A method and system for identifying and predicting rock fractures.

Abstract

This invention provides a method and system for rock fracture identification and propagation prediction. The method includes: obtaining original slice images of siltstone samples from a triaxial cyclic loading and unloading test using CT scanning, and preprocessing them; inputting the preprocessed slice images into a SAM segmentation model for rock fracture identification and extracting geometric spatial feature parameters; constructing a database for training and testing a graph convolutional neural network (GCN) based on the geometric spatial feature parameters; establishing a GCN in the PyTorch framework and training the GCN to obtain a trained GCN model; and analyzing the connectivity between fractures based on the trained GCN model to predict fracture propagation paths. This invention, based on a hybrid strategy of a large SAM model and a graph convolutional neural network, improves the efficiency and accuracy of neural networks in rock fracture identification and provides guidance for subsequent prediction of rock fracture propagation paths.

Description

Technical Field

[0001] This invention relates to the field of image recognition and prediction technology for fracture propagation paths during rock fracturing, and in particular to a method and system for rock fracture recognition and propagation prediction. Background Technology

[0002] Due to the complex geological processes that rock masses undergo over long periods, they possess numerous structural surfaces such as joints, fissures, and faults. The strength and failure characteristics of rock masses under external forces depend on these natural defects. Therefore, in various rock mass engineering projects, obtaining the geometric spatial characteristic parameters of fissures and predicting their subsequent development and propagation paths can provide a deeper understanding of the mechanical properties of rock masses, guide the design and construction of rock mass engineering projects, and prevent and manage rock mass geological hazards.

[0003] Traditionally, the acquisition of rock mass fractures relies on manual on-site measurements using tools such as measuring tapes, a method that is not only inefficient but also costly. With the development of computer vision technology, deep learning-based intelligent fracture recognition methods have been widely applied. Research on fracture propagation and evolution mainly employs methods such as theoretical analysis, laboratory experiments, and numerical simulations. Many scholars have used fracture mechanics principles and the generalized Griffith instability criterion to explain fracture propagation laws; conducted laboratory experiments using rock-like materials or real rocks with pre-existing fractures; and further investigated the mechanisms of multi-fracture problems and the macroscopic fracture process of rock masses using numerical methods such as PFC. However, these existing methods have limitations and poor accuracy, and cannot further predict the path of fracture propagation.

[0004] Graph Convolutional Networks (GCNs) are deep learning models that combine graph theory and convolutional neural networks to capture the dependencies in a graph by observing the information exchange between nodes. In recent years, they have been applied in various industries, such as social network analysis and transportation networks, and have played an important role in solving practical problems.

[0005] Rock fracture systems are complex systems that exist under long-term geostress, possessing unique and complex properties such as random size, varied shapes, and unpredictable changes. Furthermore, the interactions between fractures are complex, making them difficult to understand using conventional research methods. Therefore, applying the Segment Anything Model (SAM) combined with graph convolutional neural networks to the intelligent identification and propagation prediction of fractures in rocks during triaxial cyclic loading and unloading can compensate for the shortcomings of existing methods in rock fracture propagation research and provide a theoretical foundation and technical support for subsequent prediction of rock fracture propagation paths. Summary of the Invention

[0006] To address the aforementioned technical problems, and to further explore the connectivity between fracture networks within rocks, as well as to reveal the diverse mechanisms underlying fracture development under pressure, this invention provides a method and system for identifying and predicting rock fracture propagation. The technical solution is as follows:

[0007] On the one hand, a method for identifying and predicting rock fractures is provided, the method comprising the following steps:

[0008] S1. Use CT scanning to obtain original slice images of siltstone samples in the triaxial cyclic loading and unloading test, and perform preprocessing.

[0009] S2. Input the preprocessed slice image into the SAM segmentation model to identify rock fractures and extract geometric spatial feature parameters;

[0010] S3. Construct a database for training and testing graph convolutional neural networks based on the geometric spatial feature parameters;

[0011] S4. Build a graph convolutional neural network in the PyTorch framework, train the graph convolutional neural network to obtain the trained GCN model, and verify it.

[0012] S5. Analyze the connection relationship between fractures based on the trained GCN model and predict the fracture propagation path.

