A numerical crack prediction method, system and terminal based on graphical model

Abstract

This invention discloses a method, system, and terminal for numerical crack prediction based on a graph model. The method includes: constructing a graph model for each target region based on core scan CT data and acoustic emission technology experimental data of multiple target regions; constructing a crack model for the corresponding target region by combining a mechanical model; obtaining crack values ​​for each target region based on the crack model and user requirements; constructing a target training set based on the core scan CT data, acoustic emission technology experimental data, and crack values ​​of all target regions; and training a target convolutional neural network based on the target training set to obtain a crack prediction model; whenever core scan CT data and acoustic emission technology experimental data of a region to be predicted are obtained, processing the region to be predicted based on the crack prediction model to obtain the crack value of the region to be predicted. This invention can achieve fast and accurate crack value estimation.

Description

Technical Field

[0001] This invention relates to the field of crack prediction technology, and in particular to a crack numerical prediction method, system and terminal based on a graph model. Background Technology

[0002] During oil and gas extraction, water injection and pressurization are often used to propagate fractures and increase flow channels. However, the extension of fractures can inadvertently activate natural underground faults, damaging existing geological structures and potentially causing rock fracturing and displacement, leading to other problems. Therefore, obtaining accurate fracture data helps optimize extraction processes, improve production efficiency, and reduce environmental risks, thus aiding oil and gas exploration and development to achieve efficient resource extraction and rational utilization.

[0003] However, currently, data acquisition for oil and gas well fracture modeling often requires huge costs and the actual samples obtained are relatively scarce. Furthermore, mechanical simulation modeling methods often still have some difficulties in simulating the interrelationships such as the intersection and bifurcation behavior of a large number of fractures. At the same time, they are more suitable for simulating large-scale fractures at the scale of hundreds of kilometers and kilometers, and it is often difficult to refine them at the scale of decimeters, centimeters, or even millimeters, which makes it impossible to accurately and conveniently obtain accurate fracture values.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a numerical prediction method, system, terminal, and computer-readable storage medium for fractures based on a graphical model. This aims to address the problems in existing technologies where data acquisition before modeling fractures in oil and gas wells often incurs significant costs and actual sample availability is scarce. Furthermore, mechanical simulation modeling methods often struggle to simulate the interrelationships of numerous fractures, such as their intersections and bifurcations. These methods are more suitable for simulating large-scale fractures at the hundred-kilometer and kilometer levels, but are often difficult to refine at smaller scales such as decimeters, centimeters, or even millimeters, resulting in the inability to accurately and conveniently obtain precise fracture numerical values.

[0006] To achieve the above objectives, the present invention provides a graph-based numerical prediction method for cracks, which includes the following steps:

[0007] Based on the core scan CT data and acoustic emission technology experimental data of multiple target areas, a graphical model of each target area was constructed, and based on the graphical model of each target area, a crack model of the corresponding target area was constructed in combination with the mechanical model.

[0008] Based on the crack model of each target area, the crack values ​​of each target area are obtained in combination with user needs. According to the core scan CT data, acoustic emission technology experimental data and crack values ​​of all target areas, a target training set is constructed, and a target convolutional neural network is trained according to the target training set to obtain a crack prediction model.

[0009] Whenever core scan CT data and acoustic emission technology experimental data of the area to be predicted are obtained, the area to be predicted is processed based on the crack prediction model to obtain the crack value of the area to be predicted.

[0010] Optionally, the step of constructing a graphical model for each target region based on the acquired core scan CT data and acoustic emission technology experimental data of multiple target regions specifically includes:

[0011] Based on the core scan CT data of each target area, the fracture endpoints, fracture intersections, rock layer boundary points, and voids of each target area are obtained. Based on the fracture endpoints, fracture intersections, rock layer boundary points, and voids of each target area, nodes of each target area are generated.

[0012] Based on the core scan CT data of each target area, the fracture and rock layer boundaries are obtained, and the edges of each target area are generated based on the fracture and rock layer boundaries of each target area.

[0013] Obtain the attribute information for each target region, and add the attribute information to the nodes and edges of each target region;

[0014] Based on the acoustic emission technology experimental data of each target area, the type, damage and stress state of each crack are obtained, and the type, damage and stress state of each crack are added to the corresponding node;

[0015] Construct a graph model for each target region based on its nodes and edges.

[0016] Optionally, obtaining the attribute information of each target region and adding the attribute information to the nodes and edges of each target region specifically includes:

[0017] Obtain attribute information for each target region, wherein the attribute information includes the location and clustering coefficient of fracture endpoints, fracture intersections, rock layer boundary points and voids in each target region, as well as the permeability and elastic modulus of each fracture;

[0018] The locations and clustering coefficients of the fracture endpoints, fracture intersections, rock layer boundary points, and voids in each target region are added to the nodes of the corresponding graph model, and the permeability and elastic modulus of each fracture are added to the edges of the corresponding graph model.

