Mechanism and data-based ground stress inversion method, device, equipment and medium

Abstract

The application relates to a mechanism and data-based in-situ stress inversion method, device, equipment and medium, comprising the following steps: S1, obtaining measured in-situ stress data and sampling depth of each sampling point in a target area; S2, obtaining specific values of fitting parameters in a change relationship formula according to the measured in-situ stress data and the sampling depth, and constructing a target function based on the change relationship formula with known fitting parameters; S3, predicting first predicted in-situ stress data of each sampling point through an initial model, and calculating a target value by using the target function based on the first predicted in-situ stress data and the measured in-situ stress data of each sampling point; when the target value is greater than a preset threshold value or the number of iterations is less than a preset number, optimizing the initial model, repeating S3 based on the optimized initial model, until the target value is less than or equal to the preset threshold value, or the number of iterations is equal to the preset number, obtaining a prediction model, and inverting a stress distribution field of the target area through the prediction model. The method can have high precision and high efficiency.

Description

Technical Field

[0001] This application relates to the field of geostress inversion analysis technology, and in particular to a geostress inversion method, apparatus, equipment and medium based on mechanism and data. Background Technology

[0002] In-situ stress is a key factor affecting the stability analysis of geological engineering projects. Accurately obtaining the three-dimensional in-situ stress field of the engineering area is crucial for engineering design and disaster early warning.

[0003] Currently, methods for obtaining geostress fields mainly fall into two categories: The first is regression analysis methods based on field measured data, including multivariate nonlinear regression, support vector regression, and neural networks. These methods heavily rely on the number and distribution of measuring points, and are prone to overfitting when data is scarce or unevenly distributed. Furthermore, they struggle to reflect the controlling effect of geological structures (such as faults) on the stress field. The second category is numerical simulation methods based on mechanical principles, which involve inversion by applying boundary conditions. While this method can consider geological structures, it suffers from low inversion efficiency, high computational cost, and the inversion results are extremely sensitive to the choice of boundary conditions.

[0004] Existing technologies show that single methods have significant limitations, easily leading to inversion results that are either highly regular overall but distorted locally, or have good local fitting but unclear overall physical meaning. Therefore, there is an urgent need for a geostress inversion method that combines high efficiency and high accuracy. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, apparatus, equipment, and medium for geostress inversion based on mechanism and data that combines high accuracy and high efficiency to address the aforementioned technical problems.

[0006] A mechanism- and data-based geostress inversion method, the method comprising: S1. Obtain the measured ground stress data and sampling depth of each sampling point in the target area. The measured ground stress data includes measured horizontal stress and measured vertical stress. Preferably, the measured horizontal stress includes the maximum horizontal principal stress and the minimum horizontal principal stress; Preferably, the sampling depth refers to the depth of the sampling point from the ground. S2. Based on the measured ground stress data and the sampling depth, obtain the specific values ​​of the fitting parameters in the variation formula, and construct the objective function based on the variation formula with known fitting parameters. The relationship is a formula that characterizes the actual ratio between the measured horizontal stress and the measured vertical stress at the sampling point as a function of the sampling depth. This represents the actual ratio at sampling point i. This represents the prediction ratio obtained based on the first predicted geostress data from sampling point i, where n is the number of sampling points. W is the penalty factor, and W is the mean square error between the first predicted geostress data and the measured geostress data at the sampling point. Preferably, the variation relationship is as follows: ,in The measured horizontal stress at the sampling point. denoted as the measured vertical stress at the sampling point, D as the depth of the sampling point, and a and b as known fitting parameters in the variation formula; Preferably, the prediction ratio is the result of dividing the predicted horizontal stress in the first predicted geostress data of the sampling point by the predicted vertical stress; Preferably, the mean square error is the mean square error of the measured horizontal stress and the predicted horizontal stress at each sampling point, or the mean square error of the measured vertical stress and the predicted vertical stress at each sampling point. S3. Predict the first predicted ground stress data of each sampling point using the initial model, and calculate the target value M using the objective function based on the first predicted ground stress data and the measured ground stress data of each sampling point. Preferably, the initial model is one of the following: a deep learning model based on physical constraints, a regression prediction model based on machine learning, or a nonlinear inversion model; S4. When the target value is greater than a preset threshold or the number of iterations is less than a preset number, optimize the model parameters of the initial model based on the target value, and repeat step S3 based on the optimized initial model until the target value is less than or equal to the preset threshold, or the number of iterations is equal to the preset number, to obtain a trained prediction model, so as to obtain the stress distribution field of the target region through the prediction model.

[0007] Preferably, the penalty coefficient c and loss function parameters in the initial model are optimized using the particle swarm optimization algorithm. and kernel function parameters The penalty coefficient, loss function parameters, and kernel function parameters are all model parameters of the initial model.

[0008] In one embodiment, the measured horizontal stress includes the maximum horizontal principal stress and the minimum horizontal principal stress, and the process of solving the fitting parameters in the variation relationship in step S2 includes: Based on the measured ground stress data and sampling depth of each sampling point, a fitting operation is performed to obtain the variation relationship. The specific values ​​of the fitting parameters, The maximum horizontal principal stress at the sampling point. The minimum horizontal principal stress at the sampling point, Let be the vertical stress at the sampling point, D be the depth of the sampling point, and a and b be the fitting parameters of the variation relationship.

