Method for establishing rock multi-field coupling model and model thereof

Abstract

The application discloses a rock multi-field coupling model establishing method and model, and belongs to the technical field of engineering geological survey; through the comprehensive application of multi-scale modeling, nondestructive testing technology, machine learning algorithm and intelligent optimization, the problems of insufficient in-depth analysis of existing multi-physical field coupling, poor precision of observation method and easy damage of device are solved, the advantages of multiple observation are combined, the precision of rock multi-field coupling model establishing and analysis is improved, the intelligentization and automation level of the model are enhanced, the microstructure data and macro response are combined, a more in-depth understanding of rock behavior is provided, meanwhile, the application can provide more accurate and efficient decision support for rock engineering design and construction, and provides a new analysis and simulation method for the rock mechanics research field. Through the comprehensive scheme, the behavior of rock under complex geological conditions can be better predicted and managed, so that the safety and economy of engineering are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of engineering geological survey, and particularly relates to a method for establishing a rock multi-field coupling model and the model. BACKGROUND

[0002] The research on rock multi-field coupling problems has a history of several decades, covering the influence of single physical field, two-field coupling and three-field coupling on rock. In deep mineral resource exploitation and underground space development, the environment of rock is extremely complex, involving high temperature, high permeation pressure, high stress and complex water chemical environment. These factors jointly form the temperature-water flow-stress-chemistry (THMC) multi-field coupling effect.

[0003] Although some progress has been made in existing research, there are still deficiencies in the establishment of multi-field coupling constitutive model, the understanding of fracture evolution and deformation mechanics mechanism. Especially in deep rock mass engineering, the complexity of multi-field coupling brings challenges to the stability analysis of rock engineering, and the existing research has not fully revealed the variation law of physical and mechanical properties of rock under multi-field coupling effect.

[0004] In order to better understand the behavior of rock under multi-field coupling effect, experiment and numerical simulation are two main research means. Experimental research improves and develops multi-field coupling test system, and combines modern non-destructive detection technologies such as real-time CT scanning, SEM, NMRI, etc. to study the structure and evolution process of rock from micro to macro.

[0005] In the aspect of numerical simulation, with the development of computer technology, software such as the combination of TOUGHREACT and FLAC3D and the program of Comsol and Matlab docking provides technical support for multi-field coupling simulation (“Review of rock physical and mechanical properties and constitutive model under THMC multi-field coupling effect”, Bingqian Yan et al., Journal of Engineering Science, 2020).

[0006] For the intelligent feedback analysis method of long-term stability of rock engineering, the existing research has proposed a dynamic analysis method based on field feedback information. These methods can combine field information with geological analysis model in time, and provide suggestions for design, construction and long-term stability maintenance according to the analysis results (“Intelligent feedback analysis method and application research of long-term stability of rock engineering”, Bingrui Chen, Rock and Soil Mechanics, 2008). This method reflects the supplement and improvement of existing technology, especially in model identification and parameter identification.

[0007] The establishment of rock multi-field coupling model is the basis of research, and some patent literatures have made beneficial research on it and put forward some schemes.

[0008] For example, Chinese patent CN109284571A discloses a carbon dioxide displacement shale gas multi-scale multi-field coupling seepage mathematical modeling method. Based on the established multi-scale shale pore and fracture gas flow pattern classification model, the method establishes a multi-field coupling seepage mathematical model of carbon dioxide displacement of shale gas multi-component gas system in shale matrix nanometer pores, matrix micrometer pores and fractures, and establishes a difference numerical model in and between block center grid blocks.

[0009] In summary, the existing technology has made certain achievements in the study of the mechanical properties of rocks under multi-field coupling, but further development is still needed. Future research should focus more on in-depth analysis of multi-physical field coupling to improve the accuracy and practicality of the model.