[0013] Optionally, in step S1, a triaxial cyclic loading and unloading test is performed on the siltstone sample, and three-dimensional information of the internal cracks of the sample is obtained by CT scanning after each loading and unloading. The CT scanning points are set to 30%, 60%, 90% of the peak strength of the siltstone and after the sample breaks, so as to establish an original slice image library of cracks and perform unified preprocessing.

[0014] Optionally, in step S2, the preprocessed slice image is input into the fine-tuned SAM segmentation model to achieve intelligent identification of rock fissures; the outline of the binary image of the fissure is obtained using the OpenCV library, and the length, width, and dip angle parameters of the fissure are extracted based on the outline using library functions; then, according to the slice image information, the two-dimensional geometric spatial feature parameters of the fissure are converted into three-dimensional geometric spatial feature parameters.

[0015] Optionally, in step S3, each crack is defined as a node, the geometric spatial feature parameters of the crack are used as the relevant features of the node, the cracks are connected by edges and an adjacency matrix is ​​formed by the distance, and a database for training and testing graph convolutional neural networks is obtained.

[0016] Optionally, in step S3, during the triaxial cyclic loading and unloading test of siltstone, the data is divided into multiple stages according to different stress loading and unloading points. When training the graph convolutional neural network, the data of the previous stage is used as the input of the model, and the data of the next stage is used as the output of the model. Therefore, the data in the database is divided into multiple categories, and different graph convolutional neural networks are trained for each category to obtain a model for predicting the rock fracture propagation path at different stages.

[0017] Optionally, in step S4, the operation of the graph convolutional neural network is represented as follows:

[0018] (1)

[0019] Among them, H (l) W is the feature matrix of the nodes in the l-th layer. (l) It is the weight matrix of the l-th layer. It is the normalized adjacency matrix. yes The degree matrix, where σ is the activation function;

[0020] Each graph convolutional layer performs a linear transformation on the features of the nodes, performs a weighted summation through the adjacency matrix, and then applies a non-linear activation function to extract the higher-order features of the nodes layer by layer.

[0021] To predict the next propagation path of rock fractures at each breakpoint in a triaxial cyclic loading and unloading test, a nodal regression task of the GCN model was established, and stochastic gradient descent (SGD) was selected as the optimizer. The loss function, Loss, was calculated using the following formula:

[0022] (2)

[0023] Where N is the number of nodes, y i It is the actual value of the i-th node. It is the predicted value of the i-th node.

[0024] Optionally, in step S4, the training set in the database is input into the GCN model for training to obtain the trained GCN model. The convergence of the model is checked by the loss value, and the prediction accuracy of the model is evaluated by the accuracy, precision, recall and Dice coefficient.

[0025] Optionally, during the training of the GCN model, nodes transmit and aggregate information, so that each node contains not only its own feature information but also the feature information of other nodes. The above mechanism is used to analyze the connection relationship between the fractures, thereby revealing the mutual influence relationship and potential structural patterns between the fractures.

[0026] Optionally, in step S5, the trained GCN model is analyzed to extract and visualize node features and edge weights, quantify the connection relationships between fractures, and analyze the influence of node features on fracture propagation paths, thereby revealing the internal mechanisms of different fracture development and the interactions between fractures.

[0027] On the other hand, a rock fracture identification and propagation prediction system is provided for implementing the method described in any of the above embodiments, the system comprising:

[0028] The data acquisition module is used to obtain raw slice images of siltstone samples in triaxial cyclic loading and unloading tests using CT scans, and to perform preprocessing.

[0029] The fracture identification module is used to input the preprocessed slice image into the SAM segmentation model to identify rock fractures and extract geometric spatial feature parameters;

[0030] A database construction module is used to construct a database for training and testing graph convolutional neural networks based on the geometric spatial feature parameters.

[0031] The model building module is used to build a graph convolutional neural network in the PyTorch framework, train the graph convolutional neural network to obtain the trained GCN model, and then verify it.

[0032] The analysis and prediction module is used to analyze the connectivity between fractures based on the trained GCN model and predict the fracture propagation path.

[0033] On the other hand, an electronic device is provided, the electronic device comprising:

[0034] processor;

[0035] The memory stores computer-readable instructions, which, when loaded and executed by the processor, implement the steps of the rock fracture identification and propagation prediction method described above.