[0009] In one embodiment, obtaining the stress distribution field of the target region through the prediction model inversion includes: Obtain a geological survey report for the target area, the geological survey report including topographic contour data, borehole columnar section data, geological profile data and fault attitude data for the target area; The free surface at the top of the three-dimensional numerical model is constructed based on the topographic contour data; the rock strata interface of the three-dimensional numerical model is fitted by an interpolation algorithm based on the borehole columnar section data and the geological profile data; and the fault plane and fault influence zone entities of the three-dimensional numerical model are constructed based on the fault attitude data. Based on the rock strata interface, the free surface, the fault plane and the fault influence zone entity, a three-dimensional numerical model of the target area is constructed, and the three-dimensional numerical model is meshed. The prediction model outputs the second predicted geostress data for each grid node after the three-dimensional numerical model is divided. Based on the second predicted geostress data of each grid node, the stress distribution field of the target area is obtained.

[0010] In one embodiment, the step of outputting the second predicted geostress data for each grid node after the three-dimensional numerical model is divided by the prediction model includes: The spatial coordinates of each grid node after the three-dimensional numerical model is divided are input into the prediction model to obtain the initial geostress data of each grid node; The initial geostress data are optimized to obtain the second predicted geostress data for each grid node.

[0011] In one embodiment, optimizing the initial geostress data to obtain the second predicted geostress data for each of the grid nodes includes: The initial stress step field, which is formed by the initial geostress data of each of the grid nodes, is used as the initial stress condition, and the initial stress condition is applied to the three-dimensional numerical model. Based on the three-dimensional numerical model with the applied initial stress step field and the constitutive model of the target region, stress balance calculations are performed using finite difference software to obtain the second predicted geostress data for each of the grid nodes.

[0012] In one embodiment, the coordinate system of the spatial coordinates of the grid node is consistent with the coordinate system of the sampling point, and the method further includes: Determine the target node with consistent spatial coordinates from the sampling points and the grid nodes; Based on the measured ground stress data of the target node and the second predicted ground stress data, the stress deviation is calculated; Based on the comparison results between the stress deviation and the preset deviation threshold, the reliability of the second predicted geostress data of each grid node is determined.

[0013] Preferably, the stress deviation is the sum of one or more of the differences in maximum principal stress, minimum principal stress, and vertical stress at the target node; the difference in maximum principal stress is the difference between the measured and predicted maximum horizontal principal stress at the target node; the difference in minimum principal stress is the difference between the measured and predicted minimum horizontal principal stress at the target node; and the difference in vertical stress is the difference between the measured and predicted vertical stress at the target node.

[0014] Preferably, if there are multiple target nodes, the stress deviation of each target node is calculated separately, and the total number of nodes that are greater than a preset first deviation threshold is counted. When the total number is less than the preset number, the second predicted geostress data of each grid node is determined to be reliable.

[0015] Preferably, if there are multiple target nodes, the stress deviation of each target node is calculated separately, and the stress deviations of all target nodes are summed to obtain the total stress deviation; when the total stress deviation is less than or equal to a preset second deviation threshold, the second predicted geostress data of each grid node is determined to be reliable. The first deviation threshold and the second deviation threshold are both deviation thresholds.

[0016] In one embodiment, the measured horizontal stress includes the maximum horizontal principal stress and the minimum horizontal principal stress, the first predicted geostress data includes the predicted maximum horizontal principal stress, the predicted minimum horizontal principal stress, and the predicted vertical stress, and the mean square error is... ; in, The predicted horizontal maximum principal stress at sampling point i, The predicted horizontal minimum principal stress at sampling point i, The predicted vertical stress at sampling point i, The maximum horizontal principal stress at sampling point i, Let i be the minimum horizontal principal stress. Let be the vertical stress at sampling point i.

[0017] A mechanism- and data-based geostress inversion device for performing the above method, the device comprising: The data acquisition unit is used to acquire the measured ground stress data and sampling depth of each sampling point in the target area. The measured ground stress data includes measured horizontal stress and measured vertical stress. The objective function construction unit is used to obtain the specific values ​​of the fitting parameters in the variation relationship based on the measured in-situ stress data and the sampling depth, and to construct the objective function based on the variation relationship with the known fitting parameters. The relationship is a formula that characterizes the actual ratio between the measured horizontal stress and the measured vertical stress at the sampling point as a function of the sampling depth. This represents the actual ratio at sampling point i. This represents the prediction ratio obtained based on the first predicted geostress data from sampling point i, where n is the number of sampling points. W is the penalty factor, and W is the mean square error between the first predicted geostress data and the measured geostress data at the sampling point. The target value calculation unit is used to predict the first predicted geostress data of each of the sampling points through an initial model, and to calculate the target value M using the objective function based on the first predicted geostress data and the measured geostress data of each of the sampling points. The stress inversion unit is used to optimize the model parameters of the initial model when the target value is greater than a preset threshold or the number of iterations is less than a preset number, and repeat step S3 based on the optimized initial model until the target value is less than or equal to the preset threshold, or the number of iterations is equal to the preset number, to obtain a trained prediction model, so as to invert the stress distribution field of the target region through the prediction model.