[0010] In addition, the further development of intelligent feedback analysis methods, combined with machine learning and artificial intelligence technology, is expected to provide more efficient and accurate tools for stability analysis of rock engineering. SUMMARY

[0011] The purpose of the present application is to address the problem that the analysis of multi-physical field coupling in existing rock engineering is not deep enough. In addition to this, the advantages of non-contact observation and contact observation are combined to overcome problems such as poor accuracy and device damage in existing observation methods. A method for establishing a rock multi-field coupling model and its model are proposed. Through the comprehensive application of multi-scale modeling, non-destructive testing technology, machine learning algorithms and intelligent optimization, not only the accuracy of the establishment and analysis of the rock multi-field coupling model is improved, but also the intelligence and automation level of the model is enhanced. The microstructure data and macro response are combined to provide a deeper understanding of rock behavior. At the same time, the present application can provide more accurate and efficient decision support for rock engineering design and construction, and also provides a new analysis and simulation method for the field of rock mechanics research. Through this comprehensive solution, the behavior of rock under complex geological conditions can be better predicted and managed, thereby improving the safety and economy of the project.

[0012] To solve the above technical problems, the present application provides a method for establishing a rock multi-field coupling model and its model, comprising:

[0013] S1, obtaining rock samples by sampling, obtaining initial data and performing data preprocessing, and inputting the preprocessed data into a computer system;

[0014] S2, on a microscale, based on the obtained initial data, the micro-mechanical behavior of the rock is characterized by statistical or numerical methods;

[0015] S3, based on S2, the micro-parameters are converted to the upper scale to predict the response of the rock on a macroscale, and a multi-scale model is constructed;

[0016] S4, taking the pre-processed data as input, performing finite element analysis based on the multi-scale model to simulate the macro response of the rock under actual engineering conditions;

[0017] S5, optimizing the machine learning model parameters using a machine learning algorithm and performing model verification.

[0018] Preferably, the S1 comprises the following steps:

[0019] S11, high-resolution imaging of the rock surface and fissures, and detection of the pore structure and fluid distribution inside the rock; at the same time, three-dimensional imaging of the rock sample is performed to obtain accurate spatial distribution data of the pores and fissures;

[0020] S12, image segmentation and three-dimensional reconstruction of the pore structure of the obtained image data;

[0021] S13, normalization processing of the data obtained by S11 and S12;

[0022] S14, inputting the pre-processed data into a computer system to provide necessary input conditions for the establishment and simulation of the multi-field coupling model.

[0023] Preferably, the data pre-processing comprises image segmentation, three-dimensional reconstruction of the pore structure, and data normalization; the obtaining of the initial data comprises obtaining the microstructure data and physical property data of the rock sample through non-destructive testing techniques in the laboratory and in the field.

[0024] Preferably, the S2 comprises: analyzing the microstructure data of the rock sample through a micro-mechanical model, and quantifying the micro characteristics of the rock using a statistical method; and predicting the local stress-strain relationship through the micro-mechanical model.

[0025] The expression of the micro-mechanical model is:

[0026] σ e = σ - αp;

[0027] wherein σ is the total stress, p is the pore pressure, and α is a coefficient related to the pore geometry.

[0028] Preferably, the S3 comprises:

[0029] S31, deriving the relationship between the macro stress and the micro stress through the Hill-Mandel condition, converting the micro parameters to the macro scale to determine the macro mechanical properties, which is specifically expressed as:

[0030] <σ micro > = σ macro ; in the formula, <σ micro > is the average value of the micro stress, and σ macromacroscopic stress;

[0031] S32, based on the macroscopic mechanical properties, using standard rock mechanics test data obtained by verification.

[0032] Preferably, the standard rock mechanics test includes using uniaxial compression test data to verify the macroscopic elastic modulus, and the specific expression is:

[0033] Where, Delta sigma is the applied stress change, and Delta tau is the corresponding strain change.

