Rock mass three-dimensional joint network intelligent modeling and dynamic updating method and system

Abstract

The invention relates to the technical field of rock mass engineering geological exploration and digital rock mass, and particularly provides a rock mass three-dimensional joint network intelligent modeling and dynamic updating method and system. The core process of the method comprises the steps of constructing an integrated multi-physics field detection system, synchronously collecting rock mass surface joints, borehole internal joints and wave velocity field data, evaluating the quality and reliability of multi-source data, implementing super-resolution optimization for low-resolution data, and forming a reference factor weight system and a high-quality preprocessing data set. And dynamic updating and visualization of an imaging result are realized by utilizing a newly added drilling data dynamic verification and correction model. According to the method, the inherent defects of data fragmentation, model staticization, lack of physical constraint, update lag and the like in a traditional method are effectively overcome, and reliable technical support is provided for blasting excavation design, stability analysis and risk management and control of rock mass engineering.

Description

Technical Field

[0001] This invention belongs to the field of rock mass engineering geological exploration and digital rock mass technology, specifically relating to a method and system for intelligent modeling and dynamic updating of three-dimensional joint networks in rock masses. In particular, this invention achieves high-precision, physically consistent, and dynamically optimizable three-dimensional visualization of the joint and fracture network within the rock mass by integrating multi-physics field synchronous detection, deep learning-driven modeling, and incremental learning-based dynamic updating. Background Technology

[0002] In rock engineering practice, key aspects such as bench blasting, tunnel excavation, slope stability evaluation, and underground cavern development all rely on a precise understanding of the spatial distribution, geometric morphology, and connectivity characteristics of discontinuous structural surfaces such as joints and fissures within the rock mass. These structural features directly control the mechanical response and overall stability of the rock mass. Therefore, achieving high-precision, true three-dimensional, and dynamically updated digital reconstruction of the joint and fissure network within the rock mass has become a core prerequisite for ensuring engineering safety and optimizing structural design.

[0003] However, the existing detection and imaging technology system still has significant limitations in dealing with complex rock mass structures and dynamic construction environments, specifically as follows: (1) The data acquisition process is scattered and independent, lacking systematic integration. Existing methods usually adopt a step-by-step approach, such as obtaining surface joints through three-dimensional laser scanning, extracting the internal structure through borehole television, and detecting the physical field of the rock mass through seismic waves or resistivity methods. This heterogeneous and asynchronous operation mode leads to inconsistent spatiotemporal references between data, destroying the integrity and internal correlation of rock mass structure information, and restricting the collaborative utilization of multi-source information; (2) The multi-source data fusion mechanism is weak, making it difficult to achieve effective complementarity. Due to the essential differences in physical nature, scale characteristics and resolution of data such as optical images, elastic wave fields, and resistivity fields, and the presence of different types of noise interference, traditional fusion methods lack a unified inversion framework and adaptive registration mechanism, resulting in serious information loss during the integration process, and even introducing coupling errors, making it impossible to construct a rock mass structure model with consistency and integrity; (3) The modeling method is insufficient in characterizing the nonlinearity of the structure, limiting its predictive ability. Traditional inversion theories are mostly based on linear or simple nonlinear assumptions, which makes it difficult to truly reflect the non-uniform distribution, morphological complexity and mechanical coupling behavior of joints and fractures in space. This causes the model to deviate from reality in deduction and prediction, thus limiting its engineering guidance value. (4) The model update mechanism is rigid and difficult to adapt to the dynamic engineering process. Most existing three-dimensional models are static results, and once established, they are difficult to evolve synchronously with the construction process. Model correction relies on manual intervention and repeated exploration, which is slow and inefficient, and cannot provide real-time data support for dynamic design and risk prevention. (5) The visualization is simple and lacks interactivity and immersion. Imaging results are often presented in the form of two-dimensional maps or static three-dimensional meshes, which lack multi-dimensional and interactive visual expression capabilities, limiting the efficiency of engineers' spatial cognition and analysis of complex fracture systems.

