Underground powerhouse initial stress field inversion analysis method

Abstract

The application provides an initial stress field inversion analysis method for an underground powerhouse, which comprises the following steps of: establishing a model diagram of a mountainous area of the underground powerhouse in FLAC 3D , establishing a plurality of ground stress measuring points, measuring ground stress data, and analyzing and screening representative sample points from the measured ground stress; calculating stress component actual lateral stress coefficients k i of each measuring point, introducing the fitted stress function into FLAC 3D to form an approximate paleo-stress field; introducing a generative adversarial network to fine-tune the paleo-lateral stress coefficient, optimize the present ground stress field, and meet the demand of engineering analysis. The uniform design can greatly reduce the number of tests and keep the training samples uniformly dispersed. After each uniform design test, the training of the GAN can be performed, the data of the measuring points are input into the GAN, and the optimized lateral stress coefficient can be obtained. The optimized paleo-stress field can be obtained by calculation, and the present ground stress field can be obtained after nonlinear elastic-plastic excavation simulation.

Description

Technical Field

[0001] The present invention relates to the field of rock and soil mechanics, and in particular to an inversion analysis method for an initial stress field of an underground powerhouse. Background Art

[0002] Quantitative research on geostress magnitude involves, on the one hand, field measurement and monitoring, and, on the other hand, inversion analysis and calculation of the initial geostress field based on limited observational data, to determine the initial geostress values ​​and distribution patterns for the entire project area. Field observation of geostress can provide relatively accurate geostress data at the measurement points. With the advancement of science and technology, observation equipment and methods are becoming increasingly advanced and accurate. However, the major drawback of field observation is its high cost. Therefore, accurately predicting and analyzing the geostress distribution across the entire project area based on limited measurement points is of paramount importance. Summary of the Invention

[0003] The main purpose of the present invention is to provide an inversion analysis method for the initial stress field of an underground powerhouse to solve the problems in the above-mentioned background technology.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is: the method steps are as follows:

[0005] S1. In FLAC 3D A model diagram of the underground powerhouse mountain area is established. The X-axis of the 3D numerical model is selected as the main powerhouse axis direction, the Y-axis is the upstream and downstream direction, and the Z-axis is vertically upward;

[0006] S2. Establish multiple geostress measurement points, measure geostress data, analyze the measured geostress data, and select representative sample points;

[0007] S3. Calculate the actual lateral stress coefficient k of the stress component of each measuring point by formula (1) i The fitted stress function can be imported into FLΑC 3D In the middle, an approximate paleo-stress field is formed;

[0008] ; (1);

[0009] Among them, α i and b i is the regression factor, k i is the lateral stress coefficient of the six stress components, H is the burial depth;

[0010] S4. Input the numerical simulation parameters of each rock mass into FLAC 3DIn the ancient model, the coordinates of the elements containing the rock mass and faults are output by the FISH language and then assigned to the model according to the parameters of the equivalent elements. The equivalent elements are numerically simulated using a transversely isotropic elastic constitutive model.

[0011] S5. In FLAC 3D In the FLAC, each unit area stores 6 stress components and each unit node stores unbalanced force. 3D Guidelines in the User Manual for performing geostress balance calculations on the model;

[0012] S6. By using the generative adversarial network (GAN), the optimized regression factor is further determined to obtain the optimized current initial geostress field.

[0013] Preferably, the specific steps of step S5 are as follows:

[0014] Α1. According to formula (1), the paleo-stress field can be calculated and imported into FLΑC 3D The rock mass is then input into the model;

[0015] A2. Apply normal constraints to the boundaries around and on the bottom of the ancient model;

[0016] A3. Execute the command on the model, i.e. calculate one step and then record the unbalanced points on the unit nodes;

[0017] A4. Apply a force opposite to the imbalance in step A3 to each unit node in the ancient model;

[0018] A5. Repeat steps A3 and A4 several times to make the stress components close enough to those in the ancient model.