[0018] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of the method described above.

[0019] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0020] The aforementioned mechanism- and data-based geostress inversion method, apparatus, equipment, and medium acquire measured geostress data and sampling depths at various sampling points in the target area. The measured geostress data includes measured horizontal and vertical stresses. Based on the measured geostress data and sampling depths, the specific values ​​of the fitting parameters in the variation equation are obtained, and an objective function is constructed based on the variation equation with known fitting parameters. This allows the distribution law of geostress in the target area to be introduced as a physical constraint into the objective function, so as to avoid the defect of the initial model ignoring the physical law during the training process, making the trained prediction model more reliable, thereby improving the accuracy of the stress distribution field of the target area obtained by inversion through the prediction model, and improving the efficiency of obtaining the stress distribution field. Attached Figure Description

[0021] Figure 1 This is a diagram illustrating the application environment of a mechanism- and data-based geostress inversion method in one embodiment. Figure 2 This is a flowchart illustrating a mechanism- and data-based geostress inversion method in one embodiment. Figure 3 This is a schematic diagram of a three-dimensional numerical model of the target region in one embodiment; Figure 4 This is a schematic diagram of the stress distribution field formed by the optimized second predicted stress data in one embodiment; Figure 5 This is a schematic diagram of the initial stress distribution field in one embodiment; Figure 6 This is a schematic diagram illustrating the difference in maximum principal stress among the target nodes in one embodiment; Figure 7 This is a schematic diagram of the overall process of the geostress inversion method based on mechanism and data in another embodiment; Figure 8 This is a structural block diagram of a geostress inversion device based on mechanism and data in one embodiment; Figure 9 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0023] The mechanism- and data-based geostress inversion method provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 102 interacts with server 104 via a wired / wireless channel. The data storage system can store the data that server 104 needs to process. S1: Obtain the measured ground stress data and sampling depth of each sampling point in the target area. The measured ground stress data includes measured horizontal stress and measured vertical stress. Based on the measured ground stress data and sampling depth, obtain the specific values ​​of the fitting parameters in the variation formula, and construct the objective function based on the variation formula with known fitting parameters. The relationship is a formula that characterizes the actual ratio between the measured horizontal stress and the measured vertical stress at the sampling point as a function of the sampling depth. This represents the actual ratio at sampling point i. This represents the prediction ratio obtained based on the first predicted geostress data from sampling point i, where n is the number of sampling points. W is the penalty factor, and W is the mean square error between the first predicted ground stress data and the measured ground stress data of the sampling points. S3: Predict the first predicted ground stress data of each sampling point using the initial model, and calculate the target value M using the objective function based on the first predicted ground stress data and the measured ground stress data of each sampling point. S4: When the target value is greater than a preset threshold or the number of iterations is less than a preset number, optimize the model parameters of the initial model based on the target value, and repeat step S3 based on the optimized initial model until the target value is less than or equal to the preset threshold, or the number of iterations is equal to the preset number, to obtain the trained prediction model, which can then be used to invert the stress distribution field of the target area. The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, etc. The server 104 can be a single server, a server cluster consisting of multiple servers, or a cloud computing center consisting of multiple servers.

[0024] In one embodiment, such as Figure 2 As shown, a geostress inversion method based on mechanism and data is provided, which is then applied to... Figure 1 Taking server 104 as an example, the following steps are included: S1. Obtain the measured ground stress data and sampling depth of each sampling point in the target area. The measured ground stress data includes the measured horizontal stress and the measured vertical stress. The target area includes both the surface and the subsurface within a specific region.

[0025] Measured in-situ stress data refers to the in-situ stress data actually measured at the sampling points. Measured horizontal stress includes the maximum and minimum horizontal principal stresses.

[0026] The sampling points were not distributed throughout the entire target area, but only in a portion of it.

[0027] Sampling depth refers to the depth of the sampling point from the ground.

[0028] In one embodiment, the measured ground stress data and sampling depth of some sampling points in the target area are shown in Table 1. Here, elevation refers to the height of the sampling point relative to sea level.

[0029] Table 1 Measured ground stress data

[0030] S2. Based on the measured ground stress data and sampling depth, obtain the specific values ​​of the fitting parameters in the variation relationship, and construct the objective function based on the variation relationship of the known fitting parameters; Wherein, the objective function is The relationship is a formula that characterizes the actual ratio between the measured horizontal stress and the measured vertical stress at the sampling point as a function of the sampling depth. This represents the actual ratio at sampling point i. This represents the prediction ratio obtained based on the first predicted geostress data from sampling point i, where n is the number of sampling points. W is the penalty factor, and W is the mean square error between the first predicted geostress data and the measured geostress data at the sampling point. The change relationship can be: ,in The measured horizontal stress at the sampling point. Let be the measured vertical stress at the sampling point, D be the depth of the sampling point, and a and b be the fitting parameters in the variation formula. Based on the measured ground stress data and sampling depth, the specific values ​​of the fitting parameters in the variation formula can be obtained.

[0031] The prediction ratio is obtained by dividing the predicted horizontal stress in the first predicted geostress data of the sampling point by the predicted vertical stress.

[0032] The specific values ​​of the fitting parameters in the variation relationship can be obtained by fitting using the least squares method.