[0004] In summary, existing technologies still have systemic problems in areas such as the synchronization of data acquisition, the quality of multi-source information fusion, the physical rationality of model construction, the adaptive capability of dynamic updates, and the engineering applicability of visualization output. Therefore, there is an urgent need to develop a three-dimensional imaging method for rock mass structures that can integrate multi-source heterogeneous data, embed rock mechanics mechanisms, and possess dynamic learning and evolution capabilities, thereby promoting the digitalization and intelligentization of rock engineering. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides an intelligent modeling and dynamic updating method and system for three-dimensional joint networks in rock masses. This method solves the problems of asynchronous data acquisition, low accuracy of multi-source data fusion, lack of geomechanical constraints in model construction, difficulty in dynamically updating imaging results, and unintuitive visualization effects inherent in traditional methods, thus failing to meet the high-precision detection requirements of complex rock mass engineering. To achieve the above objectives, this invention adopts the following technical solution: The method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock masses includes the following steps: Simultaneously acquiring surface joint data, borehole internal joint data, and wave velocity field imaging data of the rock mass by integrating a multi-physics detection module; processing borehole images to extract joint features, resulting in a multi-source dataset and a joint database; evaluating data quality by analyzing the accuracy, completeness, and resolution of the multi-source dataset; optimizing low-resolution data to obtain baseline factor weights and a preprocessed dataset; extracting joint features using a convolutional neural network; combining geological background and mechanical properties; constructing an initial joint network model using a generative adversarial network to obtain the initial three-dimensional imaging results of the joint network; establishing a stress wave propagation model in jointed rock masses; using the initial three-dimensional joint network model as input to simulate the stress wave propagation process and outputting theoretical wave velocity field distribution characteristics; introducing wave velocity field data as constraints into the model; fine-tuning parameters using a conditional generative adversarial network to obtain a high-precision joint network model under wave velocity field constraints; verifying the model accuracy by adding new borehole data; dynamically correcting parameters using incremental learning to obtain three-dimensional imaging results, achieving visualization of the joint fracture network.

[0006] Furthermore, the method of simultaneously acquiring surface joint, borehole internal joint, and wave velocity field data through an integrated multiphysics detection module, processing borehole images to extract joint features, and obtaining a multi-source dataset and joint database includes the following steps: Using an integrated multiphysics detection module, surface joint data, borehole internal joint data, and wave velocity field data are simultaneously acquired; the multiphysics detection module includes a 3D laser scanner, borehole television, a standard drill pipe, an ultrasonic transmitter, and a receiver; preprocessing the acquired borehole images, including grayscale conversion to reduce data complexity, Gaussian filtering to remove noise, histogram equalization to enhance contrast, and edge detection algorithms to extract geometric and spatial feature parameters of the joints; the spatial feature parameters include the number of joints, joint length, joint width, joint development length, joint depth, and joint location; integrating the processed data with the extracted joint features to obtain a dataset containing multi-dimensional information, constructing a detailed joint database, and obtaining a multi-source dataset and joint database.

[0007] Furthermore, the process of analyzing the accuracy, completeness, and resolution of multi-source datasets to evaluate data quality, optimize low-resolution data, and obtain benchmark factor weights and a preprocessed dataset includes the following steps: employing data analysis methods to comprehensively evaluate data quality by deeply analyzing the accuracy, completeness, and resolution of surface joint, borehole joint, and wave velocity field data in the multi-source dataset; optimizing low-resolution data areas using super-resolution reconstruction technology to improve overall data accuracy; integrating the evaluation results with the optimized data, extracting key influencing factors and assigning weights to them, resulting in a benchmark factor weight system and a rigorously preprocessed dataset.

[0008] Furthermore, the establishment of a stress wave propagation model in jointed rock mass, using an initial three-dimensional joint model as input, simulates the stress wave propagation process and outputs the theoretical wave velocity field distribution, includes the following steps: based on the actual characteristics of the rock mass, selecting numerical simulation software and calculation methods, constructing a rock mass geometric model, and embedding the initial three-dimensional joint model into the rock mass geometric model; setting the density and elastic modulus of the rock mass, as well as the stiffness and friction coefficient of the joints; inputting excitation source information, running the simulation program, and simulating the stress wave propagation process in jointed rock mass; and processing the simulation results to obtain the theoretical wave velocity field distribution.

[0009] Furthermore, the method of extracting joint features through convolutional neural networks, combining geological background and mechanical properties, and constructing an initial joint network model using generative adversarial networks to obtain three-dimensional imaging results of joints includes the following steps: using convolutional neural networks to extract deep features from preprocessed multi-source data, automatically capturing the geometric morphology and spatial distribution patterns of joints through multi-layer convolution and pooling operations; fusing the extracted joint features with the geological background and mechanical property parameters of the rock mass to determine the constraints for model construction; constructing a three-dimensional mathematical model of joints based on the spatial feature parameters of joints, using multivariate surface equations to characterize the shape distribution and development characteristics of joints; using the dynamic game mechanism of the generator and discriminator in the generative adversarial network to iteratively optimize the model structure, transforming the feature data into the topological relationship of the joint network in three-dimensional space; using the feature data of the three-dimensional mathematical model of joints, the wave velocity field distribution model generated by stress waves passing through the rock mass, and the attenuation terminal velocity as the result data, and using the theoretical formula of rock mass stress wave propagation as the loss function. A generative adversarial network model is constructed; by training the model, the feature parameters of the three-dimensional mathematical model of joints are continuously adjusted so that the error of the model loss function is controlled within the allowable range, the model parameters are obtained, and the three-dimensional imaging results of joints are obtained.

