An intelligent analysis system and method for inversion of three-dimensional ground stress field of layered surrounding rock

Abstract

The present invention relates to the technical field of inversion analysis of layered surrounding rocks, and discloses a three-dimensional geostress field inversion intelligent analysis system and method for layered surrounding rocks. The system includes a three-dimensional geological numerical module, an intelligent inversion analysis module, and an inversion result verification module. The method includes collecting geological data in the target layered surrounding rocks; normalizing the geological data and performing sensitivity analysis; constructing a three-dimensional geological model and performing numerical simulation; intelligently generating a geostress field inversion intelligent analysis model based on the three-dimensional geological model; training the geostress field inversion intelligent analysis model using training samples; inputting real-time geological data into the geostress field inversion intelligent analysis model to obtain geostress inversion analysis results; visually outputting the geostress analysis results; and adjusting the geostress field inversion intelligent analysis model according to measured data. The present invention can provide a scientific geostress field analysis basis for complex projects such as tunnel projects, underground projects, and hydropower stations.

Description

Technical Field

[0001] The present invention relates to the technical field of layered surrounding rock inversion analysis, and in particular to a three-dimensional in-situ stress field inversion intelligent analysis system and method for layered surrounding rock. Background Art

[0002] In-situ stress field inversion involves building a mathematical model based on field-measured data (such as displacement, stress, and strain) to infer the initial in-situ stress state of the underground rock mass. Under complex geological conditions, such as deep fault zones or areas of high in-situ stress, traditional linear superposition methods struggle to accurately reflect the true distribution of the in-situ stress field. With the advancement of artificial intelligence (AI), the application of intelligent algorithms in in-situ stress field inversion is increasing. For example, the in-situ stress field inversion method based on support vector machines (SVR) minimizes the structural risk of eigenvectors, reducing the impact of random errors in measured data on parameter inversion. The background technology of intelligent analysis systems and methods for in-situ three-dimensional in-situ stress field inversion in layered surrounding rock ranges from basic principles to application in complex geological conditions, as well as the introduction of intelligent algorithms and their verification in engineering practice. These technological developments provide important theoretical and practical support for solving in-situ stress problems in geotechnical engineering.

[0003] Prior art one, a Chinese patent with patent number 202310538323.5, discloses an intelligent real-time rock formation inversion identification method based on LWD parameter characteristics. Specifically, the method involves conducting simulated drilling tests on different types of rock samples on an indoor drilling rig test platform. Sensors are installed on the drill rig to collect LWD parameters during rock drilling. The collected drill pipe axial vibration signals are then subjected to time-frequency analysis and training to create an artificial neural network model for rock formation identification. At the drilling site, the collected LWD data is input into the trained artificial neural network model, which then outputs real-time identification results such as the interface and lithology of the rock formation encountered, thereby classifying the stratum geology and rock mass. During the drilling process, workers can adjust drilling parameters and paths in a timely manner based on the LWD rock formation identification results. While this method is beneficial for ensuring borehole quality, improving drilling efficiency, and saving significant manpower and material resources, the inversion method has certain limitations. It ignores the nonlinear characteristics of the geostress field, is sensitive to human factors, and lacks uniqueness.

[0004] Prior art 2, Chinese patent No. 202411773937.2, discloses a method for detecting and analyzing the floor of a coal seam working face, relating to the field of detection and analysis technology. By observing the potential changes at different locations and elevations through parallel electrical instruments and performing three-dimensional electrical inversion, the resistivity distribution at different depths within the working face and its floor can be obtained. These resistivity distributions provide technical support for objective and accurate geological interpretations, helping to better understand the hydrogeological conditions of the rock formations at the working face floor, thereby providing an important reference for safe mining. Although the deviation of the measured conductivity value can be accurately corrected by evaluating and correcting the thermal and electromagnetic effects in the environment, the accuracy and reliability of the conductivity measurement data can be improved. The corrected conductivity values ​​SR(xi, yi, zi) of each grid point are then input into a three-dimensional model and annotated to construct a three-dimensional conductivity model, thereby visualizing the electrical property distribution information of the underground medium. However, the inversion method has certain limitations, ignoring the nonlinear characteristics of the ground stress field, being sensitive to human factors, and lacking uniqueness.

[0005] Prior art three, Chinese patent No. 202411699733.9, discloses a lithofacies analysis system and method based on a shale geo-mechanical coupled lithofacies classification system, relating to the field of shale lithofacies analysis. The system construction module determines the geo-mechanical coupled lithofacies classification system based on information from the data acquisition module; the prediction module determines lithofacies prediction information for a single well; the feedback adjustment module performs feedback adjustment on the lithofacies prediction information based on a set range of rock mechanics and geostress parameters to obtain adjustment information; the three-dimensional model construction module uses a genetic algorithm to perform seismic attribute mosaicking and integrated inversion based on the seismic attribute data and adjustment information, and employs a collaborative kriging method to construct a three-dimensional model for geo-mechanical coupled lithofacies well-seismic collaborative modeling. Although the parameter model determination module uses a layer-control and phase-control dual-control method to control constraints and determine the heterogeneity of the geo-mechanical characteristics of the shale layer in three-dimensional space, providing a basis for establishing production areas and conducting exploration and development of artificial oil and gas reservoirs, the inversion method has certain limitations. It ignores the nonlinear characteristics of the geostress field, is sensitive to human factors, and lacks uniqueness.

[0006] Currently, the inversion methods of existing technologies 1, 2, and 3 have certain limitations, ignore the nonlinear characteristics of the geostress field, are sensitive to human factors, and lack uniqueness. To address the above problems, the present invention provides an intelligent analysis system and method for inversion of three-dimensional geostress fields in layered surrounding rock. Summary of the Invention

[0007] The main purpose of the present invention is to provide an intelligent analysis system and method for inversion of three-dimensional geostress field of layered surrounding rock, so as to solve the problems in the prior art that the inversion method has certain limitations, ignores the nonlinear characteristics of the geostress field, is sensitive to human factors, and lacks uniqueness.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A three-dimensional inverse intelligent analysis system for layered surrounding rock stress fields, comprising:

[0010] The 3D geological numerical module is used to collect geological data on fault distribution, rock formation occurrence, physical and mechanical parameters, and ground stress measurements in the target layered surrounding rock; normalize the geological data and perform sensitivity analysis; construct a 3D geological model and perform numerical simulation;

[0011] The intelligent inversion analysis module is used to design parameter combinations for intelligent inversion through orthogonal experiments and generate training samples based on numerical simulation results; generate an intelligent analysis model for inversion of geostress field based on a three-dimensional geological model; and use the training samples to train the intelligent analysis model for inversion of geostress field.

[0012] The inversion result verification module is used to input real-time geological data into the geostress field inversion intelligent analysis model to obtain the geostress inversion analysis results; visualize the geostress analysis results; and adjust the geostress field inversion intelligent analysis model based on the measured data.

[0013] As a further improvement of the present invention, the three-dimensional geological numerical module includes:

[0014] The rock formation classification submodule is used to import drilling data, geological exploration data, geophysical data, and measured ground stress data to define the spatial range of the 3D geological model; and to determine the range of rock mass mechanical parameters and ground stress parameters in different regions through geological zoning.

[0015] The shear-compression submodule is used to perform 3D meshing and generate regular tetrahedral elements. It processes complex fault regions and assigns heterogeneous rock mechanical parameters to grid elements. It also applies normal stress to the top surface of the 3D geological model to simulate the self-weight stress field and obtain the shear-compression effect that reflects the fault structure.

[0016] The 3D geological model adjustment submodule is used to comprehensively characterize the nonlinear superposition effect of the geostress field through the pressure coefficient and rock body force; compare the measured data of the geostress components at the measuring points with the simulated values, and dynamically adjust the 3D geological model by combining the comparison results with real-time geological data.

