatial coverage is limited. The pore water pressure inversion method obtains data through conventional drilling and monitoring of pore water pressure, which is applicable to a variety of geological conditions. Its inversion results rely on model assumptions and have large errors. As a new means of studying geostress, numerical calculation has been widely used in the calculation and analysis of three-dimensional geostress fields of underground engineering in recent years. However, it requires high-performance computing resources, the accuracy of the model depends on the quality of the input data and the model assumptions, and it is highly sensitive to the model parameters. It can be seen that the traditional initial geostress field inversion method has problems such as poor economy, high regional limitations, complex calculations and large uncertainties. Summary of the invention

[0004] In order to quickly, accurately and effectively obtain the initial geostress field in the rock mass, the present invention proposes an initial geostress field inversion method.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An initial geostress field inversion method comprises the following steps:

[0007] Obtaining measured values ​​of ground stress at each on-site monitoring point obtained from borehole stress measurement;

[0008] Screening the data of measured values ​​of ground stress to obtain screened measured values ​​of ground stress and determining the parameters to be inverted;

[0009] A BP neural network model is established based on the screened measured values ​​of geostress, and it is iteratively optimized in combination with a genetic algorithm to obtain the initial geostress field for geostress inversion. Combining the BP neural network model with the genetic algorithm can fully utilize the advantages of both and achieve global high-precision optimization of network weights.

[0010] As a further solution of the present invention: the borehole stress measurement is obtained by an aperture strain measurement method, which is a method of calculating the three-dimensional stress state in the rock mass around the borehole by measuring the strain value of the borehole wall.

[0011] As a further solution of the present invention: the method for screening the data of the measured ground stress values ​​is monitoring credibility test, normality test and related homogeneity test, which can eliminate the human subjectivity in the layout of monitoring points and information extraction methods.

[0012] As a further solution of the present invention: the monitoring credibility test adopts McDonald's ω coefficient monitoring credibility test, the normality test adopts SW normality test, and the related homogeneity test adopts t test.

[0013] As a further solution of the present invention: the initial geostress field inversion method also includes the following steps: substituting the initial geostress field obtained by geostress inversion into the numerical calculation software for forward calculation to obtain the geostress calculation value of each on-site monitoring point, performing error analysis between the measured geostress value and the geostress calculation value, and analyzing the reliability of the initial geostress field.

[0014] As a further solution of the present invention: the specific steps of substituting the initial geostress field obtained by geostress inversion into the numerical calculation software for forward calculation are as follows: using Rhinoceros software to establish a three-dimensional geological body model and gridding it to obtain a gridded three-dimensional geological body model, importing the gridded three-dimensional geological body model into CASRock software and substituting it into the initial geostress field for forward calculation.

[0015] As a further solution of the present invention: the formula of the BP neural network model combined with the genetic algorithm is as follows: p=(p1,p2,...,p n ) is the input node expression of the neural network,

[0016] where D = (d1, d2, ..., d n ) is the output node expression of the neural network, NN(n,h1,h2,...,h p ,m) is the established multi-layer neural network structure, n,h1,h2,...,h p ,m is the input layer F x , hidden layer F1...F p And the input layer F yThe number of nodes is the nonlinear mapping between the boundary conditions (displacement boundary, stress boundary, etc.) of the BP neural network and the stress values ​​of the measured geostress points in the model. The nonlinear relationship between the boundary conditions to be inverted (whose initial range value is given by the multivariate linear regression result of the elastic calculation of the geostress field) and the stress value is mapped using a set of neural networks (n, h1, h2, ..., h p ,m) to describe.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: compared with the traditional technology, the present invention is applied to the inversion of the initial geostress field of rock mass, which not only has the characteristics of high data utilization, stable iterative process, fast convergence, etc., but also can significantly reduce the influence of a small amount of sample data and human subjectivity, improve the reliability of the results, and enable relevant engineering and technical personnel to simply and conveniently obtain the accurate initial geostress field of the project area, which has important application value for the stability evaluation, design and construction of underground projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The present invention is a flowchart of an initial geostress field inversion method according to an embodiment of the present invention.

[0019] Figure 2 This is a BP neural network model diagram in an embodiment of the present invention.