[0010] Furthermore, the method of introducing wave velocity field data as a constraint into the model and fine-tuning the parameters using a conditional generative adversarial network to obtain a high-precision joint network model under wave velocity field constraints includes the following steps: Using data fusion technology, wave velocity field data is introduced as a key constraint into the initial joint network model, and the model parameters are corrected using the anomalous variation characteristics of wave velocity at the joints; using a conditional generative adversarial network, through an adversarial training mechanism between the generator and the discriminator, the model parameters are fine-tuned under wave velocity field constraints to ensure that the spatial distribution of the joint network highly matches the wave velocity propagation law; defining the objective function of the inversion problem as the difference between the measured wave velocity data and the forward simulation results; and reconstructing the optimized model parameters in three-dimensional space to obtain a rock mass internal joint network model with high precision and physical consistency under wave velocity field constraints.

[0011] Furthermore, the process of verifying model accuracy using newly added borehole data, employing incremental learning to dynamically correct parameters, and obtaining 3D imaging results for immersive visualization includes the following steps: Using measured data collected from newly added boreholes, comparing and analyzing the joint distribution at corresponding locations in the 3D model to extract error features and quantify model accuracy; inputting the verification results into an incremental learning framework to dynamically adjust neural network parameters to correct model deviations, ensuring the imaging results continuously match actual geological conditions; importing the optimized 3D model into a virtual reality engine, and achieving immersive visualization through real-time rendering and interactive design, supporting multi-angle observation and geological analysis, and obtaining dynamic imaging results of internal joints and fractures in the rock mass that combine accuracy and intuitiveness.

[0012] The second aspect of this invention provides an intelligent modeling and dynamic updating system for three-dimensional rock mass joint networks. This system includes the following modules: a multi-source dataset module, used to simultaneously collect surface joint, borehole internal joint, and wave velocity field data through an integrated multi-physics detection module, process borehole images to extract joint features, and obtain a multi-source dataset and a joint database; a preprocessing module, used to evaluate data quality by analyzing the accuracy, completeness, and resolution of the multi-source dataset, optimize low-resolution data, and obtain baseline factor weights and a preprocessed dataset; and a joint network module, used to extract joint features through a convolutional neural network, combining geological background and mechanical properties, and utilizing... The system employs a generative adversarial network (GAN) to construct an initial joint network model, yielding 3D joint imaging results. A simulation propagation module establishes a stress wave propagation model within jointed rock masses, using the initial 3D joint model as input to simulate the stress wave propagation process and output the theoretical wave velocity field distribution and attenuation characteristics. A model optimization module introduces wave velocity field data as constraints into the model and fine-tunes parameters using a conditional GAN, resulting in a high-precision joint network model constrained by the wave velocity field. A model verification module verifies the model's accuracy using newly added borehole data, employs incremental learning to dynamically correct parameters, and obtains 3D imaging results, enabling immersive visualization.

[0013] A third aspect of the present invention provides an intelligent modeling and dynamic updating device for three-dimensional joint networks of rock mass, the device comprising a memory and at least one processor, the memory storing instructions; the at least one processor calling the instructions in the memory to cause the device to perform the steps of the intelligent modeling and dynamic updating method for three-dimensional joint networks of rock mass as described in any of the preceding claims.