[0036] On the other hand, a computer-readable storage medium is provided, wherein program code is stored in the computer-readable storage medium, and the program code can be called by a processor to execute the steps of the rock fracture identification and propagation prediction method described above.

[0037] The beneficial effects of the technical solution provided by this invention include at least the following:

[0038] This invention combines CT scanning technology, the SAM large-scale model, and the GCN deep learning model to acquire three-dimensional information of complex internal fractures in siltstone during triaxial cyclic loading and unloading, and characterizes them in slice form. Based on this, a CT slice image database is established. The SAM large-scale model is used to achieve intelligent fracture identification and extraction of geometric spatial feature parameters. The fracture features are then used to train a graph convolutional neural network to obtain an optimized GCN model, ultimately achieving efficient, fast, and accurate identification and prediction of rock fracture propagation paths. This invention significantly improves the efficiency and accuracy of deep learning methods in the field of rock fracture identification and prediction, providing a new perspective for subsequent analysis of the connectivity and development mechanisms between fractures. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of a rock fracture identification and propagation prediction method provided in an embodiment of the present invention;

[0041] Figure 2 This is a schematic diagram of the structure of a rock fracture identification and propagation prediction system provided in an embodiment of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0043] In embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either one or the other. In embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably; it should be noted that their intended meanings are consistent unless the distinction is emphasized.

[0044] This invention addresses the limitations of existing deep learning methods in identifying rock fractures, including low efficiency and accuracy, poor accuracy in studying fracture propagation and evolution, and the inability to accurately predict fracture propagation paths. Consequently, it fails to provide accurate guidance for the design and construction of rock engineering projects and the prevention and control of rock geological disasters. This invention delves into the connectivity between fractures and, based on a hybrid strategy of SAM and GCN, achieves efficient, rapid, and accurate identification and prediction of rock fracture propagation paths.

[0045] like Figure 1 As shown, this embodiment of the invention provides a method for identifying and predicting rock fractures, including the following steps:

[0046] S1. Use CT scans to obtain original slice images of siltstone samples from the triaxial cyclic loading and unloading test, and perform preprocessing.

[0047] Specifically, since a dataset of three-dimensional information on the complex internal fractures of siltstone during triaxial cyclic loading and unloading tests is not currently publicly available, this embodiment of the invention uses a siltstone sample with dimensions of Φ50mm×100mm and an end-face non-parallelism controlled below 0.02mm to undergo a triaxial cyclic loading and unloading test using a servo-controlled rock mechanics testing system (MTS 815). After each loading and unloading cycle, the sample is scanned in a 360-degree, spiral pattern using CT technology to obtain three-dimensional information on the complex internal fractures. The CT scan points are set at 30%, 60%, and 90% of the peak strength of the siltstone, as well as after the sample fractures, to establish a raw image library of fracture slices, which are then uniformly preprocessed.

[0048] S2. Input the preprocessed slice image into the SAM segmentation model to identify rock fractures and extract geometric spatial feature parameters.

[0049] In this embodiment of the invention, the preprocessed slice image is input into the fine-tuned SAM segmentation model to achieve intelligent identification of rock fissures; the outline of the binary image of the fissure is obtained using the OpenCV library, and the length, width, dip angle and other parameters of the fissure are extracted based on the outline using library functions; then, according to the slice image information, the two-dimensional geometric spatial feature parameters of the fissure are converted into three-dimensional geometric spatial feature parameters.

[0050] Specifically, the original slice images obtained from CT scans of siltstone undergo uniform preprocessing to align the rock samples at each stage of the experiment in terms of resolution, retaining only the slice images containing the rock samples. The preprocessed slice images are then input into the Segment Anything Model (SAM), a large-scale segmentation model trained on millions of images and over a billion masks. This model can perform pixel-level image segmentation tasks by clicking, selecting boxes, and automatically recognizing target objects in the image. Therefore, this invention uses a fine-tuned SAM model to achieve intelligent identification of rock fractures.

[0051] The preprocessed slice image is input into the fine-tuned SAM segmentation model. After recognizing the two-dimensional slice image, the model outputs a binary image of the crack. In order to extract its geometric spatial feature parameters in the two-dimensional slice image, the outline of the crack binary image is first obtained through the OpenCV library. Based on the outline, the length, width, inclination angle and other parameters of the crack are extracted through library functions. Then, according to the information of the slice image, the two-dimensional geometric spatial feature parameters of the crack are converted into three-dimensional geometric spatial feature parameters.