[0017] As a further improvement of the present invention, the rock formation classification submodule includes:

[0018] The classification and sorting unit is used to import drilling data, geological survey data, and geophysical data, divide the strata according to the drilling data, unify the numbering and sorting, classify them into the same rock unit according to the lithologic pixel row and vertical position, and handle the impact of faults on the continuity of the bottom layer;

[0019] The fault trajectory unit is used to determine the fault dip and extension depth based on the fault distribution and rock formation occurrence information in the geological survey data, combined with the geophysical inversion results, and generate the fault trajectory line;

[0020] Define spatial range units to use borehole data to correct rock mass morphology, combine seismic interpretation horizons with velocity / density data, divide different lithologic units, fuse fault geometry with stratigraphic boundary data, and define the spatial range of the 3D geological model.

[0021] As a further improvement of the present invention, the shearing and extrusion submodule includes:

[0022] Grid division units are used to grid the 3D geological model; process the fault intersection relationship in complex fault areas; perform local grid densification on the fault zone; and assign different rock mass mechanical parameters to corresponding grid units based on fault type and measured data.

[0023] The normal constraint unit is used to apply normal stress boundary conditions on the top surface of the 3D geological model, simulate the rock mass self-weight stress field by calculating the blowing response force generated by the rock mass density, and apply normal constraints to the ground and lateral boundaries of the 3D geological model;

[0024] The stress direction unit applies horizontal tectonic stress on the lateral boundary of the 3D geological model. The horizontal tectonic stress component is simulated by applying a normal unit load on the lateral boundary. The proportional relationship between horizontal stress and vertical stress is defined in combination with the pressure coefficient. The tectonic stress direction is determined by inverting the measured ground stress data.

[0025] As a further improvement of the present invention, the intelligent inversion analysis module includes:

[0026] A sample set submodule is formed to select parameter combinations that affect the geostress field and perform numerical simulations using a three-dimensional geological model. The parameter combinations are introduced into the three-dimensional geological model to calculate the geostress field distribution results under different working conditions. A training sample set is formed based on the simulation results and the parameter combinations.

[0027] Generate an inversion analysis model submodule, which is used to generate an intelligent inversion analysis model for the geostress field based on the 3D geological model; determine the center position of the radial basis function through self-organizing learning, establish a mapping relationship between input parameters and output geostress field; select key parameters and perform nonlinear fitting;

[0028] The inversion analysis model optimization submodule is used to input the generated training samples into the inversion intelligent analysis model of the geostress field, adjust the fitting coefficient until it meets the preset value, and then compare the measured value with the inversion value to optimize the inversion intelligent analysis model of the geostress field to obtain the final inversion intelligent analysis model.

[0029] As a further improvement of the present invention, a sample collection submodule is formed, including:

[0030] The stress field distribution unit is used to select key parameters that affect the ground stress field in the three-dimensional geological model and determine the parameter combination scheme; the selected parameter combination is input into the three-dimensional geological model to perform numerical simulation calculations to obtain the ground stress field distribution results under different working conditions;

[0031] The parameter combination unit is used to extract the geostress components and distribution patterns of key measuring points from the 3D geological model after each numerical simulation is completed, and record the corresponding parameter combination; the parameter combination is used as the input variable, and the corresponding simulated geostress field result is used as the output variable;

[0032] The training sample set storage unit is used to combine input variables and output variables into parameter ground stress field paired data; eliminate abnormal data due to unqualified parameters, and optimize the samples through iterative optimization; store the verified parameter combination and simulated ground stress field data as a training sample set in a preset format.

[0033] As a further improvement of the present invention, generating an inversion analysis model submodule includes:

[0034] The weight setting unit is used to generate an intelligent analysis model for geostress field inversion based on a three-dimensional geological model; calculate the stress contribution value of each parameter combination under simulated working conditions, and determine the influence weight of the sensitivity of the geostress field distribution on the output result by quantifying the sensitivity of each parameter to the geostress field distribution;

[0035] The balance relationship unit is used to sort the calculated stress contribution values ​​by magnitude and compare them with the preset error threshold. By repeatedly adjusting the parameter combination, the key parameters are retained for nonlinear fitting. The overall inversion error is evaluated at each iteration. Weight parameters are introduced to balance the relationship between data-driven and physical constraints.

[0036] The iterative optimization unit is used to establish a mapping relationship between input parameters and the output of the geostress field, determine the center position of the function through self-organizing learning, and adjust the fitting coefficient based on the screened key parameter set until the inversion intelligent analysis model meets the preset three-dimensional distribution characteristic standards of the geostress field.

[0037] As a further improvement of the present invention, the balance relationship unit includes:

[0038] A weighting subunit is used to obtain data-driven items based on the error square of numerical simulation values ​​and measured values, and to use historical stratum and surrounding rock data as physical constraints; different weights are assigned to data-driven items and physical constraints;

[0039] The weight optimization subunit is used to dynamically adjust the weight parameters according to the comparison between the current inversion error and the preset threshold during the iteration process; by quantifying the sensitivity of the parameters to the ground stress field, the key parameters are screened out and given higher weights; non-key parameters are eliminated;

[0040] The parameter combination and weight configuration subunit is used to combine the prior distribution of physical constraints with the data-driven likelihood function to obtain the posterior probability; optimize the weight parameters based on the posterior probability; iterate in sequence until the preset tolerance is reached or the weight change reaches a preset trend, terminate the iteration, and output the optimized parameter combination and weight configuration.

[0041] As a further improvement of the present invention, the inversion result verification module includes:

[0042] The forward modeling verification submodule is used to communicate the stress field inversion intelligent analysis model with the 3D geological model, input real-time geological data into the inverse intelligent analysis model for simulation, and bring the output parameters of the inverse intelligent analysis model into the 3D geological model for forward analysis and calculation;

[0043] The visualization submodule is used to generate the distribution of the geostress field based on the topography, faults, and rock structure characteristics of the 3D geological model, as well as the applied self-weight pressure and tectonic stress boundary conditions; the calculated geostress components are visualized and output through cloud maps;

[0044] The adjustment and optimization submodule is used to compare the calculated ground stress components with the real-time measured data. If they exceed the preset accuracy, the model adjustment instruction is triggered. If deviations are found during the verification, the weights of the stress field intelligent analysis model or the parameters of the three-dimensional geological model are optimized, and the inversion results are iteratively updated until the preset accuracy value is met.

[0045] To achieve the above object, the present invention also provides the following technical solutions:

[0046] A three-dimensional inverse intelligent analysis method for layered surrounding rock stress field, which is applied to the three-dimensional inverse intelligent analysis system for layered surrounding rock stress field, and the three-dimensional inverse intelligent analysis method for layered surrounding rock stress field comprises:

[0047] Collect geological data on fault distribution, rock formation occurrence, physical and mechanical parameters, and measured ground stress data in the target layered surrounding rock; normalize the geological data and perform sensitivity analysis; construct a three-dimensional geological model and perform numerical simulation;

[0048] The intelligent inversion department uses orthogonal experimental design parameter combinations and generates training samples based on numerical simulation results. It generates an intelligent analysis model for geostress field inversion based on a three-dimensional geological model. The training samples are used to train the intelligent analysis model for geostress field inversion.

[0049] Input real-time geological data into the geostress field inversion intelligent analysis model to obtain the geostress inversion analysis results; output the geostress analysis results in a visual format; and adjust the geostress field inversion intelligent analysis model based on the measured data.