[0019] Preferably, the method for calculating the equivalent unit includes dividing the rock mass into hexahedral units, and due to the existence of faults, some rock mass units are cut by faults, and these units contain both rock mass and faults, forming composite units;

[0020] A set of mechanical parameters is assigned to a unit. Therefore, the mechanical parameters of the composite unit with both rock mass and fault need to be equivalent. The local coordinate system (x´y´z´) is established on the fault plane of the composite unit, and the composite unit is simplified into a transversely isotropic equivalent unit with layered distribution.

[0021] H k is the unit layer thickness, which can be expressed as:

[0022] (2);

[0023] Where subscript k is equal to 1 or 2, representing the parameters of rock mass and fault, respectively, V kis the volume of the rock mass or fault, and A is the area of the rock mass-fault contact surface.

[0024] Preferably, for the z' direction, the deformation and stress of the equivalent element are:

[0025] (3).

[0026] where, , σ v1 and σ v2 are the stresses applied on the equivalent element, the rock mass and the fault in the z' direction, respectively, is the equivalent elastic modulus in the z' direction;

[0027] Based on equation (3), the equivalent elastic modulus in the z' direction is:

[0028] (4).

[0029] Preferably, for the x' and y' directions, assuming that the elongation of the rock mass and the fault in the equivalent element are equal, then

[0030] (5).

[0031] where, , ε h1 and ε h2 are the strains of the equivalent element, the rock mass and the fault in the x' and y' directions, respectively, , σ h1 and σ h2 are the stresses applied on the equivalent element, the rock mass and the fault in the x' and y' directions, respectively,

[0032] Equation (5) can be written as:

[0033] (6).

[0034] According to equations (5) and (6), the equivalent elastic modulus is:

[0035] (7).

[0036] In combination with equations (5) and (7), the equivalent Poisson's ratio can be expressed as

[0037] (8).

[0038] Preferably, in performing GAN, the measuring points of the underground cavern and its periphery and the measuring points near the boundary of the numerical model are selected as samples, and each regression factor a i or b iThe value range of α is between 0.5 and 1.5 times of itself. Multiple regression factors are used for uniform design experiments, so multiple regression models with different α can be designed. i or b i Paleo-stress field of the value;

[0039] Substitute each regression factor into equation (1) and equation (9) to calculate the stress component of each unit and import it into FLAC 3D Calculation is performed in

[0040] (9);

[0041] where σ1–σ6 are the stress components σ x , σ y , σ z , τ yz , τ zx and τ xy , k i is the lateral stress coefficient of the six stress components, H is the burial depth, and γ is the bulk density.

[0042] Preferably, after training with real data samples, the X and Y coordinates are input, and the present burial depth and lateral stress coefficient of the measuring point are input into the present stress field of the measuring point to obtain the regression factor of the paleo-stress field;

[0043] After GAN training, the regression factor of the optimized paleostress field can be predicted. Therefore, the optimized paleostress field can be obtained by formula (1). By mining the optimized paleostress field, the current geostress field can be obtained.

[0044] Preferably, the uniform design experiment includes designing different α i and b i Various combinations of values, α i and b i The measurement point parameters under different combinations are:

[0045] (10);

[0046] (11);

[0047] (12);

[0048] (13);

[0049] (14);

[0050] (15);

[0051] Where x and y are the coordinate vectors of the measured point in the horizontal plane, n is the number of measurement points, and H p is the depth vector of the current measured point, k j is the lateral stress coefficient matrix of the current measuring point, j is the number of uniform design tests, α j and b j is the stress component σ at the ancient measuring point x ,σ y ,σ z ,τ yz ,τ zx and τ xy The regression factor matrix of .

[0052] Preferably, after each uniform design experiment, GAN training is performed, and the real data sample input to GAN can be expressed as:

[0053] (16);

[0054] After the training is completed, the data of the measurement points are input into GAN to obtain the optimized α i and b i The optimized paleo-stress field can be calculated by formula (1), and then the present-day stress field can be obtained after nonlinear elastic-plastic excavation simulation.