[0014] In the technical solution provided by this invention, multi-physics field detection modules are integrated to simultaneously collect imaging data of rock surface joints, borehole internal joints, and wave velocity fields. Borehole images are processed to extract joint features, resulting in a multi-source dataset and a joint database. The accuracy, completeness, and resolution of the multi-source datasets are analyzed to evaluate data quality, and low-resolution data is optimized to obtain baseline factor weights and a preprocessed dataset. Joint features are extracted using a convolutional neural network, and combined with geological background and mechanical properties, an initial joint network model is constructed using a generative adversarial network (GAN) to obtain the initial joint network's three-dimensional imaging results. A stress wave propagation model in jointed rock masses is established, using the initial joint network's three-dimensional model as input to simulate the stress wave propagation process and output theoretical wave velocity field distribution characteristics. Wave velocity field data is introduced as a constraint condition into the model, and a conditional GAN ​​is used to fine-tune the parameters, resulting in a high-precision joint network model under wave velocity field constraints. The accuracy of the model is verified by adding new borehole data, and incremental learning is used to dynamically correct the parameters, obtaining the three-dimensional imaging results and realizing the visualization of the joint fracture network. This invention solves the problems of asynchronous data acquisition, low accuracy of multi-source data fusion, lack of geomechanical constraints in model construction, difficulty in dynamic updating of imaging results, and unintuitive visualization effects in traditional methods, which make it difficult to meet the high-precision detection requirements of complex rock mass engineering. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 This is a schematic diagram of the first embodiment of a method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of a second embodiment of a method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of the third embodiment of a method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to an embodiment of the present invention.

[0019] Figure 4 This is a schematic diagram of the fourth embodiment of a method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to the present invention.

[0020] Figure 5 This is a schematic diagram of the fifth embodiment of a method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0022] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0023] A method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock masses, such as Figure 1 As shown, the process includes the following steps: Simultaneously acquiring surface joint data, borehole joint data, and wave velocity field imaging data from an integrated multi-physics detection module; processing borehole images to extract joint features, resulting in a multi-source dataset and a joint database; evaluating data quality by analyzing the accuracy, completeness, and resolution of the multi-source dataset, optimizing low-resolution data, and obtaining baseline factor weights and a preprocessed dataset; extracting joint features using a convolutional neural network, combining geological background and mechanical properties, and constructing an initial joint network model using a generative adversarial network to obtain the initial joint network's three-dimensional imaging results; establishing a stress wave propagation model in jointed rock masses, using the initial joint network's three-dimensional model as input to simulate the stress wave propagation process, and outputting theoretical wave velocity field distribution characteristics; introducing wave velocity field data as constraints into the model, and fine-tuning parameters using a conditional generative adversarial network to obtain a high-precision joint network model under wave velocity field constraints; verifying the model's accuracy by adding new borehole data, dynamically correcting parameters using incremental learning, obtaining three-dimensional imaging results, and realizing the visualization of the joint fracture network.

[0024] Incremental learning is an intelligent machine learning method that allows models to continuously absorb information from new data to update and optimize themselves based on existing knowledge, without retraining the entire model. This characteristic makes it highly advantageous in scenarios involving ever-growing data, such as rock mass monitoring. As new borehole data is continuously acquired, incremental learning enables models to quickly adapt to new data, dynamically adjust internal parameters, and accurately capture the changing characteristics of joints and fractures. It not only saves significant computational resources and time but also effectively avoids catastrophic forgetting problems, ensuring that the model maintains high accuracy and stability.

[0025] like Figure 2As shown, in this embodiment, an integrated multiphysics detection module is used to simultaneously acquire surface joint data, borehole internal joint data, and wave velocity field data. The multiphysics detection module includes a 3D laser scanner, borehole television, a standard drill pipe, an ultrasonic transmitter, and a receiver. The acquired borehole images are preprocessed, including grayscale conversion to reduce data complexity, Gaussian filtering to remove noise, histogram equalization to enhance contrast, and edge detection algorithms to extract geometric and spatial feature parameters of the joints. Spatial feature parameters include the number of joints, joint length, joint width, joint development length, joint depth, and joint location. The processed data is integrated with the extracted joint features to obtain a dataset containing multi-dimensional information, and a detailed joint database is constructed to obtain a multi-source dataset and a joint database.

[0026] Employing an integrated multiphysics detection module, this system can simultaneously acquire joint data and wave velocity field data from both the surface and borehole interiors, ensuring temporal and spatial consistency of multi-source data and effectively solving the problem of data fusion difficulties. Image processing algorithms are used to extract joint geometric and spatial feature parameters, which are then integrated with the processed data to construct a detailed joint database. This embodiment improves the comprehensiveness and accuracy of detection, providing precise data support for rock engineering and contributing to enhanced construction safety and efficiency.

[0027] like Figure 3 As shown, in this embodiment, a data analysis method is used to comprehensively evaluate the data quality by deeply analyzing the accuracy, completeness, and resolution of surface joints, borehole joints, and wave velocity field data in the multi-source dataset. For low-resolution data areas, super-resolution reconstruction technology is used for optimization to improve the overall accuracy of the data. The evaluation results are integrated with the optimized data, key influencing factors are extracted and assigned weights to obtain a benchmark factor weight system and a dataset that has undergone rigorous preprocessing.