[0034] Preferably, the S4 includes the following steps:

[0035] S31, based on the preprocessed data, using computer aided design software to construct the accurate three-dimensional geometric model of rock or rock structure and carry out meshing, create mesh model;

[0036] S32, based on the mesh model, define the material properties of the rock sample and apply the actual engineering boundary conditions to the mesh model;

[0037] S33, according to S32, the rock sample is loaded and response analysis is carried out, and a multi-scale model, that is, a rock multi-field coupling model, is obtained; wherein the load includes static load and dynamic load.

[0038] Preferably, the S5 includes the following steps:

[0039] S51, select a machine learning algorithm, obtain historical experimental data or field monitoring data and train a machine learning model corresponding to the machine learning algorithm;

[0040] S52, after the machine learning model training is completed, the machine learning algorithm adjusts and optimizes the parameters in the rock multi-field coupling model according to the input data and model prediction error;

[0041] S53, verify the optimized machine learning model; wherein the verification includes calculating the difference between the predicted value and the actual value.

[0042] A rock multi-field coupling model is established by a rock multi-field coupling model establishing method.

[0043] The beneficial effects of the present application are:

[0044] 1、 This solution improves the precision of rock multi-field coupling model establishment and analysis, and enhances the intelligence and automation level of the model. The implementation effect of the invention is to provide more accurate and efficient decision support for rock engineering design and construction, and also provides a new analysis and simulation method for the field of rock mechanics research. Through this comprehensive solution, the behavior of rock under complex geological conditions can be better predicted and managed, thereby improving the safety and economy of the project.

[0045] 2、 This solution provides a deeper understanding of rock behavior by combining microstructure data with macroscopic response through the application of multi-scale modeling technology. The innovation effect is to more accurately predict the mechanical response of rock in actual geological engineering, especially under complex stress paths and multi-field coupling conditions. The establishment of multi-scale models not only improves the understanding of rock failure modes, but also provides a more reliable theoretical basis for engineering design.

[0046] 3、 This solution can obtain high-resolution internal structure images of rock by using SEM, NMRI and CT scanning technology, which are directly used for model initialization and verification. This advanced data acquisition method enables the model to simulate based on the real microstructure of rock, improving the prediction accuracy and reliability of the model. In addition, the application of non-destructive testing technology also reduces the bias and damage that may be introduced by traditional sampling and testing.

[0047] 4、 This solution uses machine learning algorithms to intelligently optimize model parameters, learning and identifying the macroscopic mechanical properties of rock such as elastic modulus and creep parameters from a large amount of data. Compared with traditional methods based on experience or theoretical derivation, machine learning algorithms can more quickly and accurately adjust model parameters to match actual observation data, thereby significantly improving the prediction accuracy and adaptability of the model. BRIEF DESCRIPTION OF DRAWINGS

[0048] Other features, objects and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments made with reference to the accompanying drawings. The drawings are for the purpose of illustrating preferred embodiments only and are not to be construed as limiting the present invention. Moreover, the same reference numerals are used throughout the accompanying drawings to refer to the same or like parts.

[0049] Figure 1 Flowchart for the embodiments of the present invention. DETAILED DESCRIPTION

[0050] In order to make the objects, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with the drawings and examples. It should be understood that the specific implementation described herein is only one of the best embodiments of the present application, which is only used to explain the present application and does not limit the protection scope of the present application. All other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0051] Embodiment 1: As shown in the figure, a method for establishing a rock multi-field coupling model and the model thereof, comprising: Figure 1

[0052] S1, initial data collection and data preprocessing, including image segmentation, three-dimensional reconstruction of pore structure and data normalization. Initial data collection includes obtaining microstructure and physical property data of rock samples through non-destructive testing techniques such as SEM, NMRI and CT scanning in the laboratory and on site. These data will provide key information about the internal pore, fracture distribution and rock type of the rock for the model. Subsequently, data preprocessing is performed to ensure data quality, including image segmentation, three-dimensional reconstruction of pore structure and data normalization. Then the preprocessed data is input into the multi-field coupling model, including the microstructure parameters and macroscopic physical properties of the rock.