[0019] Optionally, the step of constructing a crack model for each target region based on a graphical model and a mechanical model specifically includes:

[0020] Based on the material definition of the target region, the mechanical properties of the material in the mechanical model are determined, and the material is meshed to construct the mechanical model;

[0021] Discontinuities are introduced into the mechanical model, and crack propagation is simulated based on the objective function and the enhancement function at the crack tip. The corresponding force coefficients and crack propagation paths are then obtained.

[0022] The force coefficient calculation and crack propagation path are mapped to the corresponding nodes of the graphical model to generate the corresponding crack model.

[0023] Optionally, the fracture model based on each target region, combined with user requirements, obtains fracture values ​​for each target region, and constructs a target training set based on core scan CT data, acoustic emission technology experimental data, and fracture values ​​for all target regions. Specifically, this includes:

[0024] Based on the crack model of each target area, obtain the crack values ​​of each target area according to user needs;

[0025] The crack values ​​of each target region are used as labels for the core scan CT data and acoustic emission experimental data of each target region to construct training data;

[0026] Summarize all the training data and construct the target training set based on all the training data.

[0027] Optionally, training a target convolutional neural network based on the target training set to obtain a crack prediction model specifically includes:

[0028] Based on the target training set, the target convolutional neural network is iteratively trained, and the parameters of the target convolutional neural network are updated accordingly after each training session.

[0029] When the training requirements are met, training ends, and the target convolutional neural network obtained from the last training is output as the crack prediction model.

[0030] Optionally, the method based on the fracture model of each target region, combined with user requirements to obtain fracture values ​​for each target region, constructs a target training set based on core scan CT data, acoustic emission technology experimental data, and fracture values ​​of all target regions, and trains a target convolutional neural network based on the target training set to obtain a fracture prediction model, further including:

[0031] When user requirements change, based on the crack model of each target area, the first crack value of each target area is obtained according to the changed user requirements.

[0032] Based on the core scan CT data, acoustic emission technology experimental data, and the first fracture value of all target areas, a first target training set is constructed.

[0033] The first crack prediction model is obtained by training a target convolutional neural network based on the first target training set.

[0034] Furthermore, to achieve the above objectives, the present invention also provides a graph model-based numerical prediction system for cracks, wherein the graph model-based numerical prediction system for cracks includes:

[0035] The data acquisition module is used to acquire core scan CT data and acoustic emission technology experimental data of multiple target areas. Based on the acquired core scan CT data and acoustic emission technology experimental data of multiple target areas, a graphical model of each target area is constructed, and based on the graphical model of each target area, a crack model of the corresponding target area is constructed in combination with the mechanical model.

[0036] The model generation module is used to obtain the crack values ​​of each target area based on the crack model of each target area and combined with user requirements. Based on the core scan CT data, acoustic emission technology experimental data and crack values ​​of all target areas, a target training set is constructed, and a target convolutional neural network is trained based on the target training set to obtain a crack prediction model.

[0037] The result acquisition module is used to process the region to be predicted based on the crack prediction model whenever core scan CT data and acoustic emission technology experimental data of the region to be predicted are acquired, so as to obtain the crack value of the region to be predicted.

[0038] Furthermore, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a graph model-based crack numerical prediction program stored in the memory and executable on the processor, wherein when the graph model-based crack numerical prediction program is executed by the processor, it implements the steps of the graph model-based crack numerical prediction method as described above.

[0039] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a graph model-based crack numerical prediction program, which, when executed by a processor, implements the steps of the graph model-based crack numerical prediction method as described above.

[0040] In this invention, based on the core scan CT data and acoustic emission technology experimental data of multiple target areas, a graphical model of each target area is constructed. Based on the graphical model of each target area, a corresponding crack model is constructed in conjunction with a mechanical model. Based on the crack model of each target area, crack values ​​for each target area are obtained according to user requirements. A target training set is constructed based on the core scan CT data, acoustic emission technology experimental data, and crack values ​​of all target areas. A target convolutional neural network is trained based on the target training set to obtain a crack prediction model. Whenever core scan CT data and acoustic emission technology experimental data of a region to be predicted are obtained, the region to be predicted is processed based on the crack prediction model to obtain the crack values ​​for that region. This invention utilizes a graphical model derived from core scanning CT data and acoustic emission experimental data to describe microscopic crack structures, while a mechanical model can be extended to larger scales. This multi-scale integrated model can encompass different levels of physical mechanisms. The combination of the graphical and mechanical models enables global and local modeling of the entire crack process, effectively connecting macroscopic structures with microscopic crack features, and achieving the evaluation of crack behavior under multi-field coupling conditions. Finally, a crack prediction model that meets user needs can be obtained based on a transfer learning algorithm. Combined with the results obtained from the graphical model, accurate and rapid numerical estimation of cracks can be achieved. Attached Figure Description

[0041] Figure 1 This is a flowchart of a preferred embodiment of the crack numerical prediction method based on a graph model according to the present invention;