[0050] The present invention collects geological information such as fault distribution, rock formation occurrence, physical and mechanical parameters, and ground stress measured data in the target layered surrounding rock through a three-dimensional geological numerical module, uses these data for normalization processing, and combines with sensitivity analysis to construct an accurate three-dimensional geological model. Through the intelligent inversion module, combined with the orthogonal experimental design parameter combination, numerical simulation is used to generate training samples, and an intelligent analysis model for ground stress field inversion is constructed based on the three-dimensional geological model. Through the inversion result verification module, real-time geological data is input into the intelligent analysis model, the ground stress inversion analysis results are obtained, and the results are displayed through visual output. By combining the three-dimensional numerical model and the intelligent inversion method, the accuracy of ground stress field inversion can be effectively improved, especially under complex geological conditions, such as deep structural areas or tunnel projects, the ground stress field distribution characteristics can be more accurately reflected. Through visualization technology, the inversion results are intuitively displayed, which is convenient for engineers and researchers to understand the distribution law of the ground stress field. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a schematic diagram of the functional modules of an embodiment of the intelligent inversion analysis system for three-dimensional in-situ stress field of layered surrounding rock according to the present invention;

[0052] Figure 2 This is a functional module diagram of a three-dimensional geological numerical module of an embodiment of the three-dimensional in-situ stress field inversion intelligent analysis system for layered surrounding rocks of the present invention;

[0053] Figure 3 This is a functional module diagram of a rock stratum classification submodule of an embodiment of the three-dimensional in-situ stress field inversion intelligent analysis system for layered surrounding rocks of the present invention;

[0054] Figure 4 This is a functional module diagram of the shear-compression action submodule of an embodiment of the three-dimensional in-situ stress field inversion intelligent analysis system for layered surrounding rock of the present invention;

[0055] Figure 5 This is a functional module diagram of an intelligent inversion analysis module of an embodiment of the three-dimensional in-situ stress field inversion intelligent analysis system for layered surrounding rock of the present invention;

[0056] Figure 6A schematic diagram of the functional modules for forming a sample collection submodule in one embodiment of the intelligent inversion analysis system for three-dimensional in-situ stress field of layered surrounding rock according to the present invention;

[0057] Figure 7 A schematic diagram of a functional module for generating an inversion analysis model submodule for an embodiment of the intelligent inversion analysis system for three-dimensional in-situ stress field of layered surrounding rock according to the present invention;

[0058] Figure 8 This is a functional module diagram of a balance relationship unit in one embodiment of the intelligent inversion analysis system for three-dimensional in-situ stress field of layered surrounding rock according to the present invention;

[0059] Figure 9 This is a functional module diagram of an inversion result verification module of an embodiment of the three-dimensional in-situ stress field inversion intelligent analysis system for layered surrounding rock of the present invention;

[0060] Figure 10 This is a schematic diagram of the steps of an embodiment of the intelligent analysis method for inversion of three-dimensional in-situ stress field of layered surrounding rock according to the present invention;

[0061] Figure 11 A rock quality slope diagram of an embodiment of the intelligent inversion analysis method for three-dimensional in-situ stress field of layered surrounding rock according to the present invention;

[0062] Figure 12 This is a schematic structural diagram of an embodiment of an electronic device of the present invention;

[0063] Figure 13 This is a schematic structural diagram of an embodiment of a storage medium of the present invention. DETAILED DESCRIPTION

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0065] The terms "first," "second," and "third" in this disclosure are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features indicated. Therefore, features specified as "first," "second," or "third" may explicitly or implicitly include at least one of such features. In the description of this disclosure, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this disclosure are intended only to illustrate the relative positional relationships and movement of components in a specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly. Furthermore, the terms "including," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements and may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to such process, method, product, or apparatus.

[0066] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0067] like Figure 1 As shown, this embodiment provides an embodiment of a three-dimensional in-situ stress field inversion intelligent analysis system for layered surrounding rock. In this embodiment, the three-dimensional in-situ stress field inversion intelligent analysis system for layered surrounding rock specifically includes:

[0068] 3D geological numerical module 1 is used to collect geological data such as fault distribution, rock formation occurrence, physical and mechanical parameters, and measured ground stress data in the target layered surrounding rock; normalize the geological data and perform sensitivity analysis; construct a 3D geological model and perform numerical simulation;

[0069] Intelligent inversion analysis module 2 is used to design parameter combinations for intelligent inversion through orthogonal experiments and generate training samples based on numerical simulation results; generate an intelligent analysis model for inversion of geostress field based on the three-dimensional geological model; and use the training samples to train the intelligent analysis model for inversion of geostress field;

[0070] The inversion result verification module 3 is used to input real-time geological data into the geostress field inversion intelligent analysis model to obtain the geostress inversion analysis results; output the geostress analysis results in a visual manner; and adjust the geostress field inversion intelligent analysis model according to the measured data.

[0071] Preferably, this embodiment uses a three-dimensional geological numerical module to collect geological information such as fault distribution, rock formation occurrence, physical and mechanical parameters, and measured geostress data in the target layered surrounding rock. This data is then normalized and combined with sensitivity analysis to construct an accurate three-dimensional geological model. The intelligent inversion module, combined with orthogonal experimental design parameter combinations, generates training samples using numerical simulation, and constructs an intelligent geostress field inversion analysis model based on the three-dimensional geological model. The inversion result verification module inputs real-time geological data into the intelligent analysis model to obtain geostress inversion analysis results, which are then displayed through visual output. By combining the three-dimensional numerical model and the intelligent inversion method, the accuracy of geostress field inversion can be effectively improved, especially under complex geological conditions, such as deep tectonic zones or tunnel projects, to more accurately reflect the distribution characteristics of the geostress field. Visualization technology allows the inversion results to be intuitively displayed, making it easier for engineers and researchers to understand the distribution patterns of the geostress field. Furthermore, through real-time data input and model adjustment, the model can be dynamically optimized to better meet actual engineering needs, thereby improving the safety of engineering design and construction. It can provide a scientific basis for geostress field analysis for complex projects such as tunnel projects, underground projects, and hydropower stations, helping to optimize design and construction plans, reduce project risks, and improve economic benefits.

[0072] Furthermore, if Figure 2 As shown, the three-dimensional geological numerical module 1 specifically includes:

[0073] The rock layer classification submodule 11 is used to import drilling data, geological exploration data, geophysical data and measured ground stress data to define the spatial range of the three-dimensional geological model; and to determine the range of rock mass mechanical parameters and ground stress parameters in different regions through geological zoning.

[0074] Among them, geological exploration data includes fault distribution and rock formation occurrence; geophysical data includes seismic interpretation horizon and velocity / density information; spatial scope includes stratum boundaries and fault geometry;

[0075] The shear-compression submodule 12 is used to perform three-dimensional meshing and generate regular tetrahedral units; process complex fault areas and assign heterogeneous rock mechanical parameters to grid units; and apply normal stress to the top surface of the three-dimensional geological model to simulate the self-weight stress field to obtain the shear-compression effect reflecting the fault structure;

[0076] The three-dimensional geological model adjustment submodule 13 is used to comprehensively characterize the nonlinear superposition effect of the ground stress field through the pressure coefficient and the rock body force; compare the measured data of the ground stress components at the measuring points with the simulated values, and dynamically adjust the three-dimensional geological model based on the comparison results combined with the real-time geological data.

[0077] Preferably, this embodiment integrates and fuses multi-source heterogeneous data by importing drilling data, geological exploration data, geophysical data, and measured geostress data. Geological zoning is used to determine the rock mass mechanical parameters and geostress parameter ranges for different regions. Three-dimensional meshing technology is then used to generate regular tetrahedral cells. Complex fault regions are processed and the heterogeneous rock mechanical parameters are assigned to the mesh cells. The 3D geological model adjustment module 13 comprehensively characterizes the nonlinear superposition effect of the geostress field using pressure coefficients and rock mass body forces. Based on the comparison results between measured data and simulated values, the 3D geological model is dynamically adjusted. Normal stress is applied to the top surface of the 3D geological model to simulate the self-weight stress field, reflecting the shear and compression effects of the fault structure. Through the integration and dynamic adjustment of multi-source data, the 3D geological model can more realistically reflect the spatial distribution and mechanical properties of the geological body, thereby improving modeling accuracy. The 3D geological model can adapt to engineering needs under complex geological conditions, such as subway site selection and engineering construction. The use of automated meshing technology and heterogeneous rock mechanical parameter assignment methods reduces manual intervention and improves modeling efficiency.