[0033] The formula relating the actual ratio of measured horizontal stress to measured vertical stress at a sampling point to the sampling depth can be used to describe the macroscopic laws governing the distribution of geostress in the target area.

[0034] In a specific application scenario, the specific value of the fitting parameter a obtained from the measured ground stress data and sampling depth of each sampling point in Table 1 is 218.65, and the specific value of the fitting parameter b is 1.

[0035] Furthermore, the mean square error can be the mean square error of the measured horizontal stress and the predicted horizontal stress at each sampling point, or it can be the mean square error of the measured vertical stress and the predicted vertical stress at each sampling point.

[0036] S3. Predict the first predicted ground stress data of each sampling point using the initial model, and calculate the target value M using the objective function based on the first predicted ground stress data and the measured ground stress data of each sampling point. The initial model is used to predict the geostress data at the sampling points. The initial model includes, but is not limited to, physical constraint-based deep learning models, machine learning-based regression prediction models, and nonlinear inversion models. Nonlinear inversion models can be established using the SVR (Support Vector Regression) method, and the kernel function of the nonlinear inversion model can be a radial basis function (RBF).

[0037] Specifically, the spatial coordinates of each sampling point are input into the initial model, the first predicted geostress data of each sampling point are output, and the first predicted geostress data and the measured geostress data of each sampling point are substituted into the objective function to obtain the objective value of the objective function.

[0038] S4. When the target value is greater than the preset threshold or the number of iterations is less than the preset number of iterations, optimize the model parameters of the initial model based on the target value, and repeat step S3 based on the optimized initial model until the target value is less than or equal to the preset threshold, or the number of iterations is equal to the preset number of iterations, to obtain the trained prediction model, so as to obtain the stress distribution field of the target area through the prediction model inversion.

[0039] The condition "until the target value is less than or equal to a preset threshold, or the number of iterations is equal to a preset number, to obtain the trained prediction model" means that when the target value is less than or equal to the preset threshold, it can be determined that the model training is complete and the trained prediction model is obtained; when the number of iterations is equal to the preset number, it can also be determined that the model training is complete and the trained prediction model is obtained. Meeting either of these two conditions is sufficient to determine that the model training is complete and the trained prediction model is obtained.

[0040] The stress distribution field includes the stress at each node in the three-dimensional numerical model of the target area. The three-dimensional numerical model is a "geomechanical" digital model built based on the geological survey report, used for inverting or predicting initial geostress.

[0041] Furthermore, the current iteration number is obtained, and the model parameters of the initial model are optimized when the target value is greater than a preset threshold or the iteration number is less than a preset number.

[0042] Furthermore, the particle swarm optimization algorithm is used to optimize the penalty coefficient c and loss function parameters in the initial model. and kernel function parameters The penalty coefficient, loss function parameters, and kernel function parameters are all model parameters of the initial model.

[0043] Furthermore, the preset number of attempts is 1000.

[0044] In the aforementioned mechanism- and data-based geostress inversion method, measured geostress data and sampling depths are obtained from each sampling point in the target area. The measured geostress data includes measured horizontal and vertical stresses. Based on the measured geostress data and sampling depths, the specific values ​​of the fitting parameters in the variation equation are obtained, and an objective function is constructed based on the variation equation with known fitting parameters. This allows the distribution law of geostress in the target area to be introduced as a physical constraint into the objective function, so as to avoid the defect of the initial model ignoring the physical law during the training process, making the trained prediction model more reliable, thereby improving the accuracy of the stress distribution field of the target area obtained by inversion through the prediction model, and improving the efficiency of obtaining the stress distribution field.

[0045] In one embodiment, the measured horizontal stress includes the maximum horizontal principal stress and the minimum horizontal principal stress. The process of solving the fitting parameters in the variation equation in step S2 includes: The variation formula is obtained by fitting the measured ground stress data and sampling depth at each sampling point. The specific values ​​of the fitting parameters, The maximum horizontal principal stress at the sampling point. The minimum horizontal principal stress at the sampling point, Let be the vertical stress at the sampling point, D be the depth of the sampling point, and a and b be the fitting parameters of the variation relationship.

[0046] Vertical stress is the stress perpendicular to the Earth's surface experienced at the sampling point, mainly caused by the gravity of the overlying rock strata.

[0047] The maximum horizontal principal stress is the principal stress that exerts the greatest effect on the sampling point on a horizontal plane parallel to the Earth's surface. The minimum horizontal principal stress is the principal stress that exerts the least effect on the sampling point on a horizontal plane parallel to the Earth's surface.

[0048] In this embodiment, the variation relationship is obtained by fitting the measured ground stress data and sampling depth of each sampling point. The specific values ​​of the fitting parameters are obtained, so that the macroscopic laws of the distribution of geostress in the target area can be revealed through the variation formula. This allows the objective function to be constructed based on the known variation formula of the fitting parameters, avoiding the defect of the initial model ignoring physical laws during the training process, and making the trained prediction model more reliable.