[0042] Figure 2 This is a schematic diagram of scanning CT in the crack numerical prediction method based on graph model of the present invention;

[0043] Figure 3 This is a schematic diagram of the graph model in the crack numerical prediction method based on graph model of the present invention;

[0044] Figure 4 This is a schematic diagram of the crack model in the crack numerical prediction method based on graph model of the present invention;

[0045] Figure 5 This is a schematic diagram of transfer learning in the crack numerical prediction method based on graph model of the present invention;

[0046] Figure 6 This is a schematic diagram of a technical solution in the crack numerical prediction method based on a graph model of the present invention;

[0047] Figure 7 This is a structural diagram of a preferred embodiment of the crack numerical prediction system based on a graph model of the present invention;

[0048] Figure 8This is a structural diagram of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, 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 of the invention and are not intended to limit the invention.

[0050] During oil and gas extraction, water injection and pressurization are often used to propagate fractures and increase flow channels. However, the extension of fractures can inadvertently activate natural underground faults, damaging existing geological structures and potentially causing rock fracturing and displacement, leading to other problems. Therefore, obtaining accurate fracture data helps optimize extraction processes, improve production efficiency, and reduce environmental risks, thus aiding oil and gas exploration and development to achieve efficient resource extraction and rational utilization.

[0051] However, currently, data acquisition for oil and gas well fracture modeling often requires huge costs and the actual samples obtained are relatively scarce. Furthermore, mechanical simulation modeling methods often still have some difficulties in simulating the interrelationships such as the intersection and bifurcation behavior of a large number of fractures. At the same time, they are more suitable for simulating large-scale fractures at the scale of hundreds of kilometers and kilometers, and it is often difficult to refine them at the scale of decimeters, centimeters, or even millimeters, which makes it impossible to accurately and conveniently obtain accurate fracture values.

[0052] To address one or more of the above-mentioned problems, this invention constructs a graphical model for each target region based on core scan CT data and acoustic emission technology experimental data acquired from multiple target regions. Based on the graphical model of each target region, a corresponding fracture model is constructed in conjunction with a mechanical model. Based on the fracture model of each target region, fracture values ​​for each target region are obtained according to user requirements. A target training set is constructed based on the core scan CT data, acoustic emission technology experimental data, and fracture values ​​of all target regions. A target convolutional neural network is trained using the target training set to obtain a fracture prediction model. Whenever core scan CT data and acoustic emission technology experimental data for a region to be predicted are acquired, the region to be predicted is processed based on the fracture prediction model to obtain the fracture values ​​for that region.

[0053] The preferred embodiment of the crack numerical prediction method based on a graph model described in this invention, such as... Figure 1 As shown, the crack numerical prediction method based on the graph model includes the following steps:

[0054] Step S10: Based on the core scan CT data and acoustic emission technology experimental data of multiple target areas, construct a graphical model of each target area, and based on the graphical model of each target area, construct a crack model of the corresponding target area in combination with the mechanical model.

[0055] Specifically, acquiring data before modeling cracks in a region often incurs significant costs, and the actual samples obtained are relatively scarce. Furthermore, mechanical simulation modeling methods often face difficulties in simulating the interrelationships, such as the intersection and bifurcation behaviors of numerous cracks. They are also more suitable for simulating large-scale cracks at the hundred-kilometer and kilometer levels, and often struggle to achieve finer detail at the decimeter, centimeter, or even millimeter scales. Graph models, however, successfully alleviate the shortcomings of mechanical simulation modeling in understanding crack interactions, relationships, network structures, and achieving finer detail at small scales.

[0056] It should be noted that core scanning CT scans, such as Figure 2 As shown, core scanning CT can acquire information such as crack endpoints, crack intersections, rock layer boundary points, voids, cracks, and rock layer boundaries. Acoustic emission (AE) technology is a non-destructive testing method that assesses material properties or structural integrity by receiving and analyzing acoustic emission signals from materials. Acoustic emission refers to the stress wave phenomenon generated by the rapid release of strain energy due to crack propagation, plastic deformation, or phase transformation in a material. Correspondingly, data such as crack type, damage, and stress state are obtained through experiments.

[0057] Furthermore, based on the acquired core scan CT data and acoustic emission technology experimental data of multiple target areas, a graphical model of each target area is constructed, specifically including:

[0058] Based on the core scan CT data of each target area, the fracture endpoints, fracture intersections, rock layer boundary points, and voids of each target area are obtained. Based on the fracture endpoints, fracture intersections, rock layer boundary points, and voids of each target area, nodes of each target area are generated.

[0059] Based on the core scan CT data of each target area, the fracture and rock layer boundaries are obtained, and the edges of each target area are generated based on the fracture and rock layer boundaries of each target area.

[0060] Obtain the attribute information for each target region, and add the attribute information to the nodes and edges of each target region;

[0061] Based on the acoustic emission technology experimental data of each target area, the type, damage and stress state of each crack are obtained, and the type, damage and stress state of each crack are added to the corresponding node;

[0062] Construct a graph model for each target region based on its nodes and edges.