[0028] When analyzing multi-source data using data analysis methods, for accuracy, taking surface joint data as an example, the measured values ​​are compared with high-precision laser scanning results; an error within 5% is considered accurate. Regarding completeness, if borehole joint data is missing more than 10% at a certain depth, that segment is considered incomplete. For resolution, if wave velocity field imaging cannot distinguish joints smaller than 5cm, it is considered low resolution. For low-resolution areas, cubic spline interpolation is used to enhance data continuity, or deep learning-based super-resolution reconstruction techniques are employed to improve detail. Subsequently, the evaluation results are integrated with optimized data. For example, considering the impact of joint orientation on rock mass stability, a weight of 0.3 is assigned based on expert scoring and the analytic hierarchy process, ultimately forming a baseline factor weighting system and a preprocessed dataset.

[0029] The specific benchmark factor weighting system is a key framework constructed for comprehensive evaluation and decision-making. In rock mass-related studies, it assigns weights to multi-source factors such as joint characteristics and rock mass physical parameters. During construction, the influence of each factor on the target must be comprehensively considered; for example, joint density and orientation have a greater impact on rock mass stability, thus receiving higher weights, while less important factors have lower weights. Weight values ​​are determined through methods such as expert scoring and the analytic hierarchy process (AHP), ensuring that different factors are reasonably quantified according to their importance in the evaluation. This system ensures that the evaluation results are more scientific and objective, providing accurate and reliable basis for subsequent model construction and engineering decisions.

[0030] Specific data quality assessment: Based on the above-mentioned quantitative indicators of accuracy, completeness, and resolution, a comprehensive evaluation of the multi-source dataset is conducted, and a weight is assigned to each indicator. For example, the weight of accuracy is set to 0.5, the weight of completeness is set to 0.3, and the weight of resolution is set to 0.2.

[0031] like Figure 4 As shown, in this embodiment, based on the actual characteristics of the rock mass, numerical simulation software and calculation methods are selected to construct a rock mass geometric model, and the initial three-dimensional joint model is embedded into the rock mass geometric model; the density and elastic modulus of the rock mass, as well as the stiffness and friction coefficient of the joints are set; the excitation source information is input, the simulation program is run, and the propagation process of stress waves in the jointed rock mass is simulated; the simulation results are processed to obtain the theoretical wave velocity field distribution.

[0032] Numerical simulation models of stress wave propagation in jointed rock masses can accurately simulate the propagation process of stress waves under complex environments. By inputting an initial three-dimensional joint model and setting parameters, the influence of rock mass and joint characteristics on propagation can be comprehensively considered. Stress wave information from different locations and times is extracted to obtain key data, which, after processing, yields the theoretical wave velocity field distribution and stress wave attenuation characteristics. This contributes to a deeper understanding of the propagation laws of stress waves in jointed rock masses, providing theoretical support for rock engineering projects such as tunnel excavation and slope stability analysis, optimizing engineering design, and ensuring construction safety and quality.

[0033] like Figure 5As shown, in this embodiment, a convolutional neural network is used to extract deep features from the preprocessed multi-source data. Multi-layer convolution and pooling operations are used to automatically capture the geometric morphology and spatial distribution patterns of joints. The extracted joint features are fused with the geological background and mechanical properties of the rock mass to determine the constraints for model construction. A three-dimensional mathematical model of joints is constructed based on the spatial feature parameters, using multivariate surface equations to characterize the shape distribution and development characteristics of joints. The dynamic game mechanism between the generator and discriminator in a generative adversarial network is used to iteratively optimize the model structure, transforming the feature data into the joint network topology in three-dimensional space. Using the feature data of the three-dimensional mathematical model of joints, the wave velocity field distribution model generated by stress waves passing through the rock mass, and the terminal velocity of attenuation as the result data, and using the rock mass stress wave propagation theory formula as the loss function, a generative adversarial network model is constructed. By training the model, the feature parameters of the three-dimensional mathematical model of joints are continuously adjusted to control the error of the model loss function within an allowable range, thus obtaining the model parameters and the three-dimensional imaging results of the joints.