[0049] In one embodiment, obtaining the stress distribution field of the target region through prediction model inversion includes: Obtain a geological survey report for the target area, which includes topographic contour data, borehole columnar section data, geological profile data, and fault attitude data for the target area. The free surface at the top of the three-dimensional numerical model is constructed based on topographic contour data; the rock strata interface of the three-dimensional numerical model is fitted by interpolation algorithm based on borehole columnar section data and geological profile data; and the fault plane and fault influence zone entities of the three-dimensional numerical model are constructed based on fault attitude data. Based on the rock strata interface, free surface, fault plane and fault influence zone entity, a three-dimensional numerical model of the target area is constructed, and the three-dimensional numerical model is meshed. The prediction model outputs the second predicted geostress data for each grid node after the three-dimensional numerical model is divided. Based on the second predicted geostress data of each grid node, the stress distribution field of the target area is obtained.

[0050] Topographic contour data describes the shape of the target area's surface. Topographic contour data is a set of two-dimensional curves with elevation (Z-value). All points on each curve have the same elevation.

[0051] Borehole columnar section data describes the vertical lithological distribution of a target area. The specific format of borehole columnar section data can be a table or a graph. It records the rock name, color, hardness, core recovery rate, etc., at different depths (Z) below a specific coordinate point (X, Y).

[0052] Geological profile data are subsurface cross-sections of a target area inferred by geologists based on boreholes and surface outcrops. Specifically, geological profile data can be presented as a two-dimensional cross-section, where the horizontal axis represents horizontal distance and the vertical axis represents depth.

[0053] Fault attitude data describes the spatial orientation of a fault (fracture / breccia zone). Fault attitude data typically includes strike, dip, and dip angle. Strike is the angle between the fault line and true north (e.g., 30 degrees east of north). Dip is the direction in which the fault plane dips downwards (e.g., southeastwards). Dip angle is the angle between the fault plane and the horizontal plane (e.g., steep 80 degrees, or gentle 30 degrees).

[0054] A free surface refers to the outermost surface of a target area that is not covered by other rock layers; that is, the actual ground surface. Specifically, using topographic contour data, a free, undulating surface is generated through computer algorithms. For example, a free, undulating surface can be generated using a Triangulated Irregular Network (TIN).

[0055] A rock stratum interface is the contact surface between different underground rock strata. A fault plane is a fracture surface where rock strata break and the rocks on both sides slide relative to each other. The fault-affected zone is a band-shaped area on both sides of the fault plane where the rocks have become fragmented and loose due to intense compression and friction. The fault-affected zone also possesses corresponding mechanical parameters, such as elastic modulus, internal friction angle, and cohesion.

[0056] The methods for constructing fault planes and fault-affected zones include, but are not limited to, topology construction based on "wire frame model", surface reconstruction based on "discrete smooth interpolation", and modeling based on "implicit potential field".

[0057] Each grid node after division covers the entire target area, and the number of grid nodes is greater than the number of sampling points.

[0058] The coordinate system in which the spatial coordinates of the grid nodes are located is consistent with the coordinate system in which the sampling points are located.

[0059] Based on the second predicted geostress data of each grid node, a continuous stress distribution field covering the target area is generated through interpolation calculation or data mapping.

[0060] Furthermore, the spatial coordinates of each grid node are input into the prediction model, and the predicted geostress data of each grid node output is used as the second predicted geostress data, or the optimized data of the predicted geostress data of each grid node is used as the second predicted geostress data.

[0061] In a specific scenario, the constructed 3D numerical model of the target region is as follows: Figure 3 As shown. Among them, III1, III2, IV1, and IV2 represent four different types of surrounding rock, and F115 is the entity of the fault influence zone.

[0062] In this embodiment, by acquiring the geological survey report of the target area, a free surface at the top of the three-dimensional numerical model is constructed based on topographic contour data. Based on borehole columnar section data and geological profile data, an interpolation algorithm is used to fit the rock strata interface of the three-dimensional numerical model. Based on fault attitude data, the fault plane and fault influence zone entity of the three-dimensional numerical model are constructed. Based on the rock strata interface, free surface, fault plane and fault influence zone entity, a three-dimensional numerical model of the target area is constructed, and the three-dimensional numerical model is meshed. In this way, the second predicted geostress data of each grid node after the three-dimensional numerical model is divided can be output by the prediction model. Based on the second predicted geostress data of each grid node, the stress distribution field of the target area is obtained.

[0063] In one embodiment, the second predicted geostress data for each grid node after the three-dimensional numerical model is divided by the prediction model includes: The spatial coordinates of each grid node after the three-dimensional numerical model is divided are input into the prediction model to obtain the initial geostress data of each grid node. The initial geostress data are optimized to obtain the second predicted geostress data for each grid node.

[0064] The optimization of the initial geostress data can be achieved using finite difference software, and the optimized data is the second predicted stress data for each grid node.

[0065] Furthermore, the initial geostress data of each grid node can generate an initial three-dimensional geostress field that is continuously distributed in space. Figure 4 A schematic diagram of the stress distribution field formed by the optimized second predicted stress data.

[0066] In this embodiment, the initial geostress data of each grid node after the three-dimensional numerical model is divided is obtained by inputting the spatial coordinates of each grid node into the prediction model. The initial geostress data is then optimized to obtain the second predicted geostress data of each grid node, which makes the second predicted stress data obtained by inversion more accurate.