[0053] Specifically, the method comprises the following steps:

[0054] S11, the initial data collection process first involves obtaining high-quality microstructure images from rock samples. Scanning electron microscopy (SEM) is used to image the rock surface and fractures at high resolution to capture the micro features of the rock. Nuclear magnetic resonance imaging (NMRI) technology is used to detect the pore structure and fluid distribution inside the rock. In addition, computed tomography (CT) technology is used to perform three-dimensional imaging of the rock sample to obtain accurate spatial distribution data of the pores and fractures. During the acquisition process, the imaging parameters and conditions of each technology need to be recorded to ensure the consistency and comparability of the data.

[0055] S12, the obtained image data then enters the preprocessing stage, including image segmentation and three-dimensional reconstruction of pore structure. Image segmentation is to distinguish the solid matrix and pore space of the rock using image processing techniques, which can be achieved by threshold segmentation method, and the basic formula is:

[0056] T = threshold;

[0057] V = {x | I(x) > T};

[0058] ​where T is the segmentation threshold, I(x) is the intensity value of the image at point x, and V is the segmented pore region. Three-dimensional reconstruction is then performed by converting two-dimensional image data into a three-dimensional model, using voxelization methods to convert tomographic data into a three-dimensional mesh model, providing a geometric basis for subsequent finite element analysis. In image processing, "threshold" is a parameter used for image segmentation, which maps pixel values of an image into two classes in a binary image (usually black and white). Thresholding is a simple method that determines whether each pixel belongs to the foreground or background based on its intensity value.

[0059] Specifically, for each pixel in a grayscale image, if its grayscale value is higher than or equal to the threshold (T), the pixel is set to white (or some high value, representing the foreground) in the binary image; if the pixel's grayscale value is lower than the threshold, the pixel is set to black (or some low value, representing the background) in the binary image.

[0060] The choice of threshold depends on the characteristics of the image and the goal of segmentation, which can be global (using a single threshold for the entire image) or local (using different thresholds for different regions of the image). The determination of the threshold can be based on histogram analysis of the image, Otsu method or other more advanced algorithms.

[0061] In the segmentation of rock microstructure images, threshold segmentation can help distinguish the pores and fractures of the rock from the rock matrix, thus providing accurate binary images for subsequent three-dimensional reconstruction and analysis.

[0062] S13, after completing image segmentation and three-dimensional reconstruction, data normalization is performed to eliminate the dimensional and order differences between different rock samples or different imaging techniques. The normalization formula is:

[0063]

[0064] Here, X is the original data, X min and X max are the minimum and maximum values in the data set, respectively, and X n is the normalized data. Normalization ensures that data from different sources can be compared and analyzed on a unified scale. Finally, the preprocessed data is input into the multi-field coupling model, including the microstructure parameters and macroscopic physical properties of the rock, providing necessary input conditions for model establishment and simulation.

[0065] S2, after collecting and preprocessing the data, the next step is to build a multi-scale model. First, at the microscopic scale, based on the obtained microstructure data, statistical methods or numerical methods are used to characterize the micro-mechanical behavior of the rock. Then, the micro-parameters are up-scaled to predict the response of the rock at the macroscopic scale. This step involves determining the macro-mechanical properties of the rock, such as the elastic modulus and strength parameters, which can be inferred from the microstructure through appropriate homogenization techniques.

[0066] Specifically, the following steps are included:

[0067] S21, through the analysis of the microstructure data of the rock sample, statistical methods (such as maximum likelihood estimation or Bayesian inference) are used to quantify the micro-characteristics of the rock (such as porosity and fracture density). These characteristics are then used to define the micro-mechanical behavior of the rock, and then the local stress-strain relationship is predicted through the micro-mechanical model.

[0068] The micro-mechanical model used in the present application is the Hill average stress formula, which is used to estimate the effective stress, and its expression is:

[0069] σ e = σ - αp;

[0070] where σ is the total stress, p is the pore pressure, and α is a coefficient related to the pore geometry.