[0063] Specifically, CT scans can provide three-dimensional images of the interior of rock materials, revealing the distribution and geometric features of crack endpoints, crack intersections, rock layer boundaries, voids, fissures, and other discontinuities. Data from CT scans can be directly used to construct the basic structure of nodes and edges in a graphical model, helping to more vividly depict the spatial distribution of cracks. Acoustic emission technology is primarily used to monitor the dynamic changes in crack formation, connection, and propagation during loading, i.e., data on crack type, damage, and stress state. Analyzing acoustic emission signals can identify different types of cracks and damage, infer the internal stress state and potential failure modes of the material, and help understand the mechanisms of crack formation. CT scans provide static structural features for the construction of the graphical model, while acoustic emission technology promptly reflects the material's response to dynamic changes. The combination of both provides more complete and robust data support for the graphical model to vividly display cracks.

[0064] In this invention, the creation of a graph model is a process of abstracting and digitizing the topological and geometric information of the fracture network. This invention utilizes collected geological data to construct the basic framework of the graph model, defining nodes and edges to simulate the initiation and expansion of large fractures and small fractures at the centimeter and millimeter scale. Nodes typically represent important spatial locations or features in the system, such as pores, fracture endpoints or intersections, or even the boundaries of various rock layers; edges typically represent the fracture itself or connect different nodes to determine the relationships between them. That is, nodes for each target region are generated based on fracture endpoints, fracture intersections, rock layer boundary points, and pores; edges for each target region are generated based on the fractures and rock layer boundaries of the target region. With the nodes and edges obtained, a graph model can be formed, as shown below. Figure 3 As shown.

[0065] Then, the nodes and edges in the graph model are assigned corresponding attributes. That is, the nodes can be given location information or clustering coefficients, while the edges can be assigned information such as the permeability or elastic modulus of the cracks.

[0066] After constructing the graph model and clarifying the topological relationships between nodes and edges, the timing and manner of crack propagation can be determined based on the maximum stress criterion or energy criterion. Furthermore, by changing the external environmental conditions, the graph model can be used to analyze the dynamic changes of cracks, identify potential crack propagation paths, and apply them to crack connectivity analysis or crack network flow simulation.

[0067] Furthermore, the step of obtaining attribute information for each target region and adding the attribute information to the nodes and edges of each target region specifically includes:

[0068] Obtain attribute information for each target region, wherein the attribute information includes the location and clustering coefficient of fracture endpoints, fracture intersections, rock layer boundary points and voids in each target region, as well as the permeability and elastic modulus of each fracture;

[0069] The locations and clustering coefficients of the fracture endpoints, fracture intersections, rock layer boundary points, and voids in each target region are added to the nodes of the corresponding graph model, and the permeability and elastic modulus of each fracture are added to the edges of the corresponding graph model.

[0070] Specifically, when assigning corresponding attributes to nodes and edges in the graph model, the obtained locations and clustering coefficients of crack endpoints, crack intersections, rock layer boundary points, and voids are added to the corresponding nodes of the graph model, and the permeability and elastic modulus of each crack are added to the corresponding edges of the graph model.

[0071] The creation of graphical models not only vividly displays the stress distribution within small-scale cracked materials, but also obtains the interrelationships between different cracks. This provides a solution to the difficulties in mechanical simulation crack modeling when it is impossible to simplify the simulation of small-scale cracks over a large area and to vividly display the intersection of a large number of cracks.

[0072] Furthermore, the construction of a crack model for each target region based on a graphical model and a mechanical model specifically includes:

[0073] Based on the material definition of the target region, the mechanical properties of the material in the mechanical model are determined, and the material is meshed to construct the mechanical model;

[0074] Discontinuities are introduced into the mechanical model, and crack propagation is simulated based on the objective function and the enhancement function at the crack tip. The corresponding force coefficients and crack propagation paths are then obtained.

[0075] The force coefficient calculation and crack propagation path are mapped to the corresponding nodes of the graphical model to generate the corresponding crack model.

[0076] Specifically, mechanical models can be coupled with other physical phenomena such as fluid mechanics to simulate the propagation behavior and morphology of fractures during hydraulic fracturing. This is very important in the mining of unconventional areas. For example, force models can be used to simulate the behavior and propagation process of fractures in the entire structure at the macroscopic level using the extended finite element method.

[0077] In this invention, the mechanical properties of the material must first be defined before crack modeling. Then, the material is meshed: a denser mesh is placed near the crack to capture local stress and displacement changes, while a sparser mesh can be used in other areas. Next, discontinuities are introduced, and a Heaviside function and a crack tip enhancement function are used to simulate crack propagation, with the Heaviside function serving as the objective function. After these preparations, a specific parameter level set function is set to track the crack path. Loads and boundary conditions are set, including forces or displacements applied to the structure. During crack propagation simulation, the gradual increase or variation of loads sometimes needs to be considered to simulate the actual loading process in a real working environment. Subsequently, the mechanical response, deformation, and displacement distribution of the overall model can be calculated. After the simulation is completed, post-processing analysis can be used to calculate the corresponding force coefficients and crack propagation paths, mapping them to the nodes of the graphical model to generate... Figure 4 The crack model shown facilitates the analysis of stress distribution in the construction area and also allows for updating the interrelationships and influence intensity between cracks.