[0078] Further, if Figure 3 As shown, the rock formation classification submodule 11 specifically includes:

[0079] The classification and sorting unit 111 is used to import drilling data, geological survey data, and geophysical data, divide the strata according to the drilling data, uniformly number and sort them, classify them into the same rock layer unit according to the lithologic pixel row and vertical position, and handle the impact of faults on the continuity of the bottom layer;

[0080] The fault trajectory unit 112 is used to determine the fault dip and extension depth based on the fault distribution and rock formation occurrence information in the geological survey data and in combination with the geophysical inversion results, and generate the fault trajectory line;

[0081] The spatial range definition unit 113 is used to correct the rock mass morphology using the drilling data, combine the seismic interpretation layer and velocity / density data, divide different lithologic units, fuse the fault geometry with the stratigraphic boundary data, and define the spatial range of the three-dimensional geological model.

[0082] Preferably, this embodiment divides, uniformly numbers, and sorts the strata using borehole data, and classifies the strata into the same rock unit according to lithologic pixel rows and vertical positions. Based on the fault distribution and rock formation occurrence information in the geological survey data, combined with the geophysical inversion results, the fault dip and extension depth are determined, and the fault trajectory is generated. The rock mass morphology is corrected using the borehole data, and the different lithologic units are divided by combining the seismic interpretation layer and velocity / density data. The fault geometry is then fused with the stratigraphic boundary data to define the spatial extent of the three-dimensional geological model. Through the preprocessing and standardization operations of the classification and sorting unit, the raw data is converted into structured information, which facilitates subsequent analysis and modeling. The fault trajectory unit generates accurate fault trajectories by combining the geological survey data and geophysical inversion results, thereby improving the accuracy and reliability of the three-dimensional geological model. The spatial extent definition unit constructs a three-dimensional geological model containing complex geological structures by fusing the seismic interpretation layer and velocity / density data, as well as the fault geometry. The overall technology provides a more intuitive and comprehensive perspective for geological analysis through three-dimensional modeling and data fusion.

[0083] Furthermore, if Figure 4 As shown, the shearing and extrusion submodule 12 specifically includes:

[0084] The grid division unit 121 is used to perform grid division on the three-dimensional geological model; process the fault intersection relationship in the complex fault area; perform local grid densification on the fault zone; and assign different rock mass mechanical parameters to corresponding grid units according to the fault type and measured data;

[0085] The normal constraint unit 122 is used to apply normal stress boundary conditions on the top surface of the 3D geological model, calculate the blowing response force generated by the rock mass density to simulate the rock mass self-weight stress field; and apply normal constraints to the ground and lateral boundaries of the 3D geological model;

[0086] The stress direction unit 123 applies horizontal tectonic stress to the lateral boundary of the three-dimensional geological model, simulates the horizontal tectonic stress component by applying a normal unit load to the lateral boundary, and defines the proportional relationship between horizontal stress and vertical stress in combination with the pressure coefficient; and determines the tectonic stress direction by inverting the measured ground stress data.

[0087] Preferably, this embodiment divides complex geological bodies into multiple grid cells through gridding of the 3D geological model to achieve accurate modeling of geological structures. In complex fault regions, local grid refinement technology is used to improve the resolution of the fault zone, thereby more accurately simulating the impact of faults on the geological stress field. Different rock mechanical parameters (such as elastic modulus and Poisson's ratio) are assigned to corresponding grid cells based on fault type and measured data to reflect the mechanical properties of different geological bodies. Optimizing gridding methods (such as tetrahedral optimization) improves model accuracy and computational efficiency. Normal stress boundary conditions are applied to the top surface of the 3D geological model, and the blowing response force generated by the rock mass density is calculated to simulate the rock mass's self-weight stress field. Horizontal tectonic stress is applied to the lateral boundaries, and the proportional relationship between horizontal and vertical stresses is defined in combination with the pressure coefficient. The tectonic stress direction is determined by inverting the measured ground stress data. Through refined meshing and localized encryption techniques, the structure and stress distribution of complex geological bodies can be more accurately simulated, reducing computational errors. Optimized meshing methods and automated modeling processes significantly improve modeling efficiency, reduce manual intervention, and lower computational costs. Three-dimensional visualization technology enables intuitive display of geological models and stress field distribution, facilitating engineering design and risk assessment. It can effectively address engineering challenges in complex geological conditions, such as fault effects and self-weight stress field simulation, providing a scientific basis for the design and construction of tunnels, mines, and other projects.

[0088] Furthermore, if Figure 5 As shown, the intelligent inversion analysis module 2 specifically includes:

[0089] The sample set submodule 21 is used to select a parameter combination that affects the geostress field and perform numerical simulation using a three-dimensional geological model; the parameter combination is introduced into the three-dimensional geological model to calculate the geostress field distribution results under different working conditions; and a training sample set is formed based on the simulation results and the parameter combination;

[0090] The inversion analysis model generation submodule 22 is used to generate an intelligent inversion analysis model for the geostress field based on the three-dimensional geological model; determine the center position of the radial basis function through self-organizing learning, establish a mapping relationship between input parameters and output geostress field; select key parameters and perform nonlinear fitting;

[0091] The inversion analysis model optimization submodule 23 is used to input the generated training samples into the inversion intelligent analysis model of the geostress field, adjust the fitting coefficient until it meets the preset value, and then compare the measured value with the inversion value to optimize the inversion intelligent analysis model of the geostress field to obtain the final inversion intelligent analysis model.

[0092] Preferably, this embodiment selects a parameter combination that affects the geostress field, and uses a three-dimensional geological model for numerical simulation; the parameter combination is brought into the three-dimensional geological model, the geostress field distribution results under different working conditions are calculated, and a training sample set is formed based on the simulation results and the parameter combination. The center position of the radial basis function is determined by self-organizing learning, a mapping relationship between the input parameters and the output geostress field is established, and key parameters are screened for nonlinear fitting. The generated training samples are input into the geostress field inversion intelligent analysis model, and the fitting coefficient is adjusted until the preset value is met. At the same time, the model is optimized by comparing the measured values ​​with the inversion values, and finally an accurate geostress field inversion intelligent analysis model is obtained. By combining the three-dimensional geological model and the intelligent algorithm, the spatial distribution law of the geostress field can be more accurately reflected, and the inversion accuracy can be significantly improved. Accurate geostress field inversion results can provide a reliable basis for the design and construction of underground projects, reducing safety hazards caused by inaccurate stress field predictions. It is particularly suitable for geostress field inversion under complex geological conditions, such as deep-buried tunnels and hydropower station projects, and can effectively respond to challenges brought by factors such as geological structure and rock properties. The introduction of intelligent algorithms reduces the complex mechanical models and heavy calculations in traditional methods, thereby improving computational efficiency.

[0093] Further, if Figure 6 As shown, the sample set forming submodule 21 specifically includes:

[0094] The stress field distribution unit 211 is used to select key parameters affecting the geostress field in the three-dimensional geological model and determine a parameter combination scheme; the selected parameter combination is input into the three-dimensional geological model, and numerical simulation calculations are performed to obtain geostress field distribution results under different working conditions;

[0095] The parameter combination unit 212 is used to extract the geostress components and distribution patterns of key measuring points from the three-dimensional geological model after each numerical simulation is completed, and record the corresponding parameter combination; the parameter combination is used as an input variable, and the corresponding simulated geostress field result is used as an output variable;

[0096] The training sample set storage unit 213 is used to combine input variables and output variables into parameter ground stress field paired data; eliminate abnormal data due to unqualified parameters, and optimize the samples through iterative optimization; and store the verified parameter combination and simulated ground stress field data as a training sample set in a preset format.