[0071] The value of α usually ranges from 0 to 1. When α = 0, it means that the pore pressure has no effect on the effective stress; when α = 1, it means that all the pore pressure is converted into effective stress, which may occur in completely saturated rocks or soils.

[0072] The geometry and connectivity of the pores affect the way fluid pressure acts on the rock skeleton. A more connected pore network may result in a higher α value. The permeability of the rock affects the flow of pore fluid, thereby affecting the value of α. High-permeability rocks may allow more effective pressure transmission, resulting in an α value closer to 1. The loading history and path of the rock also affect the value of α. For example, under cyclic loading or complex loading paths, the pore structure of the rock may change, thereby affecting α.

[0073] S22, the micro-parameters are up-scaled to predict the response of the rock at the macroscopic scale. Scale conversion is a key step in converting the mechanical behavior at the microscopic scale to the macroscopic scale. The present application uses the Hill-Mandel condition to derive the relationship between the macroscopic stress and the microscopic stress, which can be expressed as:

[0074] <σ micro > = σ macro ;

[0075] Here, <σ micro > is the average value of microscopic stress, σ macro is the macroscopic stress. By this method, the equivalent properties of macroscopic materials can be calculated.

[0076] S23, after determining the macroscopic mechanical properties, validation through experimental data is performed. This includes using standard rock mechanics tests, such as uniaxial compression tests and triaxial compression tests, to measure the macroscopic response of the rock and compare it with the model predictions. For example, one feasible method is to use the uniaxial compression test data to validate the macroscopic elastic modulus, which is given by the formula:

[0077]

[0078] where Δσ is the applied stress change and Δτ is the corresponding strain change. If the model-predicted macroscopic properties match the experimental data, this validates the effectiveness of the multiscale model. Conversely, if there are discrepancies, adjustments to the microscopic parameters or homogenization process are needed to ensure the accuracy of the model.

[0079] S3, after the multiscale model is established, finite element analysis (FEA) is implemented to simulate the macroscopic response of the rock under actual engineering conditions. This includes creating a geometric model of the rock, dividing the finite element mesh, defining material properties, applying boundary conditions and loads. Creating a geometric model of the rock is the application of the macroscopic mechanical properties obtained from the multiscale model to the finite element analysis (FEA) of the actual rock structure. In this step, special attention should be paid to ensure that the geometric size, material properties and load conditions of the model can truly reflect the actual engineering situation. Through FEA, the stress, strain distribution and possible failure mode of the rock under the action of multiple fields can be predicted.

[0080] S31, based on the data obtained from geological exploration and mapping, use computer-aided design (CAD) software to construct an accurate three-dimensional geometric model of the rock or rock structure. The model should include all relevant geological features, such as cracks, faults and joints. Next, meshing is performed, dividing the geometric model into a finite number of small elements, such as tetrahedrons or hexahedrons. The quality of meshing directly affects the accuracy and efficiency of FEA. In this step, appropriate mesh density is used to ensure the accuracy of the calculation while avoiding unnecessary increase in computational load.

[0081] S32, After the mesh model is created, material properties are defined for the rock material. This includes elastic modulus (E), Poisson's ratio (v), shear modulus (G), and strength parameters such as cohesion (c) and internal friction angle (φ). These parameters can be obtained through laboratory tests and derived from the microscopic scale through homogenization methods. Subsequently, actual engineering boundary conditions need to be applied on the model, including fixed boundaries, free surfaces, stress boundary conditions or displacement boundary conditions. These conditions simulate the constraints and loading of the rock in the actual geological environment.

[0082] S33, After the material properties and boundary conditions are defined, the loading is applied and the response analysis is performed. The load can be static, such as self-weight stress, or dynamic, such as seismic load. In the case of multi-field coupling, the effects of temperature field, water flow field and chemical field also need to be considered. Among them, temperature change causes additional thermal stress through thermal expansion coefficient (α th ), the calculation formula is:

[0083] σ th = α th ΔT; in the formula, ΔT is the temperature change. σ th is the thermal stress caused by temperature change.