[0067] In one embodiment, optimizing the initial geostress data to obtain the second predicted geostress data for each grid node includes: The initial stress step field, composed of the initial geostress data of each grid node, is used as the initial stress condition, and the initial stress condition is applied to the three-dimensional numerical model. Based on the three-dimensional numerical model with the applied initial stress step field and the constitutive model of the target region, stress balance calculations are performed using finite difference software to obtain the second predicted geostress data for each grid node.

[0068] Among them, the constitutive model is a mathematical rule that describes the relationship between the "force" and "deformation" of rock or soil within the target area.

[0069] When performing stress equilibrium calculations, the finite difference software can cause disturbances to the initial geostress data by structural elements such as faults within the three-dimensional numerical model, thus obtaining a second predicted geostress data that satisfies the mechanical equilibrium conditions and includes the influence of geological structures. Figure 3 The schematic diagram of the initial stress distribution field corresponding to the three-dimensional numerical model shown is as follows: Figure 5 As shown, the initial stress distribution field reflects the basic trend of stress variation with depth and topography.

[0070] In this embodiment, the initial stress step field, which is composed of the initial geostress data of each grid node, is used as the initial stress condition, and the initial stress condition is applied to the three-dimensional numerical model. Based on the three-dimensional numerical model with the applied initial stress step field and the constitutive model of the target area, stress balance calculation is performed using finite difference software to obtain the second predicted geostress data of each grid node. This can effectively characterize local "abnormal" phenomena such as stress release and concentration caused by geological structures such as faults, making the second predicted stress data obtained by inversion closer to the actual engineering situation.

[0071] In one embodiment, the coordinate system of the spatial coordinates of the grid nodes is consistent with the coordinate system of the sampling points, and the method further includes: Determine the target node with consistent spatial coordinates from the sampling points and grid nodes; The stress deviation is calculated based on the measured in-situ stress data and the second predicted in-situ stress data of the target node. Based on the comparison results between the stress deviation and the preset deviation threshold, the reliability of the second predicted geostress data of each grid node is determined.

[0072] The target node exists in both the sampling points and the grid nodes, and is a point shared by both sampling points and grid nodes. Therefore, the target node contains both measured geostress data and second predicted geostress data.

[0073] The second predicted geostress data includes the predicted maximum horizontal principal stress, the predicted minimum horizontal principal stress, and the predicted vertical stress.

[0074] Stress deviation is the sum of one or more of the differences in maximum principal stress, minimum principal stress, and vertical stress at the target node. The difference in maximum principal stress is the difference between the measured and predicted maximum horizontal principal stress at the target node. The difference in minimum principal stress is the difference between the measured and predicted minimum horizontal principal stress at the target node. The difference in vertical stress is the difference between the measured and predicted vertical stress at the target node.

[0075] In a specific scenario, a schematic diagram illustrating the difference in maximum principal stresses among the target nodes is shown below. Figure 6 As shown, the horizontal axis represents the target node number, and the maximum principal stress error is the difference between the maximum principal stresses. Figure 6 This indicates that the difference in the maximum principal stress at each target node is less than 4 MPa, which is significantly lower than that of the initial field of a single SVR, verifying the effectiveness and superiority of this application.

[0076] Based on the comparison results between the stress deviation and the preset deviation threshold, the reliability of the second predicted geostress data of each grid node is determined as follows: when the stress deviation is greater than the deviation threshold, the second predicted geostress data of each grid node is determined to be unreliable; when the stress deviation is less than or equal to the deviation threshold, the second predicted geostress data of each grid node is determined to be reliable.

[0077] Furthermore, if the second predicted geostress data of each grid node is unreliable, the parameters of the prediction model are adjusted to obtain reliable second predicted geostress data.

[0078] Furthermore, if there are multiple target nodes, the stress deviation of each target node is calculated separately, and the total number of nodes with a deviation greater than the preset first threshold is counted. When the total number is less than the preset number, the second predicted geostress data of each grid node is determined to be reliable.

[0079] Furthermore, if there are multiple target nodes, the stress deviation of each target node is calculated separately, and the stress deviations of all target nodes are summed to obtain the total stress deviation. When the total stress deviation is less than or equal to a preset second deviation threshold, the second predicted geostress data of each grid node is determined to be reliable. Both the first and second deviation thresholds are deviation thresholds.

[0080] In this embodiment, by determining the target node with consistent spatial coordinates from the sampling points and grid nodes, and calculating the stress deviation based on the measured geostress data and the second predicted geostress data of the target node, the reliability of the second predicted geostress data of each grid node can be determined based on the comparison result of the stress deviation and the preset deviation threshold.

[0081] In one embodiment, the measured horizontal stress includes the maximum horizontal principal stress and the minimum horizontal principal stress, and the first predicted geostress data includes the predicted maximum horizontal principal stress, the predicted minimum horizontal principal stress, and the predicted vertical stress, with a mean square error of [missing value]. ; in, The predicted horizontal maximum principal stress at sampling point i, The predicted horizontal minimum principal stress at sampling point i, The predicted vertical stress at sampling point i, The maximum horizontal principal stress at sampling point i, Let i be the minimum horizontal principal stress. Let be the vertical stress at sampling point i.

[0082] In this embodiment, by using By calculating the mean square error, the prediction results of both vertical and horizontal stress are combined, making the objective function more consistent with the actual stress distribution. This ensures that the training of the initial model follows the actual stress distribution, thereby improving the training effect of the initial model.