[0020] Figure 3 It is a trend diagram of the level changes of various influencing factors of the genetic algorithm in the embodiment of the present invention.

[0021] Figure 4 This is a graph of error changes in iterative analysis of the genetic algorithm in an embodiment of the present invention.

[0022] Figure 5 This is a three-dimensional geological model diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Reference Figure 1-Figure 5 , an initial geostress field inversion method, comprising the following steps:

[0025] 1) Obtain the measured value of the ground stress at the on-site monitoring point

[0026] This embodiment relies on the GS hydropower station as an actual engineering project. On the basis of detailed survey of the site address area, 10 groups of rock mass spatial in-situ stress measurements were carried out at different positions of the PD4 exploration adit and its branch tunnels in the right bank underground powerhouse using the aperture strain measurement method. They are numbered P10-P19, and the tunnel entrance elevation is 2093.54m. The stress components and principal stress values ​​of each spatial in-situ stress measurement point are measured as shown in Table 1 and Table 2 respectively:

[0027] Table 1 Stress distribution table of spatial ground stress measurement points

[0028]

[0029] Table 2 Calculation value table of principal stress at spatial monitoring points

[0030]

[0031] Note: (1) The inclination angle α is positive in the horizontal plane; (2) β is the azimuth angle.

[0032] 2) Screen the measured values ​​of geostress and determine the parameters to be inverted

[0033] Due to human subjectivity in the layout of monitoring instruments, monitoring points and information extraction methods, it is necessary to conduct reliability analysis on the data of on-site monitoring points. Monitoring credibility analysis, normality test, correlation and homogeneity test are effective means to analyze the reliability of sample data. McDonald's ω reliability coefficient was selected for reliability analysis. This method has relatively low requirements for sample size and data distribution. It is easy to interpret and apply while providing accurate reliability estimates. The value range of the ω coefficient is between 0 and 1. Similar to other reliability indicators, the closer its value is to 1, the higher the reliability. Values ​​greater than 0.8 are generally considered to have high reliability, and values ​​less than 0.5 are considered to have low reliability. As a prerequisite for many statistical methods, normality test can help determine whether data are suitable for specific statistical methods, thereby ensuring the reliability of statistical inference. The t-test was selected for correlation and homogeneity test. The t-test can effectively compare the mean differences between two groups of data when the sample size is small. When analyzing the t-test, the p-value is analyzed first. If this value is less than 0.05, it means that there is a difference, otherwise there is no difference. Based on this, the reliability analysis of the research data obtained the following results:

[0034] Monitoring credibility test, see Table 3:

[0035] Table 3 Monitoring credibility table

[0036]

[0037] The calculated McDonald's ω coefficient values ​​for the in-situ stress monitoring data used for the study listed in Table 3 are higher than 0.8, indicating that the data have high reliability quality and can be used for further analysis.

[0038] Normality test, see Table 4:

[0039] Table 4 Normality test results

[0040]

[0041] The sample size of the research data in Table 4 is less than 50, and the SW test is used. Specifically, the sample of P13 shows significance (p<0.05), which means that the data of P13 does not have normal characteristics. The absolute value of the kurtosis of the data of the sample of P10 is less than 10 and the absolute value of the skewness is less than 3, which can be basically accepted as a normal distribution. The data of the remaining samples did not show significance (p>0.05), which means that they have normal characteristics. After analyzing the results of the comprehensive normal test result table, it can be seen that the data of measuring point No. 13 is abnormal (p<0.01), and the data of the remaining measuring points are in line with the normal characteristics;

[0042] Correlation homogeneity test:

[0043] Using paired t-test, a total of one set of paired data, the calculation results showed that there was no significant difference between measuring points 18 and 19 (p>0.05), but according to the sensitivity distribution of measuring point error to parameter changes, the sensitivity of No. 18 was relatively greater than that of No. 19.

[0044] Based on the above results, the data from measuring points 13 and 19 were eliminated, and the data from eight on-site geostress monitoring points, 10, 11, 12, 14, 15, 16, 17, and 18, were finally determined as the calculation data for geostress inversion.