[0084] FEA software will use these input data to calculate the stress and strain distribution inside the rock, and predict the possible failure mode, such as shear failure or tensile failure. Through post-processing tools, the analysis results can be visualized to evaluate the stability and safety of the rock structure.

[0085] S4, The last step is to use machine learning algorithms to intelligently optimize the model parameters and perform model validation. By inputting experimental data or field monitoring data into the machine learning model, automatically adjust the parameters in the rock multi-field coupling model, such as permeability coefficient and creep parameter, to improve the prediction accuracy of the model. In addition, use independent data sets or through comparative analysis with actual engineering cases to verify the accuracy and reliability of the model. Through this step, it can be ensured that the model is not only theoretically sound, but also effective in practical application.

[0086] S41, The beginning of the intelligent optimization process is to select a suitable machine learning algorithm. Common algorithms include neural networks, support vector machines (SVM), decision trees and random forests, etc. The choice of algorithm depends on the characteristics of the data and the optimization goal. Once the algorithm is selected, the next step is the training phase, in which historical experimental data or field monitoring data are used to "train" the machine learning model. This process involves dividing the data set into training and test sets, using the training set to adjust the model parameters to minimize the prediction error. For example, for neural networks, the weights and biases can be updated through the backpropagation algorithm, whose basic update rule can be expressed as:

[0087]

[0088] where w is the weight, η is the learning rate, F is the error function, and t is the iteration number.

[0089] S42, After the model training is completed, the machine learning algorithm can automatically adjust the parameters in the rock multi-field coupling model. For example, the permeability coefficient k and the creep parameters A, n (usually used in the creep law) can be optimized according to the input data and the model prediction error. This process can be achieved through gradient descent or other optimization algorithms, with the goal of finding a set of parameter values that minimize the difference between the model predictions and the actual observed data. The optimization of creep parameters can be based on the integral form of the creep law, such as:

[0090]

[0091] where μ c is the creep strain, σ is the applied stress, and B is a time-dependent constant.

[0092] In the integral form of the creep law, the determination of the time-dependent constant B is usually based on experimental data or obtained by fitting the actual observed creep behavior. The following are several steps and methods for determining the time-dependent constant B, one of which can be selected according to the actual engineering situation or conditions, or a combination of two or more methods:

[0093] (1) Creep test: First, perform a creep test to obtain the creep strain data of the rock or material under different stress levels. Then collect the strain readings at different time points in the creep test, as well as the corresponding stress levels. Use nonlinear least squares or other curve fitting techniques to fit the test data with the creep law prediction, thereby determining the parameters A, n and B.

[0094] (2) Numerical optimization method

[0095] Gradient descent method: Use the gradient descent method to optimize the creep model parameters, update the parameter values by calculating the gradient of the loss function (such as mean square error) with respect to the parameters.

[0096] Global optimization algorithm: Use global optimization algorithms (such as particle swarm optimization, genetic algorithm) to avoid local minimum, ensure to find the global optimal solution.

[0097] (3) Parameter sensitivity analysis

[0098] Sensitivity analysis: Perform parameter sensitivity analysis to determine the degree of influence of each parameter on model prediction, which helps to identify the most critical parameters for model output.

[0099] (4) Adjustment in actual application

[0100] Actual data calibration: In practical applications, parameter B is adjusted based on the latest monitoring data or engineering feedback to ensure the continued accuracy and applicability of the model.

[0101] S43. Finally, the intelligently optimized model needs to be validated to test its accuracy and reliability. This can be done by comparing the model's predictions with observations from independent test datasets or real-world engineering cases. The validation process includes calculating the difference between predicted and actual values, such as the mean squared error (MSE) or the coefficient of determination (R²). 2 For example, MSE can quantify prediction error, defined as:

[0102]

[0103] Where N is the number of data points, P rei and A ci These are the predicted and actual values ​​for the i-th data point, respectively. A low MSE indicates that the model's predictions match the actual observations, thus validating the model's effectiveness. This step ensures the reliability and effectiveness of the model in practical engineering applications.