[0083] This application also provides an application scenario in which the above-mentioned mechanism- and data-based geostress inversion method is applied. Specifically, the application of the mechanism- and data-based geostress inversion method in this scenario is as follows: like Figure 7 As shown, measured ground stress data at each sampling point in the target area are obtained. Based on the measured ground stress data and sampling depth, the variation relationship is obtained. The specific values ​​of the fitting parameters are determined, and the objective function is constructed based on the known variation relationship of the fitting parameters. The process involves: predicting the first predicted geostress data at each sampling point using an initial model; calculating the target value M using an objective function based on the first predicted geostress data and measured geostress data at each sampling point; optimizing the model parameters of the initial model based on the target value when the target value is greater than a preset threshold or the number of iterations is less than a preset number; repeatedly predicting the geostress data at each sampling point based on the optimized initial model until the target value is less than or equal to the preset threshold or the number of iterations is equal to the preset number, thus obtaining a trained prediction model. Simultaneously, a geological survey report of the target area is obtained, including topographic contour data, borehole columnar section data, geological profile data, and fault attitude data. A free surface at the top of the 3D numerical model is constructed based on the topographic contour data. Based on the borehole columnar section data and geological profile data, an interpolation algorithm is used to fit the rock strata interfaces of the 3D numerical model. The fault planes and fault influence zones of the 3D numerical model are constructed based on the fault attitude data. Finally, a 3D numerical model of the target area is constructed based on the rock strata interfaces, free surface, fault planes, and fault influence zones, and the 3D numerical model is meshed. The spatial coordinates of each grid node after the 3D numerical model is divided are input into the prediction model to obtain the initial geostress data of each grid node. The initial stress distribution field formed by the initial geostress data of each grid node is used as the initial stress condition, and this initial stress condition is applied to the 3D numerical model. Based on the 3D numerical model with applied initial stress conditions and the constitutive model of the target region, stress balance calculations are performed using finite difference software to obtain the second predicted geostress data of each grid node. Based on the second predicted geostress data of each grid node, the stress distribution field of the target region is obtained. Target nodes with consistent spatial coordinates are determined from the sampling points and grid nodes. Based on the measured geostress data and the second predicted geostress data of the target nodes, the stress deviation is calculated. Based on the comparison between the stress deviation and a preset deviation threshold, the reliability of the second predicted geostress data of each grid node is determined.

[0084] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0085] Based on the same inventive concept, this application also provides a mechanism- and data-based geostress inversion device for implementing the mechanism- and data-based geostress inversion method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more mechanism- and data-based geostress inversion device embodiments provided below can be found in the limitations of the mechanism- and data-based geostress inversion method described above, and will not be repeated here.

[0086] In one embodiment, such as Figure 8 As shown, a geostress inversion device based on mechanism and data is provided, comprising: The data acquisition unit is used to acquire the measured ground stress data and sampling depth of each sampling point in the target area. The measured ground stress data includes the measured horizontal stress and the measured vertical stress. The objective function construction unit is used to obtain the specific values ​​of the fitting parameters in the variation formula based on the measured geostress data and sampling depth, and to construct the objective function based on the variation formula with known fitting parameters. The relationship is a formula that characterizes the actual ratio between the measured horizontal stress and the measured vertical stress at the sampling point as a function of the sampling depth. This represents the actual ratio at sampling point i. This represents the prediction ratio obtained based on the first predicted geostress data from sampling point i, where n is the number of sampling points. W is the penalty factor, and W is the mean square error between the first predicted geostress data and the measured geostress data at the sampling point. The target value calculation unit is used to predict the first predicted ground stress data of each sampling point through the initial model, and to calculate the target value M using the objective function based on the first predicted ground stress data and the measured ground stress data of each sampling point. The stress inversion unit is used to optimize the model parameters of the initial model when the target value is greater than a preset threshold or the number of iterations is less than a preset number. Based on the optimized initial model, step S3 is repeated until the target value is less than or equal to the preset threshold, or the number of iterations is equal to the preset number, to obtain the trained prediction model, so as to obtain the stress distribution field of the target area through the prediction model inversion.

[0087] The modules in the aforementioned mechanism- and data-based geostress inversion device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0088] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores various types of data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a mechanism- and data-based geostress inversion method.

[0089] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0090] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0091] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0092] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0093] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0094] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0095] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0096] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A geostress inversion method based on mechanism and data, characterized in that, The method includes: S1. Obtain the measured ground stress data and sampling depth of each sampling point in the target area. The measured ground stress data includes measured horizontal stress and measured vertical stress. S2. Based on the measured ground stress data and the sampling depth, obtain the specific values ​​of the fitting parameters in the variation formula, and construct the objective function based on the variation formula with known fitting parameters. The aforementioned relationship is a formula characterizing the actual ratio between the measured horizontal stress and the measured vertical stress at the sampling point as a function of the sampling depth. This represents the actual ratio at sampling point i. This represents the prediction ratio obtained based on the first predicted geostress data from sampling point i, where n is the number of sampling points. W is the penalty factor, and W is the mean square error between the first predicted geostress data and the measured geostress data at the sampling point. S3. Predict the first predicted ground stress data of each sampling point using the initial model, and calculate the target value M using the objective function based on the first predicted ground stress data and the measured ground stress data of each sampling point. S4. When the target value is greater than a preset threshold or the number of iterations is less than a preset number, optimize the model parameters of the initial model based on the target value, and repeat step S3 based on the optimized initial model until the target value is less than or equal to the preset threshold, or the number of iterations is equal to the preset number, to obtain a trained prediction model, so as to obtain the stress distribution field of the target region through the prediction model.