[0078] In this invention, the crack model integrates a large-scale mechanical model and a small-scale graphical model to perform a more comprehensive analysis of the system at different scales and obtain a more comprehensive perspective. At the same time, when processing large-scale data, the message passing mechanism of the graphical model can also simplify the calculation process of the mechanical model and improve the model efficiency.

[0079] Step S20: Based on the crack model of each target area, obtain the crack value of each target area in combination with user needs, construct a target training set according to the core scan CT data, acoustic emission technology experimental data and crack values ​​of all target areas, and train the target convolutional neural network according to the target training set to obtain the crack prediction model.

[0080] Specifically, by combining graphical and mechanical models, the detailed situation during fracture propagation can be observed comprehensively and at multiple scales, including fracture orientation, stress distribution, and stress intensity coefficient at the fracture tip. This data provides a rich training dataset for subsequent transfer learning models, i.e., training of target convolutional neural networks. Simultaneously, by incorporating real-time monitoring parameters such as drilling fluid flow rate, applied pressure, and formation pore pressure, the simulation results of fracture propagation can be meticulously verified and adjusted, providing a precise theoretical foundation for the overall performance of subsequent model training.

[0081] Furthermore, the fracture model based on each target region, combined with user requirements, obtains fracture values ​​for each target region. Based on core scan CT data, acoustic emission experimental data, and fracture values ​​from all target regions, a target training set is constructed, specifically including:

[0082] Based on the crack model of each target area, obtain the crack values ​​of each target area according to user needs;

[0083] The crack values ​​of each target region are used as labels for the core scan CT data and acoustic emission experimental data of each target region to construct training data;

[0084] Summarize all the training data and construct the target training set based on all the training data.

[0085] In this invention, a dataset is constructed before training the corresponding network. During the construction process, corresponding crack values ​​are obtained from the crack model according to the user's needs, thereby constructing the target training set. That is, the crack values ​​are set according to the corresponding region, and the crack values ​​can be obtained from the crack model.

[0086] Further, the step of training a target convolutional neural network based on the target training set to obtain a crack prediction model specifically includes:

[0087] Based on the target training set, the target convolutional neural network is iteratively trained, and the parameters of the target convolutional neural network are updated accordingly after each training session.

[0088] When the training requirements are met, training ends, and the target convolutional neural network obtained from the last training is output as the crack prediction model.

[0089] like Figure 5 As shown, transfer learning models can delve into the complex characteristics of crack propagation from large datasets generated by combining graphical models and mechanical simulations, and can automatically identify their changing trends and patterns. Through transfer learning, a robust data-driven model can be established in a mature region and then transferred to a new region with scarce data to perform early crack prediction. This computation enhances the model's applicability to various geological conditions and crack characteristics.

[0090] In the transfer learning model construction stage, this invention employs convolutional neural networks and various algorithms to train the model. The training process, based on the corresponding target training set, allows the machine learning model to conduct in-depth analysis of the intrinsic mechanism and evolution pattern of crack propagation. For specific geomechanical backgrounds, it accurately grasps the generation, propagation, and evolution patterns of cracks during the operation of the target project, involving specific parameters such as location, width, depth, and directionality. Then, the model trained in the source region is transferred to a new region that is different from but related to the training region, where crack prediction is performed. Some new regions may possess specific characteristics, allowing for fine-tuning based on the new region. This can be achieved by freezing certain layers and updating only the last few layers to adapt to the characteristics of the new region.

[0091] Furthermore, in this invention, the training requirements are user-defined, such as reaching a certain number of training iterations or a certain level of model accuracy. Through repeated model training, continuous modification of algorithm parameter settings and architecture design, combined with cross-validation, model efficiency is optimized to achieve outstanding generalization ability and accurate prediction. After the model training reaches a certain level of effectiveness, existing field crack detection data can be used for comparison and analysis to ensure the effectiveness of transfer learning.

[0092] Furthermore, the fracture model based on each target region, combined with user requirements to obtain fracture values ​​for each target region, constructs a target training set based on core scan CT data, acoustic emission technology experimental data, and fracture values ​​from all target regions, and trains a target convolutional neural network based on the target training set to obtain a fracture prediction model, further including:

[0093] When user requirements change, based on the crack model of each target area, the first crack value of each target area is obtained according to the changed user requirements.

[0094] Based on the core scan CT data, acoustic emission technology experimental data, and the first fracture value of all target areas, a first target training set is constructed.