[0104] The above-described specific embodiments are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape, structure, and method of the present invention are within the protection scope of the present invention.

Claims

1. A method for establishing a multi-field coupling model of rock, characterized in that, Includes the following steps: S1. Obtain rock samples by sampling, acquire initial data and perform data preprocessing, and input the preprocessed data into the computer system; S2. At the microscale, based on the acquired initial data, the micromechanical behavior of rocks is characterized using statistical or numerical methods. S2 includes: The microstructure data of rock samples were analyzed by micromechanical models, the microscopic properties of rocks were quantified by statistical methods, and the local stress-strain relationship was predicted by micromechanical models. The expression for the micromechanical model is: ； Where σ is the total stress, p is the pore pressure, and α is a coefficient related to the pore geometry. It is the effective stress; S3. Based on S2, the microscopic parameters are transformed to the macroscale to predict the response of rocks at the macroscale and construct a multiscale model; S3 includes: S31. The relationship between macroscopic and microscopic stresses is derived using the Hill-Mandel condition, and the microscopic parameters are converted to the macroscopic scale to determine the macroscopic mechanical properties, specifically expressed as follows: ； In the formula, It is the average value of micro-stress. It is macroscopic stress; S32. Based on the aforementioned macroscopic mechanical properties, the experimental data obtained from standard rock mechanics tests are used for verification. S4. Using the preprocessed data as input, perform finite element analysis based on the multi-scale model to simulate the macroscopic response of rocks under actual engineering conditions; S5. Optimize the parameters of the machine learning model using machine learning algorithms and validate the model.

2. The method for establishing a multi-field coupling model of rock according to claim 1, characterized in that, S1 includes the following steps: S11. Perform high-resolution imaging of the rock surface and fissures, and detect the pore structure and fluid distribution inside the rock; at the same time, perform three-dimensional imaging of the rock sample to obtain accurate spatial distribution data of pores and fissures. S12. Perform image segmentation and three-dimensional reconstruction of the pore structure on the acquired image data; S13. Normalize the data obtained in S11 and S12; S14. Input the preprocessed data into the computer system to provide the necessary input conditions for the establishment and simulation of the multi-field coupling model.

3. The method for establishing a multi-field coupling model of rock according to claim 2, characterized in that: The data preprocessing includes image segmentation, three-dimensional reconstruction of pore structure, and data normalization; the acquisition of initial data includes obtaining microstructure and physical property data of rock samples through non-destructive testing techniques in the laboratory and on-site.

4. The method for establishing a multi-field coupling model of rock according to claim 1, characterized in that: The standard rock mechanics test includes using uniaxial compression test data to verify the macroscopic elastic modulus. The specific expression is: ； Where Δσ is the applied stress change, It is the corresponding strain change, It is the macroscopic elastic modulus.

5. The method for establishing a multi-field coupling model of rock according to claim 1, characterized in that, S4 includes the following steps: S41. Based on the preprocessed data, use computer-aided design software to construct an accurate three-dimensional geometric model of the rock or rock structure and perform mesh generation to create a mesh model. S42. Define material properties for rock samples based on the mesh model and apply actual engineering boundary conditions to the mesh model; S43. Apply loads to the rock sample according to S32 and perform response analysis; wherein the loads include static loads and dynamic loads.

6. The method for establishing a multi-field coupling model of rock according to claim 1, characterized in that, S5 includes the following steps: S51. Select a machine learning algorithm, obtain historical experimental data or on-site monitoring data, and train the machine learning model corresponding to the machine learning algorithm. S52. After the machine learning model is trained, the machine learning algorithm adjusts and optimizes the parameters in the rock multi-field coupling model based on the input data and the model prediction error. S53. Validate the optimized machine learning model; whereby validation includes calculating the difference between the predicted value and the actual value.

7. A multi-field coupling model for rocks, characterized in that... It is established by the method described in claim 1.

Images