[0055] The present invention provides an inversion analysis method for the initial stress field of an underground powerhouse. This method uses a generative adversarial network (GAN) to fine-tune the ancient lateral stress coefficients, optimizing the current geostress field to meet the needs of engineering analysis. The advantage of uniform design is that it can significantly reduce the number of tests and maintain a uniform distribution of training samples. After each uniform design experiment, the GAN can be trained. After training, the data from the measurement points are input into the GAN to obtain the optimized lateral stress coefficients. The optimized ancient geostress field can be calculated, and the current geostress field can be obtained after nonlinear elastic-plastic excavation simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The present invention will be further described below with reference to the accompanying drawings and examples:

[0057] Figure 1 is the relationship between the loss function and iteration of the GAN generator and discriminator of the present invention;

[0058] Figure 2 It is the calculation process of GAN of the present invention;

[0059] Figure 3 This is a schematic diagram of a fault-cut rock mass composite unit according to the present invention;

[0060] Figure 4 It is a schematic diagram of an equivalent unit of a fault-cut rock mass according to the present invention. DETAILED DESCRIPTION

[0061] like Figures 1 to 4 As shown in the figure, the inverse analysis method of the initial stress field of the underground powerhouse is 3D A model diagram of the underground powerhouse mountain area is established in the 3D numerical model. The X-axis of the 3D numerical model is selected as the axis direction of the main powerhouse, the Y-axis is the upstream and downstream direction, and the Z-axis is vertically upward. In this example, 10 ground stress measurement points are selected. Since the measured ground stress is obtained in the geodetic coordinate system, it is necessary to convert it into the coordinates in the model. According to the distribution characteristics of the ground stress at the measured points, representative sample points are selected and the actual lateral stress coefficient k of the six stress components of each measuring point is calculated. i Based on formula (1), we can get the current burial depth and k i The fitted stress function can be imported into FLΑC 3D In the middle, an approximate paleo-stress field is formed.

[0062] ; (1)

[0063] Among them, α i and b i is the regression factor, k i is the lateral stress coefficient of the six stress components, H is the burial depth;

[0064] In this example, the Z-axis range of the paleomodel is from the bottom elevation of 1970 m to the ancient surface. The upper six layers of the model are excavated paleostrata, which are used to simulate the erosion process of the valley. 3D The parameters of the paleomodel are shown in Table 1. The coordinates of the elements containing the rock mass and faults were exported from the FISH language. Then, the parameters of the equivalent elements were calculated and assigned to the model. The equivalent elements were numerically simulated using a transversely isotropic elastic constitutive model.

[0065] Table 1: Parameters of numerical simulation

[0066]

[0067] In FLΑC 3D In the model, each unit area stores 6 stress components and each unit node stores unbalanced force. If the model unit node does not apply unbalanced force in the X, Y, and Z directions, the stress in the unit area will change, and the model calculation results will deviate greatly from the actual results. In order to optimize the model in subsequent calculations, according to FLAC 3D Following the guidelines in the User Manual, the stress balance calculations for the model are performed as follows:

[0068] (1) According to formula (1), the paleo-stress field can be calculated and imported into FLAC 3DThen the rock mass parameters are input into the model;

[0069] (2) Apply normal constraints to the boundaries around and on the bottom of the ancient model;

[0070] (3) Execute the "step 1" command on the model, that is, calculate one step and then record the unbalanced force on the unit node;

[0071] (4) Apply a force opposite to the unbalanced force in step (3) to each unit node in the ancient model;

[0072] (5) Repeat (3) and (4) several times until the stress components are close enough to those in the ancient model.

[0073] The above process completes the equilibrium calculation of the approximate paleostress field. By excavating the ancient strata layer by layer, the present-day geostress field can be obtained, and the six stress components and lateral stress coefficients at each measured point can be derived. However, these lateral stress coefficients are derived from the approximate paleostress field and may contain unacceptable errors, requiring further determination of optimized regression factors.