[0095] The first crack prediction model is obtained by training a target convolutional neural network based on the first target training set.

[0096] Specifically, since the prediction requirements differ for each region, the required crack values ​​vary. Therefore, in this invention, when user needs change, a first crack value is obtained from the crack model based on the corresponding user needs to construct a first target training set. Then, a corresponding target convolutional neural network is trained to obtain a first crack prediction model. This first crack prediction model can be applied to the current region to obtain the required crack values. Furthermore, in this invention, different crack prediction models are stored after acquisition so that they can be directly applied when needed.

[0097] Step S30: Whenever core scan CT data and acoustic emission technology experimental data of the area to be predicted are obtained, the area to be predicted is processed based on the crack prediction model to obtain the crack value of the area to be predicted.

[0098] Specifically, after obtaining the corresponding crack prediction model, whenever core scan CT data and acoustic emission technology experimental data of the area to be predicted are acquired, they can be input into the corresponding crack prediction model and the required crack values ​​can be output.

[0099] Furthermore, such as Figure 6The diagram illustrates one implementation process of the present invention. In this invention, data is first collected to construct a graphical model. Then, a mechanical model is integrated onto the graphical model to obtain a crack model, which includes the omnidirectional crack propagation path, propagation rate, and stress field distribution. Subsequently, the corresponding data can be obtained based on the crack model to realize model transfer and application, that is, the corresponding neural network or model is trained from the data to obtain the corresponding crack prediction model. The obtained crack prediction model can perform multi-level and multi-angle numerical prediction of cracks.

[0100] This invention constructs a graphical model for each target region based on core scan CT data and acoustic emission technology experimental data acquired from multiple target regions. Based on the graphical model of each target region, a corresponding fracture model is constructed by combining it with a mechanical model. Based on the fracture model of each target region, fracture values ​​for each target region are obtained according to user requirements. A target training set is constructed based on the core scan CT data, acoustic emission technology experimental data, and fracture values ​​of all target regions. A target convolutional neural network is then trained using the target training set to obtain a fracture prediction model. Whenever core scan CT data and acoustic emission technology experimental data for a region to be predicted are acquired, the region to be predicted is processed based on the fracture prediction model to obtain the fracture values ​​for that region. This invention utilizes a graphical model derived from core scanning CT data and acoustic emission experimental data to describe microscopic crack structures, while a mechanical model can be extended to larger scales. This multi-scale integrated model can encompass different levels of physical mechanisms. The combination of the graphical and mechanical models enables global and local modeling of the entire crack process, effectively connecting macroscopic structures with microscopic crack features, and achieving the evaluation of crack behavior under multi-field coupling conditions. Finally, a crack prediction model that meets user needs can be obtained based on a transfer learning algorithm. Combined with the results obtained from the graphical model, accurate and rapid numerical estimation of cracks can be achieved.

[0101] Furthermore, such as Figure 7 As shown, based on the above-mentioned graph model-based numerical prediction method for cracks, the present invention also provides a graph model-based numerical prediction system for cracks, wherein the graph model-based numerical prediction system for cracks includes:

[0102] The data acquisition module 71 is used to construct a graphical model of each target area based on the core scanning CT data and acoustic emission technology experimental data of multiple target areas, and to construct a corresponding crack model of the target area based on the graphical model of each target area and the mechanical model.

[0103] The model generation module 72 is used to obtain the crack values ​​of each target area based on the crack model of each target area and combined with the user's requirements. Based on the core scan CT data, acoustic emission technology experimental data and crack values ​​of all target areas, a target training set is constructed, and a target convolutional neural network is trained based on the target training set to obtain a crack prediction model.

[0104] The result acquisition module 73 is used to process the region to be predicted based on the crack prediction model whenever core scan CT data and acoustic emission technology experimental data of the region to be predicted are acquired, so as to obtain the crack value of the region to be predicted.

[0105] Furthermore, such as Figure 8 As shown, based on the above-mentioned graph model-based crack numerical prediction method and system, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 8 Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0106] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a graph model-based crack numerical prediction program 40, which can be executed by the processor 10 to implement the graph model-based crack numerical prediction method of the present invention.

[0107] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the graph model-based crack numerical prediction method.

[0108] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components 10-30 of the terminal communicate with each other via a system bus.

[0109] In one embodiment, the processor 10 implements the steps of the above-described graph model-based crack numerical prediction method when executing the graph model-based crack numerical prediction program 40 in the memory 20.

[0110] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a graph model-based crack numerical prediction program, which, when executed by a processor, implements the steps of the graph model-based crack numerical prediction method as described below:

[0111] Based on the core scan CT data and acoustic emission technology experimental data of multiple target areas, a graphical model of each target area was constructed, and based on the graphical model of each target area, a crack model of the corresponding target area was constructed in combination with the mechanical model.

[0112] Based on the crack model of each target area, the crack values ​​of each target area are obtained in combination with user needs. According to the core scan CT data, acoustic emission technology experimental data and crack values ​​of all target areas, a target training set is constructed, and a target convolutional neural network is trained according to the target training set to obtain a crack prediction model.