[0097] Preferably, this embodiment screens key parameters within a three-dimensional geological model, determines parameter combinations, and then inputs these parameters into the model for numerical simulation calculations, thereby obtaining geostress field distribution results under different operating conditions. After each numerical simulation, the geostress components and their distribution patterns at key measurement points are extracted from the three-dimensional geological model, and the corresponding parameter combinations are recorded. By combining input and output variables into paired data, unqualified anomalous data is eliminated, and the samples are optimized, ultimately storing them in a preset format as a training sample set. By optimizing parameter combinations and iterative optimization, the model can more accurately reflect the distribution patterns of the geostress field, addressing the problem of insufficient fitting accuracy of local stress anomaly areas in traditional models. By storing and optimizing samples, the model can better adapt to the geostress field distribution under different operating conditions, improving the model's generalization ability and making it more practical in actual engineering applications. The automated process reduces manual intervention and improves the model's efficiency and reliability. Furthermore, iterative optimization further enhances the model's accuracy and stability. By combining three-dimensional geological modeling with numerical simulation techniques, efficient coupling between the geological model and the stress field is achieved, providing a new solution for stress field analysis under complex geological conditions.

[0098] Further, if Figure 7 As shown, the generating inversion analysis model submodule 22 specifically includes:

[0099] The weight setting unit 221 is used to generate an intelligent analysis model for geostress field inversion based on the three-dimensional geological model; calculate the stress contribution value of each parameter combination under the simulated working condition, and determine the influence weight of the sensitivity of the geostress field distribution on the output result by quantifying the sensitivity of each parameter to the geostress field distribution;

[0100] Among them, the weight adaptation kernel function of the weight is:

[0101]

[0102] Where, For the k Parameter combination in rock mass Measuring point j The stress contribution density tensor of ; Characterization m Curvature correction factor of the geological structure (calculated by the three-dimensional geological model); For measuring points j The principal value vector of the triaxial stress; Corresponding rock layer k Anisotropic damping factor; For the n Rock mass heterogeneity coefficient ν The gamma distribution probability correction term; Characterize the parallel / orthogonal axial strain coupling strength caused by tectonic movement; Represents the parameter vector and low-dimensional latent space The covariant gradient interaction of It is the weighted coefficient of directional permeability measurement; it is a weighted tensor integral equation that integrates the cross-theory of geomechanics, probability statistics and machine learning. Its theoretical basis and function can be analyzed in layers as follows: Geological structure curvature correction factor Derived from the curvature-stress coupling theory in continuum mechanics, it uses the m-order curvature tensor to correct the secondary stress field generated by rock bending, quantifying the topological influence of geological folds / faults and other structural forms on the ground stress field. When m=2, it corresponds to the stress concentration effect of the double curvature structure. Gamma distribution correction term The Bayesian rock mass parameter inversion framework is used to describe the prior probability of rock mass heterogeneity parameters through gamma distribution; probability weights are introduced in the integral kernel, and when a certain type of rock mass parameter The contribution is automatically reduced when it deviates from the typical value. Based on the anisotropic constitutive relationship of porous media, They correspond to the coaxial / non-coaxial strains generated by tectonic movements, respectively; they characterize the stress principal axis rotation effect caused by tectonic movements such as thrust faults.

[0103] Gradient interaction term Achieve nonlinear dimensionality reduction from parameter space to latent space, and ensure that geomechanical constraints are maintained during parameter optimization through covariant gradients; when a certain parameter Adjustment direction and potential space Automatically increase the penalty term when the mechanical response trend is orthogonal. Directed penetration measure The conductivity tensor, derived from seepage theory, is used here to control the directionality of stress propagation; in sedimentary rock layers, =cos²φ (φ is the normal angle of the bedding plane), reflecting the preferential conduction of stress along the bedding plane.

[0104] The dynamic weight allocation mechanism realizes the "localized" calculation of parameter contribution through the exponential decay characteristics of the integral kernel, avoiding the boundary effect caused by traditional global weighting. - - The coupling relationship fully describes the "tectonic stress (σ)-lithologic damping (τ)-strain history ( )" closed-loop feedback. The model achieves a good agreement with the measured stress inversion in deep coal mines. Its innovation lies in the integration of geological knowledge ( , term) and data driven (Γ, Items) are organically unified through tensor operations, which has better cross-scale adaptability than traditional inversion methods;

[0105] The balance relationship unit 222 is used to sort the calculated stress contribution values ​​by magnitude and compare them with a preset error threshold; adjust the parameter combination through repeated iterations to retain the key parameters for nonlinear fitting; evaluate the overall inversion error at each iteration; and introduce weight parameters to balance the relationship between data-driven and physical constraints;

[0106] The iterative optimization unit 223 is used to establish a mapping relationship between the input parameters and the output of the geostress field, determine the center position of the function through self-organizing learning, and adjust the fitting coefficient based on the screened key parameter set until the inversion intelligent analysis model meets the preset three-dimensional distribution characteristic standard of the geostress field.

[0107] Preferably, this embodiment generates an inversion intelligent analysis model for geostress field based on a three-dimensional geological model, calculates the stress contribution value of each parameter combination under simulated working conditions, and determines the weight of its influence on the output result by quantifying the sensitivity of each parameter to the geostress field distribution; sorts the calculated stress contribution values ​​by magnitude and compares them with a preset error threshold; adjusts the parameter combination through repeated iterations, retains the key parameters for nonlinear fitting; introduces weight parameters to balance the relationship between data drive and physical constraints; establishes a mapping relationship between input parameters and geostress field output; determines the function center position through self-organizing learning; adjusts the fitting coefficient based on the screened key parameter set until the inversion intelligent analysis model meets the preset geostress field three-dimensional distribution characteristic standard; and Analysis and weight adjustment can more accurately reflect the impact of various parameters on the distribution of the geostress field, thereby improving the accuracy of the inversion results. Combining iterative optimization and nonlinear fitting further enhances the model's adaptability to complex geological conditions. By introducing weight parameters and balancing the relationship between data-driven and physical constraints, the model's robustness in the face of sparse data or uncertainty is enhanced. The iterative optimization unit screens key parameters and adjusts fitting coefficients to ensure that the model maintains high fitting accuracy under different working conditions. This provides an efficient and reliable solution for geostress field inversion under complex geological conditions, with important theoretical significance and practical engineering application value. The combination of a three-dimensional geological model and an intelligent analysis model can provide a scientific basis for engineering design and construction, reducing engineering risks. Self-organized learning and nonlinear fitting reduce the amount of computation and improve inversion efficiency. The iterative optimization unit screens key parameters, avoiding unnecessary waste of computing resources and further improving computational efficiency.