[0034] By utilizing convolutional neural networks to deeply extract features from multi-source data, the geometric morphology and spatial distribution patterns of joints can be accurately captured, laying the foundation for subsequent analysis. Integrating joint features with relevant rock mass parameters to determine constraints makes the model construction more rational. The constructed three-dimensional mathematical model of joints can intuitively represent the shape and development characteristics of joints. By optimizing the model structure with generative adversarial networks and combining it with stress wave propagation theory, and adjusting parameters through training, errors can be effectively controlled, obtaining accurate three-dimensional imaging results of joints, providing a reliable basis for rock mass engineering research, design, and construction.

[0035] In this embodiment, data fusion technology is employed. Wave velocity field data is introduced as a key constraint into the initial joint network model, and the model parameters are corrected using the anomalous variation characteristics of wave velocity at the joints. A conditional generative adversarial network is used, and the model parameters are fine-tuned under the wave velocity field constraint through an adversarial training mechanism between the generator and the discriminator, so that the spatial distribution of the joint network is highly consistent with the wave velocity propagation law. The objective function of the inversion problem is defined as the difference between the measured wave velocity data and the forward simulation results. The optimized model parameters are reconstructed in three-dimensional space to obtain a rock mass internal joint network model with high accuracy and physical consistency under the wave velocity field constraint.

[0036] When employing data fusion technology, wave velocity field data is introduced as a key constraint into the initial joint network model. Specifically, based on the close correlation between wave velocity and joints, the measured wave velocity field data is mapped to the model space. For correction of anomalous change characteristics, regions with abnormal wave velocity are first identified, and the degree of difference between their wave velocities and those of the surrounding normal regions is analyzed. Based on the magnitude of the difference, the geometric and mechanical parameters of the joints at the corresponding locations are adjusted according to preset rules, so that the wave velocity simulated by the model gradually converges with the measured wave velocity, improving the physical consistency of the model. Through the adversarial training mechanism of a conditional generative adversarial network, the model parameters are fine-tuned under the aforementioned wave velocity field constraints, ensuring that the spatial distribution of the joint network highly matches the wave velocity propagation law. After three-dimensional spatial reconstruction, a high-precision joint network model of the rock mass is obtained, providing a reliable basis for engineering.

[0037] In this embodiment, measured data collected from newly added boreholes are compared and analyzed with the joint distribution at corresponding locations in the 3D model to extract error features and quantify model accuracy. The verification results are input into an incremental learning framework to dynamically adjust neural network parameters to correct model deviations, ensuring that the imaging results continuously match actual geological conditions. The optimized 3D model is imported into a virtual reality engine, and immersive visualization is achieved through real-time rendering and interactive design, supporting multi-angle observation and geological analysis, resulting in dynamic imaging results of joints and fractures inside the rock mass that combine accuracy and intuitiveness.

[0038] The accuracy of the model is verified by comparing and analyzing newly added borehole measurement data with the 3D model. If the verification is inaccurate, the sources of error are analyzed in depth, such as data acquisition errors and model algorithm defects. Corresponding measures are taken for different causes: if the problem is with the data, the data is re-acquired or corrected; if the problem is with the algorithm, the model structure is optimized. The verification results are input into an incremental learning framework, and the neural network parameters are dynamically adjusted according to the error characteristics to continuously correct model deviations. By continuously incorporating new data, the model is dynamically updated, making the imaging results more consistent with actual geological conditions. The optimized 3D model is imported into a virtual reality engine to achieve immersive visualization and provide a scientific basis for rock engineering decisions.

[0039] This invention also provides an intelligent modeling and dynamic updating system for three-dimensional joint networks in rock masses, comprising the following modules: a multi-source dataset module, used to simultaneously collect surface joint, borehole internal joint, and wave velocity field data through an integrated multi-physics detection module, process borehole images to extract joint features, and obtain a multi-source dataset and a joint database; a preprocessing module, used to evaluate data quality by analyzing the accuracy, completeness, and resolution of the multi-source dataset, optimize low-resolution data, and obtain baseline factor weights and a preprocessed dataset; and a joint network module, used to extract joint features through a convolutional neural network, combine geological background and mechanical properties, and utilize generated pairs... An adversarial network is used to construct an initial joint network model and obtain 3D imaging results of the joints. The simulation propagation module is used to establish a propagation model of stress waves in jointed rock masses. The initial 3D joint model is used as input to simulate the stress wave propagation process and output the theoretical wave velocity field distribution and attenuation characteristics. The model optimization module is used to introduce wave velocity field data as constraints into the model and use a conditional generative adversarial network to fine-tune the parameters to obtain a high-precision joint network model under wave velocity field constraints. The model verification module is used to verify the accuracy of the model by adding new borehole data, use incremental learning to dynamically correct the parameters, obtain 3D imaging results, and realize immersive visualization.