[0113] Whenever core scan CT data and acoustic emission technology experimental data of the area to be predicted are obtained, the area to be predicted is processed based on the crack prediction model to obtain the crack value of the area to be predicted.

[0114] Specifically, the construction of a graphical model for each target region based on core scan CT data and acoustic emission technology experimental data from multiple target regions includes:

[0115] Based on the core scan CT data of each target area, the fracture endpoints, fracture intersections, rock layer boundary points, and voids of each target area are obtained. Based on the fracture endpoints, fracture intersections, rock layer boundary points, and voids of each target area, nodes of each target area are generated.

[0116] Based on the core scan CT data of each target area, the fracture and rock layer boundaries are obtained, and the edges of each target area are generated based on the fracture and rock layer boundaries of each target area.

[0117] Obtain the attribute information for each target region, and add the attribute information to the nodes and edges of each target region;

[0118] Based on the acoustic emission technology experimental data of each target area, the type, damage and stress state of each crack are obtained, and the type, damage and stress state of each crack are added to the corresponding node;

[0119] Construct a graph model for each target region based on its nodes and edges.

[0120] Specifically, obtaining the attribute information of each target region and adding the attribute information to the nodes and edges of each target region includes:

[0121] Obtain attribute information for each target region, wherein the attribute information includes the location and clustering coefficient of fracture endpoints, fracture intersections, rock layer boundary points and voids in each target region, as well as the permeability and elastic modulus of each fracture;

[0122] The locations and clustering coefficients of the fracture endpoints, fracture intersections, rock layer boundary points, and voids in each target region are added to the nodes of the corresponding graph model, and the permeability and elastic modulus of each fracture are added to the edges of the corresponding graph model.

[0123] Specifically, the construction of a crack model for each target region based on a graphical model and a mechanical model includes:

[0124] Based on the material definition of the target region, the mechanical properties of the material in the mechanical model are determined, and the material is meshed to construct the mechanical model;

[0125] Discontinuities are introduced into the mechanical model, and crack propagation is simulated based on the objective function and the enhancement function at the crack tip. The corresponding force coefficients and crack propagation paths are then obtained.

[0126] The force coefficient calculation and crack propagation path are mapped to the corresponding nodes of the graphical model to generate the corresponding crack model.

[0127] Specifically, the fracture model for each target region, combined with user requirements, obtains fracture values ​​for each target region. Based on core scan CT data, acoustic emission experimental data, and fracture values ​​from all target regions, a target training set is constructed, including:

[0128] Based on the crack model of each target area, obtain the crack values ​​of each target area according to user needs;

[0129] The crack values ​​of each target region are used as labels for the core scan CT data and acoustic emission experimental data of each target region to construct training data;

[0130] Summarize all the training data and construct the target training set based on all the training data.

[0131] Specifically, the step of training a target convolutional neural network based on the target training set to obtain a crack prediction model includes:

[0132] Based on the target training set, the target convolutional neural network is iteratively trained, and the parameters of the target convolutional neural network are updated accordingly after each training session.

[0133] When the training requirements are met, training ends, and the target convolutional neural network obtained from the last training is output as the crack prediction model.

[0134] The process includes: 1) Obtaining fracture values ​​for each target region based on a fracture model for that region, combined with user requirements; 2) Constructing a target training set based on core scan CT data, acoustic emission experimental data, and fracture values ​​from all target regions; 3) Training a target convolutional neural network using the target training set to obtain a fracture prediction model; and 4) Further steps include:

[0135] When user requirements change, based on the crack model of each target area, the first crack value of each target area is obtained according to the changed user requirements.

[0136] Based on the core scan CT data, acoustic emission technology experimental data, and the first fracture value of all target areas, a first target training set is constructed.

[0137] The first crack prediction model is obtained by training a target convolutional neural network based on the first target training set.

[0138] 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 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. 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 that includes that element.

[0139] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The computer-readable storage medium can be a memory, magnetic disk, optical disk, etc.

[0140] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A numerical prediction method for cracks based on a graphical model, characterized in that, The graph-based numerical prediction method for cracks includes: Based on the core scan CT data and acoustic emission experimental data of multiple target areas, a graphical model of each target area was constructed, and based on the graphical model of each target area, a crack model of the corresponding target area was constructed in combination with the mechanical model. Based on the crack model of each target area, the crack values ​​of each target area are obtained in combination with user needs. According to the core scan CT data, acoustic emission technology experimental data and crack values ​​of all target areas, a target training set is constructed, and a target convolutional neural network is trained according to the target training set to obtain a crack prediction model. Whenever core scan CT data and acoustic emission technology experimental data of the area to be predicted are obtained, the area to be predicted is processed based on the crack prediction model to obtain the crack value of the area to be predicted.