[0108] Furthermore, if Figure 8 As shown, the balance relationship unit 222 specifically includes:

[0109] The weighting subunit 2221 is used to obtain a data-driven term based on the error square of the numerical simulation calculation value and the measured value, and use the historical stratum rock surrounding rock data as a physical constraint term; and assign different weights to the data-driven term and the physical constraint term;

[0110] Among them, the constrained state phase space projection of the physical constraint term :

[0111]

[0112] Where, Λ={ , , , , } is a composite set of super parameters; Indicates η Inverse diffusion mapping of probability density in dimensional non-Euclidean space; Characterization of the historical lithologic database p Class-confined nuclear energy; The catalog operation is the property quotient on anisotropic stress manifolds; is the physical constraint elevation angle; Drive the pitch angle for data; represents the parameter vector to be optimized; express the corresponding predicted strain tensor residual; represents the matrix trace operation; represents the error tolerance exponential expansion factor; Represents the historical lithologic database p The weight coefficient of the class constraint; Indicates attribute quotient operation; represents the projection operator; represents the forward covariant gradient operator; represents the reverse covariant gradient operator; Represents the cotangent of the physical constraint elevation angle; Represents the tangent of the data-driven pitch angle;

[0113] The weight optimization subunit 2222 is used to dynamically adjust the weight parameters according to the comparison between the current inversion error and the preset threshold value during the iteration process; by quantifying the sensitivity of the parameters to the ground stress field, the key parameters are screened out and assigned higher weights; non-key parameters are eliminated;

[0114] The parameter combination and weight configuration subunit 2223 is used to combine the prior distribution of physical constraints with the data-driven likelihood function to obtain the posterior probability; optimize the weight parameters based on the posterior probability; iterate in sequence until the preset tolerance is reached or the weight change reaches a preset trend, terminate the iteration, and output the optimized parameter combination and weight configuration.

[0115] Among them, the hybrid variational value configuration of the convergence criterion of the parameter combination and weight configuration subunit 2223 :

[0116]

[0117] Where, ∈Θ is the feasible parameter half space; Define the temperature parameter of the reverse diffusion process; [·] is the event measure q The characteristic indicator functional of The corresponding impact super-relaxation Péclet number; Representational Mode q Hyperbolic mass loss differential of ; It is a tensor fusion of non-connected addition and shaped rings; is the asymptotic critical value of steady-state phase transition; represents the feasible parameter vector; represents the weight coefficient of event q; express The adjoint operator of ; represents the regularized mapping function; express controlled inverse diffusion operator; represents the optimal weight configuration tensor; represents a symbolic function; represents the phase transition index of path s; represents the inversion event measure geometry;

[0118] The mathematical architecture of the equilibrium relationship unit 222 demonstrates the cutting-edge methodology of data-physics fusion inversion, the core of which is to establish a variational assimilation framework with dynamic weight optimization.

[0119] Constrained state phase space projection The deep mechanism of non-Euclidean space inverse diffusion mapping Derived from the inversion of heat nuclear diffusion in Riemannian geometry, the probability density on the η-dimensional manifold is reversely evolved; when η=2.3 (the empirical value of shale reservoirs), the anisotropic noise in the drilling data can be effectively suppressed; the nuclear energy constraint term Based on the rock mass memory effect theory, Characterize current parameters With historical data similarity.

[0120] Angular tensor product , build physical constraints ( ∈[0,π / 2]) and data-driven ( ∈[π / 2,π]) symplectic dual space; dynamic equilibrium: when ‖ - When ‖<π / 6, the constraint relaxation mechanism is automatically triggered.

[0121] Mixed variational configuration Innovative feature, super relaxation during impact , integrating the Péclet number of fluid mechanics and the relaxation time of non-equilibrium statistical physics; When >1, the parameter update enters the impact iteration mode. Hyperbolic quality loss , reflecting the parameter space and physical manifolds The geodesic deviation of When the value is >0.1, parameter reorganization will be automatically triggered. Shaped ring fusion (⊕F), a cyclic cohomology operation in non-commutative geometry, completes tensor chain derivation with GPU acceleration.

[0122] The method utilizes a weighted dynamic game mechanism with physically constrained weights; a topological criterion for parameter screening employs persistent homology to analyze parameter importance; and a convergence criterion for phase transition detection captures topological mutations in parameter space through the sgn[ζ(s)-ψ(∞)] scheme. The breakthrough lies in transforming geological prior knowledge into a computable topologically constrained manifold while maintaining data-driven adaptive properties.

[0123] Preferably, this embodiment uses the sum of squared errors between numerically simulated and measured values ​​to derive a data-driven term, and uses historical stratum rock data as a physical constraint. Different weights are assigned to the data-driven and physical constraint terms, reflecting a dynamic balance between information from different sources and avoiding the limitations of single-source information. During the iteration process, the weight parameters are dynamically adjusted based on a comparison of the current inversion error with a preset threshold. By quantifying the sensitivity of the parameters to the geostress field, key parameters are screened and assigned higher weights, while non-critical parameters are eliminated. The prior distribution of the physical constraint is combined with the data-driven likelihood function to obtain the posterior probability. The weight parameters are optimized based on the posterior probability, and the optimal solution is gradually approached through iteration. Iterations are terminated by a preset tolerance or weight change trend to ensure that the model converges to a stable state and avoid overfitting or underfitting. The overall process embodies the concept of multi-objective optimization, namely, simultaneously considering the balance between data-driven and physical constraints. Model performance is gradually optimized by iteratively adjusting the weight parameters. This dynamic weight adjustment method can automatically optimize the weights based on the error distribution, improving the model's adaptability to complex problems. By dynamically adjusting weight parameters, the model can better fit actual observation data, reduce errors, and improve prediction accuracy; combining data-driven and physical constraints, it can fully utilize information from different sources and avoid bias from a single source, thereby improving the overall reliability of the model; dynamic weight optimization and sensitivity analysis can identify and eliminate non-critical parameters, reduce computational complexity, and improve the model's robustness to outliers; through iterative optimization and posterior probability analysis, the model can maintain stable performance under different conditions and avoid overfitting or underfitting problems. Through sensitivity analysis and dynamic weight adjustment, the model can quickly identify key parameters, reduce unnecessary calculations, and improve computational efficiency. During the iterative optimization process, the iteration is terminated by presetting tolerances or weight change trends to avoid unnecessary waste of computing resources; the dynamic weight adjustment method can flexibly adjust model parameters according to actual observation data and prior knowledge to adapt to different application scenarios.

[0124] Further, if Figure 9 As shown, the inversion result verification module 3 specifically includes:

[0125] The forward modeling verification submodule 31 is used to communicate the stress field inversion intelligent analysis model with the three-dimensional geological model, input real-time geological data into the inverse intelligent analysis model for simulation, and bring the output parameters of the inverse intelligent analysis model into the three-dimensional geological model for forward analysis and calculation;

[0126] The visualization submodule 32 is used to generate the distribution of the geostress field based on the topography, faults, and rock structure characteristics of the three-dimensional geological model and the applied self-weight pressure and tectonic stress boundary conditions; and to visualize the calculated geostress components through cloud maps;

[0127] The adjustment and optimization submodule 33 is used to compare the calculated ground stress components with the real-time measured data. If the calculated ground stress components exceed the preset limits, the model adjustment instruction is triggered. If deviations are found during the verification, the weights of the stress field intelligent analysis model or the parameters of the three-dimensional geological model are optimized, and the inversion results are iteratively updated until the preset accuracy value is met.

[0128] Preferably, this embodiment can input real-time geological data into the geostress field inversion intelligent analysis model for simulation, and feed the results back to the three-dimensional geological model for forward analysis and calculation. It can handle geostress field inversion problems under complex geological conditions, such as self-weight pressure, tectonic stress boundary conditions, etc.; intuitively display the geostress field distribution in the form of cloud maps, so that engineering personnel can quickly understand the analysis results and provide decision support; by comparing with measured data, the model weights or geological model parameters are continuously optimized until the preset accuracy value is met. It can quickly respond to real-time data input, combined with forward verification and optimization adjustment, greatly shortening the time-consuming and high-cost problems of traditional methods. Through modular design and iterative optimization mechanism, the system can adapt to complex engineering needs under different geological conditions, such as deep-buried tunnels, underground powerhouses of hydropower stations, and other engineering scenarios; the visual output in the form of cloud maps makes the geostress field distribution clear at a glance, providing an intuitive basis for engineering design and disaster prevention and control; through precise geostress field inversion and optimization adjustment, the accuracy of surrounding rock stability analysis is improved, thereby effectively reducing the risk of disasters such as rock bursts and large deformations.