[0052] S3. Construct a database for training and testing graph convolutional neural networks based on the geometric spatial feature parameters.

[0053] In this step, the data used for training and testing the graph convolutional neural network consists of the geometric spatial feature parameters of the cracks. Each crack is defined as a node, and the geometric spatial feature parameters of the cracks are used as the relevant features of the nodes. The cracks are connected by edges and form an adjacency matrix based on their distance, thus obtaining a database for training and testing the graph convolutional neural network.

[0054] After the above data processing, the data from the triaxial cyclic loading and unloading test of siltstone can be divided into multiple stages based on different stress loading and unloading points. When training the graph neural network, the data from the previous stage is used as the model input, and the data from the next stage is used as the model output. Therefore, the data can be divided into multiple categories, and each category is randomly divided into training and test sets in a 7:3 ratio. Different graph convolutional neural networks are trained on the training sets of each category, thereby obtaining a model for predicting the rock fracture propagation path at different stages.

[0055] S4. Build a graph convolutional neural network in the PyTorch framework, train the graph convolutional neural network to obtain the trained GCN model, and verify it.

[0056] Graph Convolutional Networks (GCNs) are an extension of Convolutional Neural Networks (CNNs) to graph-structured data. They can process non-Euclidean data and efficiently handle tasks such as node classification, graph classification, and edge prediction by leveraging the local structure and node features of the graph.

[0057] In this step, GCN captures the complex relationships between nodes by performing convolution operations on the adjacency matrix (N*D) and node feature matrix (N*N) of the graph. Its core idea is to gradually aggregate the local information of nodes through graph convolutional layers, thereby obtaining a global representation of each node.

[0058] Specifically, the operations of a graph convolutional neural network are represented as follows:

[0059] (1)

[0060] Among them, H (l) W is the feature matrix of the nodes in the l-th layer. (l) It is the weight matrix of the l-th layer. It is the normalized adjacency matrix. yes The degree matrix is ​​σ, where σ is the activation function.

[0061] Each graph convolutional layer performs a linear transformation on the features of the nodes, performs a weighted summation through the adjacency matrix, and then applies a nonlinear activation function to extract the higher-order features of the nodes layer by layer.

[0062] Since the ultimate goal of this invention is to predict the next propagation path of rock fractures at each breakpoint during triaxial cyclic loading and unloading, this falls under the node regression task in GCN networks. Therefore, in this embodiment of the invention, stochastic gradient descent (SGD) is selected as the optimizer, and the loss function Loss is calculated as follows:

[0063] (2)

[0064] Where N is the number of nodes, y i It is the actual value of the i-th node. It is the predicted value of the i-th node.

[0065] The GCN network was repeatedly trained to obtain the optimal GCN model after training, and the convergence and accuracy of the model were verified by qualitative and quantitative methods.

[0066] Specifically, the training set from the database is input into the GCN network for training. During training, the network's hyperparameters are continuously adjusted until the loss no longer decreases, indicating model convergence and the optimal trained GCN model is obtained. Furthermore, during model training, changes in the loss value are recorded to monitor the training progress and address any anomalies that occur, thus maintaining efficient model training. This not only helps reflect the model's convergence but also allows for performance comparisons between different models, determining the convergence endpoint of the best model.

[0067] To evaluate the model's prediction accuracy, this invention comprehensively tests the model's performance using multiple metrics to ensure accurate predictions under various conditions. These metrics include accuracy, precision, recall, and the Dice coefficient. The calculation formulas are as follows:

[0068] (3)

[0069] (4)

[0070] (5)

[0071] (6)

[0072] In the formula: TP is the total number of the same crack in both the real and predicted coordinates in three-dimensional space; FP is the total number of the same crack predicted as the background in the real; TN is the total number of the same crack predicted as the background in both the real and predicted coordinates; and FN is the total number of the same crack predicted as the background in the real.

[0073] S5. Analyze the connection relationship between fractures based on the trained GCN model and predict the fracture propagation path.