[0045] 3) Establish a nonlinear mapping relationship between the parameters to be inverted and the stress detection information

[0046] like Figure 2 As shown in Figure 1, based on the selected geostress measurement point data, a BP neural network model was established, and iterative optimization analysis was performed in combination with a genetic algorithm. After 21 iterative calculations, the parameters tended to be stable, and the iterative error was reduced to about 6.54% and was relatively stable. Figure 3 and Figure 4 The result of the 21st iteration is taken as the initial boundary condition obtained by the inversion of geostress, as shown in Table 5:

[0047] Table 5 In-situ stress boundary obtained by back analysis

[0048]

[0049] 4) Verify the anti-analysis results

[0050] like Figure 5 As shown, a numerical model of the dam site area of ​​GS Hydropower Station is established, with a size range of X*Y=1316m*3500m in length and width, and about 1855m in height. Rhinoceros software is combined with CASRock software to establish a three-dimensional geological model with Rhinoceros software, and the meshed model is imported into CASRock software for numerical calculation. The number of model units is about 3.29 million, the number of nodes is about 640,000, the total number of materials is 14, the average unit size is 6m*6m*6m, and the structural resolution is about 1m. The material parameter table is shown in Table 6:

[0051] Table 6 Physical and mechanical parameters of rock properties

[0052]

[0053] Substitute the boundary conditions obtained from the 21st iteration into the model forward calculation to obtain the calculated values ​​of the in-situ stress at each stress measuring point. Compare the errors with the measured values. The final calculation results of the in-situ stress measuring point values ​​are shown in Table 7:

[0054] Table 7 Earth stress measurement error results

[0055]

[0056] In the process of inversion calculation of the geostress field, there may be an error of 20% to 30% between the calculated geostress value and the measured geostress value. The principal stress value of the geostress measuring point within the error range can be considered reasonable. The results in Table 7 reveal that the calculation error of the principal stress value of most geostress measuring points in this inversion calculation is below 30%. In general, based on the comparison results of the measured geostress values ​​and the inversion values ​​at the above-mentioned stress measuring point locations, the geostress field distribution results of the GS hydropower station area obtained by this inversion can be considered reasonable and reliable, and the calculated initial geostress field can be applied to the engineering in this geological area for numerical calculation of design and construction.

[0057] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for inversion of initial geostress field, characterized in that: The following steps are involved: Obtaining measured values ​​of ground stress at each on-site monitoring point obtained from borehole stress measurement; Screening the data of measured values ​​of ground stress to obtain screened measured values ​​of ground stress and determining the parameters to be inverted; A BP neural network model is established based on the screened measured values ​​of geostress, and it is iteratively optimized in combination with a genetic algorithm to obtain the initial geostress field for geostress inversion.

2. The initial geostress field inversion method according to claim 1, characterized in that: The borehole stress measurement is obtained through the aperture strain measurement method, which is a method of calculating the three-dimensional stress state in the rock mass around the borehole by measuring the strain value of the borehole wall.

3. The initial geostress field inversion method according to claim 1 or 2, characterized in that: The methods for screening the data of measured ground stress values ​​are monitoring credibility test, normality test and related homogeneity test.

4. The initial geostress field inversion method according to claim 3, characterized in that: The monitoring credibility test adopts McDonald's ω coefficient monitoring credibility test, the normality test adopts SW normality test, and the related homogeneity test adopts t test.

5. The initial geostress field inversion method according to claim 1, characterized in that: The formula of the BP neural network model combined with the genetic algorithm is as follows: ,in is the input node expression of the neural network, is the output node expression of the neural network, It is a multi-layer neural network structure established. For the input layer , hidden layer and the input layer The number of nodes.

6. The initial geostress field inversion method according to claim 1, characterized in that: The following steps are also included: The initial geostress field obtained by geostress inversion is substituted into the numerical calculation software for forward calculation to obtain the calculated geostress value of each on-site monitoring point, and the error analysis is performed between the measured geostress value and the calculated geostress value.

7. The initial geostress field inversion method according to claim 6, characterized in that: The specific steps of substituting the initial geostress field obtained by geostress inversion into the numerical calculation software for forward calculation are as follows: using Rhinoceros software to establish a three-dimensional geological body model and grid it to obtain a gridded three-dimensional geological body model, importing the gridded three-dimensional geological body model into CASRock software and substituting it into the initial geostress field for forward calculation.

Images