[0129] like Figure 10 As shown, this embodiment also provides an embodiment of a three-dimensional in-situ stress field inversion intelligent analysis method for layered surrounding rock. In this embodiment, the three-dimensional in-situ stress field inversion intelligent analysis method for layered surrounding rock is applied to the three-dimensional in-situ stress field inversion intelligent analysis system for layered surrounding rock in the above embodiment. The three-dimensional in-situ stress field inversion intelligent analysis method for layered surrounding rock specifically includes the following steps:

[0130] Step S1: Collect geological data such as fault distribution, rock formation occurrence, physical and mechanical parameters, and measured ground stress data in the target layered surrounding rock; normalize the geological data and perform sensitivity analysis; construct a three-dimensional geological model and perform numerical simulation;

[0131] Step S2: The intelligent inversion unit generates training samples by designing parameter combinations through orthogonal experiments and combining them with numerical simulation results; generates an intelligent analysis model for geostress field inversion based on the three-dimensional geological model; and trains the intelligent analysis model for geostress field inversion using the training samples;

[0132] Step S3: input the real-time geological data into the geostress field inversion intelligent analysis model to obtain the geostress inversion analysis results; output the geostress analysis results in a visual manner; and adjust the geostress field inversion intelligent analysis model according to the measured data.

[0133] Table 1 Geological description

[0134]

[0135] Preferably, this embodiment collects geological data of the target layered surrounding rock (such as fault distribution, stratum occurrence, physical and mechanical parameters, and measured geostress data), performs normalization and sensitivity analysis, and provides standardized basic data for subsequent modeling and inversion. By constructing an accurate three-dimensional geological model, the spatial distribution characteristics of the geological body can be more comprehensively reflected. Combined with numerical simulation techniques, this method enables accurate inversion of the geostress field. Orthogonal experimental design parameter combinations are used, and training samples are generated based on numerical simulation results. An intelligent analysis model for geostress field inversion is generated based on the three-dimensional geological model. Real-time geological data input is emphasized. By feeding the latest measured data into the intelligent analysis model, model parameters are dynamically adjusted to improve inversion accuracy. This dynamic adjustment mechanism adapts to changes in complex geological conditions, ensuring the real-time and accuracy of model results. Geostress analysis results are intuitively presented through visualization technology, and the model is further optimized based on measured data. By combining numerical simulation and intelligent algorithms, this method significantly improves the accuracy of geostress field inversion. For complex geological conditions such as deep tunnels and large fault tectonic zones, this method can effectively address multivariate and nonlinear problems, providing reliable data support for engineering design. The input of real-time geological data and the dynamic adjustment mechanism of the model enable this method to adapt to changes in geological conditions and ensure the timeliness and accuracy of the inversion results. The visual output not only improves the intuitiveness of the results, but also provides a direct reference for engineering design and construction, helping to reduce engineering risks and improve economic benefits (for specific principles, please refer to the attached Figure 11 ).

[0136] like Figure 12 As shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 4 includes a processor 41 and a memory 42 coupled to the processor 41.

[0137] The memory 42 stores program instructions for implementing the layout method of the three-dimensional inversion intelligent analysis system for layered surrounding rock stress field of any of the above embodiments.

[0138] The processor 41 is used to execute program instructions stored in the memory 42 to perform the layout of the intelligent analysis system for inversion of the three-dimensional in-situ stress field of layered surrounding rocks.

[0139] The processor 41 may also be referred to as a CPU (Central Processing Unit). The processor 41 may be an integrated circuit chip with signal processing capabilities. The processor 41 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The general-purpose processor may be a microprocessor or any conventional processor.

[0140] Furthermore, Figure 13 This is a schematic diagram of the structure of a storage medium in an embodiment of the present application. The storage medium 5 in the embodiment of the present application stores program instructions 51 that can implement all of the above methods. The program instructions 51 can be stored in the above storage medium in the form of a software product and specifically include several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or terminal devices such as a computer, server, mobile phone, and tablet.

[0141] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0142] In addition, the functional units in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or in the form of software functional units. The above is only an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

[0143] The above detailed description of the specific embodiments of the invention is intended to be illustrative only, and the present invention is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of the present invention. Therefore, equivalent changes, modifications, and improvements made without departing from the spirit and scope of the present invention are also encompassed within the scope of the present invention.

Claims

1. An intelligent analysis system for inversion of three-dimensional ground stress field of layered surrounding rock, characterized by: The three-dimensional inverse intelligent analysis system for layered surrounding rock stress field includes: The 3D geological numerical module is used to collect geological data on fault distribution, rock formation occurrence, physical and mechanical parameters, and ground stress measurements in the target layered surrounding rock; normalize the geological data and perform sensitivity analysis; construct a 3D geological model and perform numerical simulation; The intelligent inversion analysis module is used to design parameter combinations for intelligent inversion through orthogonal experiments and generate training samples based on numerical simulation results; generate an intelligent analysis model for inversion of geostress field based on a three-dimensional geological model; and use the training samples to train the intelligent analysis model for inversion of geostress field. The inversion result verification module is used to input real-time geological data into the geostress field inversion intelligent analysis model to obtain the geostress inversion analysis results; output the geostress analysis results in a visual manner; and adjust the geostress field inversion intelligent analysis model based on the measured data; 3D geological numerical module, including: The rock formation classification submodule is used to import drilling data, geological exploration data, geophysical data, and measured ground stress data to define the spatial range of the 3D geological model; and to determine the range of rock mass mechanical parameters and ground stress parameters in different regions through geological zoning. The shear-compression submodule is used to perform 3D meshing and generate regular tetrahedral elements. It processes complex fault regions and assigns heterogeneous rock mechanical parameters to grid elements. It also applies normal stress to the top surface of the 3D geological model to simulate the self-weight stress field and obtain the shear-compression effect that reflects the fault structure. Adjust the 3D geological model submodule, which is used to comprehensively characterize the nonlinear superposition effect of the ground stress field through the lateral pressure coefficient and rock mass body force; compare the measured data of the ground stress components at the measuring points with the simulated values, and dynamically adjust the 3D geological model based on the comparison results and real-time geological data; Intelligent inversion analysis module, including: A sample set submodule is formed to select parameter combinations that affect the geostress field and perform numerical simulations using a three-dimensional geological model. The parameter combinations are introduced into the three-dimensional geological model to calculate the geostress field distribution results under different working conditions. A training sample set is formed based on the simulation results and the parameter combinations. Generate an inversion analysis model submodule, which is used to generate an intelligent inversion analysis model for the geostress field based on the 3D geological model; determine the center position of the radial basis function through self-organizing learning, establish a mapping relationship between input parameters and output geostress field; select key parameters and perform nonlinear fitting; The inversion analysis model optimization submodule is used to input the generated training samples into the inversion intelligent analysis model of the geostress field, adjust the fitting coefficient until it meets the preset value, and then compare the measured value with the inversion value to optimize the inversion intelligent analysis model of the geostress field to obtain the final inversion intelligent analysis model.

2. The intelligent analysis system for inversion of three-dimensional in-situ stress field of layered surrounding rock according to claim 1 is characterized in that: The rock formation classification submodule includes: The classification and sorting unit is used to import drilling data, geological survey data, and geophysical data, divide the strata according to the drilling data, unify the numbering and sorting, classify them into the same rock unit according to the lithologic pixel row and vertical position, and handle the impact of faults on the continuity of the bottom layer; The fault trajectory unit is used to determine the fault dip and extension depth based on the fault distribution and rock formation occurrence information in the geological survey data, combined with the geophysical inversion results, and generate the fault trajectory line; Define spatial range units to use borehole data to correct rock mass morphology, combine seismic interpretation horizons with velocity / density data, divide different lithologic units, fuse fault geometry with stratigraphic boundary data, and define the spatial range of the 3D geological model.