[0074] Optimizing the current geostress field using a generative adversarial network (GAN) requires constructing a sufficient number of real-world samples for training. The initial geostress inversion for this engineering example aims to provide accurate geostress for underground cavern excavation, so measurement points within and around the cavern are selected as samples. Furthermore, boundary effects can occur in numerical excavation simulations, so measurement points near the boundaries of the numerical model are selected as samples.

[0075] Based on a large number of trial calculations, the floating range of the regression factors in formula (1) can be determined. i or b i The value range is between 0.5 and 1.5 times of itself. Take 12 regression factors as factors of uniform design test, and regard different regression factor values ​​as different levels. According to the principle of uniform design, there are 13 levels of regression factors. Then, levels 1 to 13 correspond to regression factors 0.52, 0.60, 0.68, 0.76, 0.84, 0.92, 1.00, 1.08, 1.16, 1.24, 1.32, 1.40, and 1.48 times, respectively. The regression factors of each test can be substituted into formulas (1) and (9) to calculate the stress components of each unit and imported into FLAC 3D Calculations are performed. In each test, the calculated stress components and lateral stress coefficients at the measured points are obtained. There are 8 valid measurement points, so 8 training samples can be created for each test, for a total of 13 × 8 = 104 samples.

[0076] After training with real data samples, the x and y coordinates are input, and the current burial depth and lateral stress coefficient of the measuring point are input into the current stress field of the measuring point to obtain the regression factor of the paleostress field. Then, the optimized paleostress field can be obtained using Equation (9). Figure 1 The relationship between the loss function and iteration of the generator and discriminator in GAN is shown. The generator G and the discriminator network D constantly compete to minimize the loss function of the two networks, indicating that the sample data output by the generator is stable. After GAN training, the regression factor of the optimized paleostress field can be predicted, so the optimized paleostress field can be obtained by formula (1). Mining the optimized paleostress field can obtain the current geostress field. The regression factors and calculated stress values ​​of the measurement points are listed in Tables 2 and 3. The relative error Δ of the measurement points can be calculated as:

[0077]

[0078] in, is the calculated stress component, is the measured stress component, is the 2-norm.

[0079] Table 2: Regression factors of paleo-stress field obtained using GAN and BP neural network

[0080]

[0081] Table 3: Stress component values ​​at each measuring point based on GAN and BP neural network

[0082]

[0083] In order to compare with the GAN inversion method, the stress component regression factors obtained by BP neural network are also listed in Tables 2 and 3. For both methods, σ x , σ y and σ z The relative error is less than 15%, which is better than τ yz , τ xz and τ xy More accurate. According to statistics, the error of ground stress measurement results can reach 25%-30%. The maximum relative error of the GAN measurement point is 17.46%, and the distribution pattern of the ground stress field is reasonable. For the BP method, the relative error of measurement point 8 reaches 28.4%, and the average relative error of GAN is smaller than that of the BP method, especially τ yz , τ xz and τ xy .

[0084] Example 2

[0085] According to the further illustration of the embodiment 1, based on the approximate paleo-stress field, the present stress field can be obtained by numerically simulating the layer-by-layer excavation of the paleo-stratum. Since the paleo-stress field is approximately estimated, the obtained present geostress field may have a large error. In view of this problem, the generative adversarial network (GAN) is introduced in this example to fine-tune the paleo-lateral stress coefficient, optimize the present geostress field, and meet the needs of engineering analysis.