[0040] This invention also provides an intelligent modeling and dynamic updating device for three-dimensional rock joint networks. This device may further include one or more power supplies, one or more wired or wireless network interfaces, one or more input / output interfaces, and / or one or more operating systems, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that the structure of this intelligent modeling and dynamic updating device for three-dimensional rock joint networks does not constitute a limitation on the computer equipment provided by this invention, and may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements.

[0041] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock masses, characterized in that, The intelligent modeling and dynamic updating method for three-dimensional joint networks of rock masses includes the following steps: By integrating a multi-physics field detection module, rock surface joints, borehole internal joints and wave velocity field imaging data are collected simultaneously. Borehole images are processed to extract joint features, resulting in a multi-source dataset and joint database. By analyzing the accuracy, completeness, and resolution of multi-source datasets, data quality is evaluated, low-resolution data is optimized, and baseline factor weights and preprocessed datasets are obtained. Joint features are extracted using a convolutional neural network. Combined with geological background and mechanical properties, an initial joint network model is constructed using a generative adversarial network to obtain the three-dimensional imaging results of the initial joint network. A stress wave propagation model in jointed rock mass is established. The initial three-dimensional model of the joint network is used as input to simulate the stress wave propagation process and output the theoretical wave velocity field distribution characteristics. By introducing wave velocity field data as a constraint into the model and using a conditional generative adversarial network to fine-tune the parameters, a high-precision joint network model under wave velocity field constraints is obtained. The accuracy of the model was verified by adding new borehole data, and the parameters were dynamically corrected by incremental learning to obtain three-dimensional imaging results, thereby realizing the visualization of the joint and fracture network.

2. The method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to claim 1, characterized in that, The process of simultaneously acquiring surface joint, borehole internal joint, and wave velocity field data through an integrated multi-physics detection module, processing borehole images to extract joint features, and obtaining a multi-source dataset and joint database includes the following steps: An integrated multiphysics detection module is used to simultaneously collect surface joint data, borehole internal joint data, and wave velocity field data. The multiphysics detection module includes a 3D laser scanner, a borehole television, a standard drill pipe, an ultrasonic transmitter, and a receiver; The acquired borehole images are preprocessed, including grayscale conversion to reduce data complexity, Gaussian filtering to remove noise, histogram equalization to enhance contrast, and edge detection algorithm to extract geometric and spatial feature parameters of joints. The spatial characteristic parameters include the number of joints, joint length, joint width, joint development length, joint depth, and joint location; The processed data is integrated with the extracted joint features to obtain a dataset containing multi-dimensional information, and a detailed joint database is constructed to obtain a multi-source dataset and a joint database.

3. The method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to claim 1, characterized in that, The process of analyzing the accuracy, completeness, and resolution of multi-source datasets to evaluate data quality, optimize low-resolution data, and obtain baseline factor weights and preprocessed datasets includes the following steps: Using data analysis methods, we comprehensively evaluate data quality by deeply analyzing the accuracy, completeness, and resolution of surface joint, borehole joint, and wave velocity field data from multi-source datasets. For low-resolution data areas, super-resolution reconstruction technology is used for optimization to improve the overall accuracy of the data. The evaluation results are integrated with the optimized data, key influencing factors are extracted and assigned weights, resulting in a baseline factor weighting system and a rigorously preprocessed dataset.

4. The method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to claim 1, characterized in that, The establishment of a stress wave propagation model in jointed rock mass, using an initial three-dimensional joint model as input, simulates the stress wave propagation process and outputs the theoretical wave velocity field distribution, includes the following steps: Based on the actual characteristics of the rock mass, numerical simulation software and calculation methods were selected to construct a rock mass geometric model, and the initial three-dimensional joint model was embedded into the rock mass geometric model. Set the density and elastic modulus of the rock mass, as well as the stiffness and friction coefficient of the joints; input the excitation source information, run the simulation program, and simulate the propagation process of stress waves in jointed rock mass; The simulation results are processed to obtain the theoretical wave velocity field distribution.