[0074] During the training of the GCN model, nodes transmit and aggregate information, so that each node contains not only its own feature information, but also the feature information of other nodes. The above mechanism is used to analyze the connection relationship between the fractures, thereby revealing the mutual influence relationship and potential structural patterns between the fractures.

[0075] In this step, the trained GCN model is analyzed to extract and visualize node features and edge weights, quantify the connectivity between fractures, and analyze the influence of node features (such as location and shape) on fracture propagation paths, thereby revealing the underlying mechanisms of diverse fracture development. Furthermore, the GCN model can be used to identify fractures that significantly impact the overall structure of the rock sample, as well as the significant interactions between fractures.

[0076] Finally, through parameter analysis of the GCN model, not only can the connectivity between fractures be quantified, but the complex mechanisms of fracture development can also be understood from the perspective of graph convolutional neural networks. This method provides a powerful tool for studying fracture networks, which can help to better predict and control fracture behavior in rock engineering.

[0077] Accordingly, embodiments of the present invention also provide a rock fracture identification and propagation prediction system. Figure 2 This is a structural block diagram illustrating a rock fracture identification and propagation prediction system according to an exemplary embodiment. Figure 2 As shown, the system includes:

[0078] The data acquisition module 201 is used to obtain raw slice images of siltstone samples in the triaxial cyclic loading and unloading test by using CT scanning, and to perform preprocessing.

[0079] The fracture identification module 202 is used to input the preprocessed slice image into the SAM segmentation model to identify rock fractures and extract geometric spatial feature parameters;

[0080] Database construction module 203 is used to construct a database for training and testing graph convolutional neural networks based on the geometric space feature parameters;

[0081] The model building module 204 is used to build a graph convolutional neural network in the PyTorch framework, train the graph convolutional neural network to obtain the trained GCN model, and verify it.

[0082] The analysis and prediction module 205 is used to analyze the connection relationship between fractures based on the trained GCN model and predict the fracture propagation path.

[0083] For ease of explanation, Figure 2 Only the main components of the system are shown. The system of this embodiment can be used to perform... Figure 1 The technical solutions of the method embodiments shown are similar in principle and in effect, and will not be described again here.

[0084] This invention provides a method and system for rock fracture identification and propagation prediction. It utilizes CT scanning technology to accurately and rapidly acquire three-dimensional information on complex internal fractures in siltstone during triaxial cyclic loading and unloading. Based on this slice-based characterization, a large-scale SAM model is employed to intelligently identify rock fractures and extract their geometric spatial feature parameters. The fracture geometric spatial feature parameters and other data are then used to train a GCN network to obtain the optimal model, enabling the prediction of fracture propagation during triaxial compression. Notably, this invention starts from the nodes and edges of a graph convolutional neural network, focusing not only on the propagation path prediction of a single fracture but also revealing the diverse mechanisms of fracture development through the connections between fractures. The research approach and results of this invention can provide effective technical means for predicting fracture propagation paths in other rocks and even rock masses under different compression conditions, and have certain guiding significance for the design and construction of subsequent rock engineering projects, as well as the prevention and control of rock geological disasters.

[0085] In an exemplary embodiment, the present invention also provides an electronic device, the electronic device comprising:

[0086] processor;

[0087] The memory stores computer-readable instructions, which, when loaded and executed by the processor, implement the steps of the rock fracture identification and propagation prediction method described above.

[0088] In an exemplary embodiment, the present invention also provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the steps of the rock fracture identification and propagation prediction method described above. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