3. The intelligent analysis system for inversion of three-dimensional stress field of layered surrounding rock according to claim 1 is characterized in that: Shear extrusion submodule, including: Grid division units are used to grid the 3D geological model; process the fault intersection relationship in complex fault areas; perform local grid densification on the fault zone; and assign different rock mass mechanical parameters to corresponding grid units based on fault type and measured data. The normal constraint unit is used to apply normal stress boundary conditions on the top surface of the 3D geological model. It simulates the rock mass self-weight stress field by calculating the normal stress generated by the rock mass density; and applies normal constraints to the ground and lateral boundaries of the 3D geological model. The stress direction unit applies horizontal tectonic stress on the lateral boundary of the 3D geological model. The horizontal tectonic stress component is simulated by applying a normal unit load on the lateral boundary. The proportional relationship between horizontal stress and vertical stress is defined in combination with the lateral pressure coefficient. The tectonic stress direction is determined by inverting the measured ground stress data.

4. The intelligent analysis system for inversion of three-dimensional in-situ stress field of layered surrounding rock according to claim 1, characterized in that: Forming a sample set submodule, including: The stress field distribution unit is used to screen the key parameters that affect the geostress field in the 3D geological model and determine the parameter combination scheme; the selected parameter combination is input into the 3D geological model to perform numerical simulation calculations to obtain the geostress field distribution results under different working conditions; The parameter combination unit is used to extract the geostress components and distribution patterns of key measuring points from the 3D geological model after each numerical simulation is completed, and record the corresponding parameter combination; the parameter combination is used as the input variable, and the corresponding simulated geostress field result is used as the output variable; The training sample set storage unit is used to combine input variables and output variables into parameter ground stress field paired data; eliminate abnormal data due to unqualified parameters, and optimize the samples through iterative optimization; store the verified parameter combination and simulated ground stress field data as a training sample set in a preset format.

5. The intelligent analysis system for inversion of three-dimensional in-situ stress field of layered surrounding rock according to claim 4 is characterized in that: Generate inversion analysis model submodule, including: The weight setting unit is used to generate an intelligent analysis model for geostress field inversion based on a three-dimensional geological model; calculate the stress contribution value of each parameter combination under the simulated working condition, and determine the influence weight of the sensitivity of the geostress field distribution on the output result by quantifying the sensitivity of each parameter to the geostress field distribution; The balance relationship unit is used to sort the calculated stress contribution values ​​by magnitude and compare them with the preset error threshold. By repeatedly adjusting the parameter combination, the key parameters are retained for nonlinear fitting. The overall inversion error is evaluated at each iteration. Weight parameters are introduced to balance the relationship between data-driven and physical constraints. The iterative optimization unit is used to establish a mapping relationship between input parameters and the output of the geostress field, determine the center position of the function through self-organizing learning, and adjust the fitting coefficient based on the screened key parameter set until the inversion intelligent analysis model meets the preset three-dimensional distribution characteristic standards of the geostress field.

6. The intelligent analysis system for inversion of three-dimensional in-situ stress field of layered surrounding rock according to claim 5, characterized in that: Balance relationship unit, including: A weighting subunit is used to obtain data-driven items based on the error square of numerical simulation values ​​and measured values, and to use historical stratum and surrounding rock data as physical constraints; different weights are assigned to data-driven items and physical constraints; The weight optimization subunit is used to dynamically adjust the weight parameters according to the comparison between the current inversion error and the preset threshold during the iteration process; by quantifying the sensitivity of the parameters to the ground stress field, the key parameters are screened and assigned higher weights; non-key parameters are eliminated; The parameter combination and weight configuration subunit is used to combine the prior distribution of physical constraints with the data-driven likelihood function to obtain the posterior probability; optimize the weight parameters based on the posterior probability; iterate in sequence until the preset tolerance is reached or the weight change reaches a preset trend, terminate the iteration, and output the optimized parameter combination and weight configuration.

7. The intelligent analysis system for inversion of three-dimensional in-situ stress field of layered surrounding rock according to claim 1 is characterized in that: Inversion result verification module, including: The forward modeling verification submodule is used to communicate the stress field inversion intelligent analysis model with the 3D geological model, input real-time geological data into the inverse intelligent analysis model for simulation, and bring the output parameters of the inverse intelligent analysis model into the 3D geological model for forward analysis and calculation; The visualization submodule is used to generate the geostress field distribution based on the terrain, faults, and rock structure characteristics of the 3D geological model, as well as the applied self-weight pressure and tectonic stress boundary conditions; the calculated geostress components are visualized and output through cloud maps; The adjustment and optimization submodule is used to compare the calculated ground stress components with real-time measured data. If they exceed the preset range, the model adjustment instruction is triggered. If deviations are found during the verification, the weights of the stress field intelligent analysis model or the parameters of the three-dimensional geological model are optimized, and the inversion results are iteratively updated until the preset accuracy value is met.

8. A method for inversion intelligent analysis of three-dimensional in-situ stress field of layered surrounding rock, which is applied to the inversion intelligent analysis system of three-dimensional in-situ stress field of layered surrounding rock according to any one of claims 1 to 7, characterized in that: The intelligent analysis method for inversion of three-dimensional ground stress field of layered surrounding rock includes: Collect geological data on fault distribution, rock formation occurrence, physical and mechanical parameters, and measured ground stress data in the target layered surrounding rock; normalize the geological data and perform sensitivity analysis; construct a three-dimensional geological model and perform numerical simulation; The intelligent inversion department uses orthogonal experimental design parameter combinations and generates training samples based on numerical simulation results. It generates an intelligent analysis model for geostress field inversion based on a three-dimensional geological model. The training samples are used to train the intelligent analysis model for geostress field inversion. Input real-time geological data into the geostress field inversion intelligent analysis model to obtain geostress inversion analysis results; output the geostress analysis results in a visual format; and adjust the geostress field inversion intelligent analysis model based on the measured data; Build a 3D geological model, including: The rock formation classification submodule is used to import drilling data, geological exploration data, geophysical data, and measured ground stress data to define the spatial range of the 3D geological model; and to determine the range of rock mass mechanical parameters and ground stress parameters in different regions through geological zoning. The shear-compression submodule is used to perform 3D meshing and generate regular tetrahedral elements. It processes complex fault regions and assigns heterogeneous rock mechanical parameters to grid elements. It also applies normal stress to the top surface of the 3D geological model to simulate the self-weight stress field and obtain the shear-compression effect that reflects the fault structure. The 3D geological model adjustment submodule is used to comprehensively characterize the nonlinear superposition effect of the ground stress field through the lateral pressure coefficient and rock mass body force; compare the measured data of the ground stress components at the measuring points with the simulated values, and dynamically adjust the 3D geological model based on the comparison results and real-time geological data; Generate an intelligent analysis model for geostress field inversion based on a 3D geological model, including: The sample set submodule is used to select parameter combinations that affect the geostress field and perform numerical simulations using a three-dimensional geological model. The parameter combinations are introduced into the three-dimensional geological model to calculate the geostress field distribution results under different working conditions. A training sample set is constructed based on the simulation results and the parameter combinations. The inversion analysis model generation submodule is used to generate an intelligent inversion analysis model for the geostress field based on the 3D geological model; determine the center position of the radial basis function through self-organizing learning, establish a mapping relationship between input parameters and output geostress field; select key parameters and perform nonlinear fitting; The optimized inversion analysis model submodule is used to input the generated training samples into the inversion intelligent analysis model of the geostress field, adjust the fitting coefficient until it meets the preset value, and then compare the measured value with the inversion value to optimize the inversion intelligent analysis model of the geostress field to obtain the final inversion intelligent analysis model.

Images