2. The method according to claim 1, characterized in that, The measured horizontal stress includes the maximum horizontal principal stress and the minimum horizontal principal stress. The process of solving the fitting parameters in the variation relationship in step S2 includes: Based on the measured ground stress data and sampling depth of each sampling point, a fitting operation is performed to obtain the variation relationship. The specific values ​​of the fitting parameters, The maximum horizontal principal stress at the sampling point. The minimum horizontal principal stress at the sampling point, Let be the vertical stress at the sampling point, D be the depth of the sampling point, and a and b be the fitting parameters of the variation relationship.

3. The method according to claim 1, characterized in that, The stress distribution field of the target region obtained by inversion through the prediction model includes: Obtain a geological survey report for the target area, the geological survey report including topographic contour data, borehole columnar section data, geological profile data and fault attitude data for the target area; The free surface at the top of the three-dimensional numerical model is constructed based on the topographic contour data; the rock strata interface of the three-dimensional numerical model is fitted by an interpolation algorithm based on the borehole columnar section data and the geological profile data; and the fault plane and fault influence zone entities of the three-dimensional numerical model are constructed based on the fault attitude data. Based on the rock strata interface, the free surface, the fault plane and the fault influence zone entity, a three-dimensional numerical model of the target area is constructed, and the three-dimensional numerical model is meshed. The prediction model outputs the second predicted geostress data for each grid node after the three-dimensional numerical model is divided. Based on the second predicted geostress data of each grid node, the stress distribution field of the target area is obtained.

4. The method according to claim 3, characterized in that, The second predicted geostress data for each grid node after the three-dimensional numerical model is divided by the prediction model includes: The spatial coordinates of each grid node after the three-dimensional numerical model is divided are input into the prediction model to obtain the initial geostress data of each grid node; The initial geostress data are optimized to obtain the second predicted geostress data for each grid node.

5. The method according to claim 4, characterized in that, The optimization of each of the initial geostress data to obtain the second predicted geostress data for each of the grid nodes includes: The initial stress step field, which is formed by the initial geostress data of each of the grid nodes, is used as the initial stress condition, and the initial stress condition is applied to the three-dimensional numerical model. Based on the three-dimensional numerical model with the applied initial stress step field and the constitutive model of the target region, stress balance calculations are performed using finite difference software to obtain the second predicted geostress data for each of the grid nodes.

6. The method according to claim 3, characterized in that, The coordinate system in which the spatial coordinates of the grid nodes are located is consistent with the coordinate system in which the sampling points are located, and the method further includes: Determine the target node with consistent spatial coordinates from the sampling points and the grid nodes; Based on the measured ground stress data of the target node and the second predicted ground stress data, the stress deviation is calculated; Based on the comparison results between the stress deviation and the preset deviation threshold, the reliability of the second predicted geostress data of each grid node is determined.

7. The method according to claim 1, characterized in that, The measured horizontal stress includes the maximum and minimum horizontal principal stresses, and the first predicted geostress data includes the predicted maximum horizontal principal stress, the predicted minimum horizontal principal stress, and the predicted vertical stress. The mean square error is... ; in, The predicted horizontal maximum principal stress at sampling point i, The predicted horizontal minimum principal stress at sampling point i, The predicted vertical stress at sampling point i, The maximum horizontal principal stress at sampling point i, Let i be the minimum horizontal principal stress. Let be the vertical stress at sampling point i.

8. A mechanism- and data-based geostress inversion device for performing the method according to any one of claims 1-7, characterized in that, The device includes: The data acquisition unit is used to acquire the measured ground stress data and sampling depth of each sampling point in the target area. The measured ground stress data includes measured horizontal stress and measured vertical stress. The objective function construction unit is used to obtain the specific values ​​of the fitting parameters in the variation relationship based on the measured in-situ stress data and the sampling depth, and to construct the objective function based on the variation relationship with the known fitting parameters. The relationship is a formula that characterizes the actual ratio between the measured horizontal stress and the measured vertical stress at the sampling point as a function of the sampling depth. This represents the actual ratio at sampling point i. This represents the prediction ratio obtained based on the first predicted geostress data from sampling point i, where n is the number of sampling points. W is the penalty factor, and W is the mean square error between the first predicted geostress data and the measured geostress data at the sampling point. The target value calculation unit is used to predict the first predicted geostress data of each of the sampling points through an initial model, and to calculate the target value M using the objective function based on the first predicted geostress data and the measured geostress data of each of the sampling points. The stress inversion unit is used to optimize the model parameters of the initial model when the target value is greater than a preset threshold or the number of iterations is less than a preset number, and repeat step S3 based on the optimized initial model until the target value is less than or equal to the preset threshold, or the number of iterations is equal to the preset number, to obtain a trained prediction model, so as to invert the stress distribution field of the target region through the prediction model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Images