[0086] The GAN is composed of a generator and a discriminator trained on an adversarial learning mechanism, as shown in FIG. 1; the goal of the GAN is to estimate and predict the distribution law of the obtained data, and to generate new data from the same distribution law by using the generator G, which converts the random variable z into new data that can be forged by continuously learning the probability distribution of the real data. The discriminator D is a binary classifier, which is used to distinguish whether the input data is real or generated data. The two networks compete with each other while enhancing during training, so that the two networks constitute a dynamic game process until a Nash equilibrium is reached. The generator and the discriminator can be designed in combination with the current deep neural network. The calculation process of the GAN can be summarized as a binary minimax game, and the objective function can be defined as: Figure 2

[0087]

[0088] wherein V(D, G) is the cross-entropy loss of two classifications, P dαtα(x) is the distribution of real data, P g(z) is the distribution of random variables, G (z) is the generator based on random variables, and E(·) represents the calculated expected value; when P dαtα = P g , a global optimal solution is reached, and the loss functions of the generator and the discriminator of the GAN are represented as log(D(G(z))) and log(D(x)) + log(1-D(G(z))).

[0089] Embodiment 3

[0090] ​Further explanation according to Example 1: The parameter differences between faults and rock masses are very large. If the influence of faults is not considered in the inversion of the current stress field, large errors will occur. Therefore, before simulating the excavation of ancient strata, the fault parameters should be assigned to the corresponding units. In numerical simulation analysis, they are often simplified to thin units, but faults and underground caves usually do not intersect in an orthogonal form. Due to the limited thickness of the faults and the complex intersection relationship, this method of handling faults may bring difficulties to the mesh generation of the model. In this example, the rock mass is divided into hexahedral units. Due to the existence of faults, some rock mass units are cut by faults. These units contain both rock mass and faults, forming composite units, such as Figure 3 As shown in Figure 2. For a unit, only one set of mechanical parameters can be assigned. Therefore, the mechanical parameters of the composite unit with both rock mass and fault need to be equivalent. The local coordinate system (x´y´z´) established on the fault plane of the composite unit simplifies the composite unit into a layered, transversely isotropic equivalent unit with parameters such as Figure 4 As shown. k is the unit layer thickness, which can be expressed as:

[0091] (2)

[0092] Where subscript k is equal to 1 or 2, representing the parameters of rock mass and fault, respectively, V k is the volume of the rock mass or fault, and A is the area of ​​the contact surface between the rock mass and the fault.

[0093] For the z´ direction, the deformation and stress of the equivalent element are:

[0094] (3)

[0095] in, , σ v1 and σ v2 are the stresses applied to the equivalent unit, rock mass and fault in the z' direction, is the equivalent elastic modulus in the z´ direction;

[0096] Based on formula (3), the equivalent elastic modulus in the z´ direction is:

[0097] . (4)

[0098] For the x' and y' directions, assuming that the elongation of the rock mass and the fault in the equivalent unit is equal, then

[0099] (5)

[0100] in, , ε h1 and ε h2are the strains of the equivalent unit, rock mass and fault in the x' and y' directions, respectively, , σ h1 and σ h2 are the stresses applied to the equivalent element, rock mass and fault in the x' and y' directions, respectively,

[0101] Formula (5) can be written as:

[0102] (6)

[0103] According to equations (5) and (6), the equivalent elastic modulus is:

[0104] (7)

[0105] Combining equations (5) and (7), the equivalent Poisson's ratio It can be expressed as

[0106] . (8)

[0107] Example 4

[0108] Further explanation is given in conjunction with Examples 1 and 2: When performing GAN, the measurement points in the underground cavern and its surroundings and the measurement points near the boundary of the numerical model are selected as samples. Each regression factor α i or b i The value range of is between 0.5 and 1.5 times of itself. Multiple regression factors are used for uniform design experiments, so multiple regression models with different α can be designed. i or b i Paleo-stress field of the value;

[0109] Substitute each regression factor into equation (1) and equation (9) to calculate the stress component of each unit and import it into FLAC 3D Calculation is performed in

[0110] (9)

[0111] where σ1–σ6 are the stress components σ x , σ y , σ z , τ yz , τ zx and τ xy , k i is the lateral stress coefficient of the six stress components, H is the burial depth, and γ is the bulk density.