[0089] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0090] The use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0091] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0092] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0093] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0094] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0095] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0096] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0097] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0098] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying and predicting rock fractures, characterized in that, Includes the following steps: S1. Use CT scanning to obtain original slice images of siltstone samples in the triaxial cyclic loading and unloading test, and perform preprocessing. In step S1, a siltstone sample with dimensions of Φ50mm×100mm and end face non-parallelism controlled below 0.02mm is subjected to a triaxial cyclic loading and unloading test using a servo-controlled rock mechanics testing system MTS 815. After each loading and unloading, the sample is scanned in all directions and spirally using CT to obtain three-dimensional information of the internal fractures. The CT scanning points are set to 30%, 60%, 90% of the peak strength of the siltstone and after the sample fractures, thereby establishing an original slice image library of fractures. The original slice images obtained from CT scans of siltstone samples were uniformly preprocessed to align the rock samples at different stages of the experiment in terms of resolution, and only slice images containing rock samples were retained. S2. Input the preprocessed slice image into the SAM segmentation model to identify rock fractures and extract geometric spatial feature parameters; In step S2, the preprocessed slice image is input into the fine-tuned SAM segmentation model to achieve intelligent identification of rock fissures; the outline of the binary image of the fissure is obtained using the OpenCV library, and the length, width, and dip angle parameters of the fissure are extracted based on the outline using library functions; then, according to the slice image information, the two-dimensional geometric spatial feature parameters of the fissure are converted into three-dimensional geometric spatial feature parameters. S3. Construct a database for training and testing graph convolutional neural networks based on the geometric spatial feature parameters; In step S3, each crack is defined as a node, and the geometric spatial feature parameters of the crack are used as the relevant features of the node. The cracks are connected by edges and form an adjacency matrix based on their distance, thus obtaining a database for training and testing graph convolutional neural networks. During the triaxial cyclic loading and unloading test of siltstone, the data was divided into multiple stages according to different stress loading and unloading points. When training the graph convolutional neural network, the data of the previous stage was used as the input of the model, and the data of the next stage was used as the output of the model. The data in the database was divided into multiple categories, and different graph convolutional neural networks were trained for each category to obtain a model for predicting the rock fracture propagation path at different stages. S4. Build a graph convolutional neural network in the PyTorch framework, train the graph convolutional neural network to obtain the trained GCN model, and verify it. In step S4, the operation of the graph convolutional neural network is represented as follows: Among them, H (l) W is the feature matrix of the nodes in the l-th layer. (l) It is the weight matrix of the l-th layer. It is the normalized adjacency matrix. yes The degree matrix, where σ is the activation function; Each graph convolutional layer performs a linear transformation on the features of the nodes, performs a weighted summation through the adjacency matrix, and then applies a non-linear activation function to extract the higher-order features of the nodes layer by layer. To predict the next propagation path of rock fractures at each breakpoint in a triaxial cyclic loading and unloading test, a nodal regression task of the GCN model was established, and stochastic gradient descent (SGD) was selected as the optimizer. The loss function, Loss, was calculated using the following formula: Where N is the number of nodes, y i It is the actual value of the i-th node. It is the predicted value of the i-th node; The training set in the database is input into the GCN model for training to obtain the trained GCN model. During the training process of the GCN model, nodes transmit and aggregate information, so that nodes not only contain their own feature information, but also the feature information of other nodes. The above mechanism is used to analyze the connection relationship between fractures, thereby revealing the mutual influence relationship and potential structural patterns between fractures. The convergence of the model is tested by the loss value, and the prediction accuracy of the model is evaluated by the accuracy, precision, recall and Dice coefficient. The calculation formula is as follows: In the formula: TP is the total number of the same crack in both the real and predicted coordinates in three-dimensional space; FP is the total number of the same crack in the real and predicted coordinates; TN is the total number of the same crack in both the real and predicted coordinates; and FN is the total number of the same crack in the real and predicted coordinates. S5. Analyze the connection relationship between fractures based on the trained GCN model and predict the fracture propagation path; In step S5, the trained GCN model is analyzed to extract and visualize node features and edge weights, quantify the connection relationships between fractures, and analyze the influence of node features on fracture propagation paths, thereby revealing the internal mechanisms of different fracture development and the interactions between fractures.

2. A rock fracture identification and propagation prediction system, said system being used to implement the method as described in claim 1, characterized in that, The system includes: The data acquisition module is used to obtain raw slice images of siltstone samples in triaxial cyclic loading and unloading tests using CT scans, and to perform preprocessing. The fracture identification module is used to input the preprocessed slice image into the SAM segmentation model to identify rock fractures and extract geometric spatial feature parameters; A database construction module is used to construct a database for training and testing graph convolutional neural networks based on the geometric spatial feature parameters. The model building module is used to build a graph convolutional neural network in the PyTorch framework, train the graph convolutional neural network to obtain the trained GCN model, and then verify it. The analysis and prediction module is used to analyze the connectivity between fractures based on the trained GCN model and predict the fracture propagation path.

Images