2. The crack numerical prediction method based on a graph model according to claim 1, characterized in that, Based on the core scan CT data and acoustic emission technology experimental data of multiple target areas, a graphical model of each target area is constructed, specifically including: Based on the core scan CT data of each target area, the fracture endpoints, fracture intersections, rock layer boundary points, and voids of each target area are obtained. Based on the fracture endpoints, fracture intersections, rock layer boundary points, and voids of each target area, nodes of each target area are generated. Based on the core scan CT data of each target area, the fracture and rock layer boundaries are obtained, and the edges of each target area are generated based on the fracture and rock layer boundaries of each target area. Obtain the attribute information for each target region, and add the attribute information to the nodes and edges of each target region; Based on the acoustic emission technology experimental data of each target area, the type, damage and stress state of each crack are obtained, and the type, damage and stress state of each crack are added to the corresponding node; Construct a graph model for each target region based on its nodes and edges.

3. The crack numerical prediction method based on a graph model according to claim 2, characterized in that, The step of obtaining attribute information for each target region and adding the attribute information to the nodes and edges of each target region specifically includes: Obtain attribute information for each target region, wherein the attribute information includes the location and clustering coefficient of fracture endpoints, fracture intersections, rock layer boundary points and voids in each target region, as well as the permeability and elastic modulus of each fracture; The locations and clustering coefficients of the fracture endpoints, fracture intersections, rock layer boundary points, and voids in each target region are added to the nodes of the corresponding graph model, and the permeability and elastic modulus of each fracture are added to the edges of the corresponding graph model.

4. The crack numerical prediction method based on a graph model according to claim 2, characterized in that, The construction of a crack model for each target region based on a graphical model and a mechanical model specifically includes: Based on the material definition of the target region, the mechanical properties of the material in the mechanical model are determined, and the material is meshed to construct the mechanical model; Discontinuities are introduced into the mechanical model, and crack propagation is simulated based on the objective function and the enhancement function at the crack tip. The corresponding force coefficients and crack propagation paths are then obtained. The force coefficient calculation and crack propagation path are mapped to the corresponding nodes of the graphical model to generate the corresponding crack model.

5. The crack numerical prediction method based on a graph model according to claim 1, characterized in that, The fracture model based on each target region, combined with user requirements, obtains fracture values ​​for each target region. Based on core scan CT data, acoustic emission experimental data, and fracture values ​​from all target regions, a target training set is constructed, specifically including: Based on the crack model of each target area, obtain the crack values ​​of each target area according to user needs; The crack values ​​of each target region are used as labels for the core scan CT data and acoustic emission experimental data of each target region to construct training data; Summarize all the training data and construct the target training set based on all the training data.

6. The crack numerical prediction method based on a graph model according to claim 1, characterized in that, The step of training a target convolutional neural network based on the target training set to obtain a crack prediction model specifically includes: Based on the target training set, the target convolutional neural network is iteratively trained, and the parameters of the target convolutional neural network are updated accordingly after each training session. When the training requirements are met, training ends, and the target convolutional neural network obtained from the last training is output as the crack prediction model.

7. The crack numerical prediction method based on a graph model according to claim 1, characterized in that, The fracture model based on each target region, combined with user requirements, obtains fracture values ​​for each target region. Based on core scan CT data, acoustic emission experimental data, and fracture values ​​from all target regions, a target training set is constructed. A target convolutional neural network is then trained using this training set to obtain a fracture prediction model. The model further includes: When user requirements change, based on the crack model of each target area, the first crack value of each target area is obtained according to the changed user requirements. Based on the core scan CT data, acoustic emission technology experimental data, and the first fracture value of all target areas, a first target training set is constructed. The first crack prediction model is obtained by training a target convolutional neural network based on the first target training set.

8. A numerical prediction system for cracks based on a graphical model, characterized in that, The graph-based numerical prediction system for cracks includes: The data acquisition module is used to construct a graphical model of each target area based on the core scan CT data and acoustic emission technology experimental data of multiple target areas, and to construct a corresponding crack model of the target area based on the graphical model of each target area and the mechanical model. The model generation module is used to obtain the crack values ​​of each target area based on the crack model of each target area and combined with user requirements. Based on the core scan CT data, acoustic emission technology experimental data and crack values ​​of all target areas, a target training set is constructed, and a target convolutional neural network is trained based on the target training set to obtain a crack prediction model. The result acquisition module is used to process the region to be predicted based on the crack prediction model whenever core scan CT data and acoustic emission technology experimental data of the region to be predicted are acquired, so as to obtain the crack value of the region to be predicted.

9. A terminal, characterized in that, The terminal includes: a memory, a processor, and a graph model-based crack numerical prediction program stored in the memory and executable on the processor. When the graph model-based crack numerical prediction program is executed by the processor, it implements the steps of the graph model-based crack numerical prediction method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a graph-based numerical prediction program for cracks, which, when executed by a processor, implements the steps of the graph-based numerical prediction method for cracks as described in any one of claims 1-7.

Images