[0112] After training with real data samples, the regression factor of the paleo-stress field can be obtained by inputting the X and Y coordinates, the present burial depth and lateral stress coefficient of the measuring point in the present stress field of the measuring point;

[0113] After GAN training, the regression factor of the optimized paleostress field can be predicted. Therefore, the optimized paleostress field can be obtained by formula (1). By mining the optimized paleostress field, the current geostress field can be obtained.

[0114] Uniform design experiments include designing different α i and b i Various combinations of values, α i and b i The measurement point parameters under different combinations are:

[0115] (10)

[0116] (11)

[0117] (12)

[0118] (13)

[0119] (14)

[0120] (15)

[0121] Where x and y are the coordinate vectors of the measured point in the horizontal plane, n is the number of measurement points, and H p is the depth vector of the current measured point, k j is the lateral stress coefficient matrix of the current measuring point, j is the number of uniform design tests, α j and b j is the stress component σ at the ancient measuring point x ,σ y ,σ z ,τ yz ,τ zx and τ xy The regression factor matrix of .

[0122] After each uniformly designed experiment, GAN is trained and the real data sample input to GAN can be expressed as:

[0123] (16)

[0124] After the training is completed, the data of the measurement points are input into GAN to obtain the optimized α i and b i The optimized paleo-stress field can be calculated by formula (1), and then the present-day stress field can be obtained after nonlinear elastic-plastic excavation simulation.

[0125] The above embodiments are only preferred technical solutions of the present application, and should not be regarded as a limitation of the present application. The protection scope of the present application should be the technical solutions recited in the claims, including equivalent replacement solutions of the technical features recited in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present application.

Claims

1. The inverse analysis method of the initial stress field of the underground powerhouse is as follows: S1. In FLAC 3D A model diagram of the underground powerhouse mountain area is established. The X-axis of the 3D numerical model is selected as the main powerhouse axis direction, the Y-axis is the upstream and downstream direction, and the Z-axis is vertically upward; S2. Establish multiple geostress measurement points, measure geostress data, analyze the measured geostress data, and select representative sample points; S3. Calculate the actual lateral stress coefficient k of the stress component of each measuring point by formula (1) i , the fitted stress function is imported into FLΑC 3D In the middle, an approximate paleo-stress field is formed; ； （1）； in, α i and b i is the regression factor, k i is the lateral stress coefficient of the six stress components, H is the burial depth; S4. Input the numerical simulation parameters of each rock mass into FLAC 3D In the ancient model, the coordinates of the elements containing the rock mass and faults are output by the FISH language and then assigned to the model according to the parameters of the equivalent elements. The equivalent elements are numerically simulated using a transversely isotropic elastic constitutive model. S5. In FLAC 3D In the FLAC, each unit area stores 6 stress components and each unit node stores unbalanced force. 3D Guidelines in the User Manual for performing geostress balance calculations on the model; S6. By using the generative adversarial network (GAN), the optimized regression factor is further determined to obtain the optimized current initial geostress field.

2. The inversion analysis method for the initial stress field of an underground powerhouse according to claim 1 is characterized by: The specific steps of step S5 are as follows: Α1. Calculate the paleo-stress field according to formula (1) and import it into FLΑC 3D The rock mass is then input into the model; A2. Apply normal constraints to the boundaries around and on the bottom of the ancient model; A3. Execute the command on the model, i.e. calculate one step and then record the unbalanced points on the unit nodes; A4. Apply a force opposite to the imbalance in step A3 to each unit node in the ancient model; A5. Repeat steps A3 and A4 several times to make the stress components close enough to those in the ancient model.