5. The method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to claim 1, characterized in that, The process of extracting joint features through a convolutional neural network, combining geological background and mechanical properties, constructing an initial joint network model using a generative adversarial network, and obtaining three-dimensional imaging results of joints includes the following steps: Convolutional neural networks are used to extract deep features from preprocessed multi-source data, and multi-layer convolution and pooling operations are used to automatically capture the geometric shape and spatial distribution pattern of joints. The extracted joint features are integrated with the geological background and mechanical property parameters of the rock mass to determine the constraints for model construction. A three-dimensional mathematical model of joints is constructed based on the spatial characteristic parameters of joints, and the shape distribution and development characteristics of joints are characterized by multivariate surface equations. By utilizing the dynamic game mechanism between the generator and discriminator in generative adversarial networks, the model structure is iteratively optimized step by step, transforming feature data into joint network topology in three-dimensional space; Using the characteristic data of the three-dimensional mathematical model of joints, the wave velocity field distribution model and the terminal velocity of the stress wave passing through the rock mass as the result data, and the theoretical formula of rock mass stress wave propagation as the loss function, a generative adversarial network model is constructed. By training the model and continuously adjusting the characteristic parameters of the three-dimensional mathematical model of joints, the error of the model loss function is controlled within the allowable range, the model parameters are obtained, and the three-dimensional imaging results of joints are obtained.

6. The method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to claim 1, characterized in that, The process of introducing wave velocity field data as a constraint into the model and using a conditional generative adversarial network to fine-tune the parameters to obtain a high-precision joint network model under wave velocity field constraints includes the following steps: By employing data fusion technology, wave velocity field data is introduced as a key constraint into the initial joint network model, and the model parameters are corrected by utilizing the anomalous variation characteristics of wave velocity at the joints. A conditional generative adversarial network is adopted, and the parameters of the model are fine-tuned under the constraint of wave velocity field through the adversarial training mechanism between the generator and the discriminator, so that the spatial distribution of the joint network is highly consistent with the wave velocity propagation law. The objective function of the inversion problem is defined as the difference between the measured wave velocity data and the forward simulation results. The optimized model parameters are reconstructed in three-dimensional space to obtain a rock mass internal joint network model with high accuracy and physical consistency under wave velocity field constraints.

7. The method for intelligent modeling and dynamic updating of three-dimensional joint networks in rock mass according to claim 1, characterized in that, The process of verifying the model's accuracy by adding new borehole data, using incremental learning to dynamically correct parameters, obtaining 3D imaging results, and achieving immersive visualization includes the following steps: By using the measured data collected from the newly added boreholes and comparing it with the joint distribution at the corresponding locations in the 3D model, error features were extracted and the accuracy of the model was quantified. The verification results are input into the incremental learning framework, and the neural network parameters are dynamically adjusted to correct model biases, so that the imaging results continuously match the actual geological conditions. The optimized 3D model is imported into the virtual reality engine, and immersive visualization is achieved through real-time rendering and interactive design. It supports multi-angle observation and geological analysis, and obtains dynamic imaging results of joints and fractures inside the rock mass that are both accurate and intuitive.

8. A three-dimensional joint network intelligent modeling and dynamic updating system for rock masses, characterized in that, The intelligent modeling and dynamic updating system for three-dimensional joint networks of rock masses includes the following modules: The multi-source dataset module is used to synchronously collect surface joint, borehole internal joint and wave velocity field data by integrating a multi-physics field detection module, process borehole images to extract joint features, and obtain multi-source datasets and joint databases. The preprocessing module is used to evaluate data quality by analyzing the accuracy, completeness, and resolution of multi-source datasets, optimize low-resolution data, and obtain the baseline factor weights and preprocessed datasets. The joint network module is used to extract joint features through a convolutional neural network, combine geological background and mechanical properties, and use a generative adversarial network to construct an initial joint network model to obtain three-dimensional imaging results of joints. The simulation propagation module is used to establish a propagation model of stress waves in jointed rock masses. It takes the initial three-dimensional joint model as input, simulates the propagation process of stress waves, and outputs the theoretical wave velocity field distribution and attenuation characteristics. The model optimization module is used to introduce wave velocity field data as a constraint into the model and fine-tune the parameters using a conditional generative adversarial network to obtain a high-precision joint network model under wave velocity field constraints. The model verification module is used to verify the accuracy of the model by adding new borehole data, dynamically correct parameters using incremental learning, obtain 3D imaging results, and achieve immersive visualization.

9. A device for intelligent modeling and dynamic updating of three-dimensional joint networks in rock masses, characterized in that, The intelligent modeling and dynamic updating device for three-dimensional joint networks of rock mass includes a memory and at least one processor. The memory stores instructions, and the at least one processor calls the instructions in the memory to cause the intelligent modeling and dynamic updating device for three-dimensional joint networks of rock mass to perform the various steps of the intelligent modeling and dynamic updating method for three-dimensional joint networks of rock mass as described in any one of claims 1-7.

Images