3. The inversion analysis method for the initial stress field of an underground powerhouse according to claim 1 is characterized by: The method of calculating equivalent units includes dividing the rock mass into hexahedral units. Due to the existence of faults, some rock mass units are cut by faults. These units contain both rock mass and faults, forming composite units. A set of mechanical parameters is assigned to a unit. Therefore, the mechanical parameters of the composite unit with both rock mass and fault need to be equivalent. The local coordinate system (x´y´z´) is established on the fault plane of the composite unit, and the composite unit is simplified into a transversely isotropic equivalent unit with layered distribution. H k is the unit layer thickness, which can be expressed as: （2）； Where subscript k is equal to 1 or 2, representing the parameters of rock mass and fault, respectively, V k is the volume of the rock mass or fault, and A is the area of ​​the contact surface between the rock mass and the fault.

4. The inversion analysis method for the initial stress field of an underground powerhouse according to claim 3 is characterized by: For the z´ direction, the deformation and stress of the equivalent element are: （3）； in, , σ v1 and σ v2 are the stresses applied to the equivalent unit, rock mass and fault in the z' direction, is the equivalent elastic modulus in the z´ direction; Based on formula (3), the equivalent elastic modulus in the z´ direction is: （4）。 5. The inversion analysis method for the initial stress field of an underground powerhouse according to claim 3 is characterized by: For the x' and y' directions, assuming that the elongation of the rock mass and the fault in the equivalent unit is equal, then （5）； in, , ε h1 and ε h2 are the strains of the equivalent unit, rock mass and fault in the x' and y' directions, respectively, , σ h1 and σ h2 are the stresses applied to the equivalent element, rock mass and fault in the x' and y' directions, respectively, Formula (5) is written as: （6）； According to equations (5) and (6), the equivalent elastic modulus is: （7）； Combining equations (5) and (7), the equivalent Poisson's ratio It can be expressed as: （8）。 6. The inverse analysis method for the initial stress field of an underground powerhouse according to claim 1 is characterized by: When performing GAN, the measurement points in the underground cavern and its surroundings and the measurement points near the boundary of the numerical model are selected as samples. Each regression factor α i or b i The value range of is between 0.5 and 1.5 times of itself. Multiple regression factors are used to conduct uniform design experiments, thus designing multiple regression factors with different α i or b i Paleo-stress field of the value; Substitute each regression factor into equation (1) and equation (9) to calculate the stress component of each unit and import it into FLAC 3D Calculation is performed in （9）； Where σ1 –σ6 is the stress component σ x , σ y , σ z , τ yz , τ zx and τ xy , k i is the lateral stress coefficient of the six stress components, H is the burial depth, and γ is the bulk density.

7. The inverse analysis method for the initial stress field of an underground powerhouse according to claim 6 is characterized by: After training with real data samples, the X and Y coordinates are input, and the present burial depth and lateral stress coefficient of the measuring point are input into the present stress field of the measuring point to obtain the regression factor of the paleo-stress field. After GAN training, the regression factor of the optimized paleostress field is predicted. Therefore, the optimized paleostress field is obtained by formula (1). By mining the optimized paleostress field, the current geostress field can be obtained.

8. The inverse analysis method for the initial stress field of an underground powerhouse according to claim 6 is characterized by: Uniform design experiments include designing different α i and b i Various combinations of values, α i and b i The measurement point parameters under different combinations are: （10）； （11）； （12）； （13）； （14）； （15）； Where x and y are the coordinate vectors of the measured point in the horizontal plane, n is the number of measurement points, and H p is the depth vector of the current measured point, k j is the lateral stress coefficient matrix of the current measuring point, j is the number of uniform design tests, α j and b j is the stress component σ at the ancient measuring point x ,σ y ,σ z ,τ yz ,τ zx and τ xy The regression factor matrix of .

9. The inverse analysis method for the initial stress field of an underground powerhouse according to claim 8 is characterized by: After each uniformly designed experiment, GAN is trained and the real data sample input to GAN is expressed as: （16）； After the training is completed, the data of the measurement points are input into GAN to obtain the optimized α i and b i The optimized paleo-stress field is calculated by formula (1), and then the present-day stress field is obtained after nonlinear elastic-plastic excavation simulation.

Images