Hydraulic fracturing problem finite element modeling system and method based on large language model

Abstract

The invention discloses a hydraulic fracturing problem finite element modeling system and method based on a large language model. The hydraulic fracturing problem finite element modeling system based on the large language model comprises a client, a cloud service platform and a computing server. The client is composed of an interactive interface and a display interface; the cloud service platform comprises a storage space, a first input end, a second input end, a third input end, a transmission end, a trained large language model, a grid tool and a parameter grid coupling tool; the computing server comprises a storage unit, an operation unit and a visualization unit. The invention provides a hydraulic fracturing problem finite element modeling system based on a large language model, the large language model is trained, the large language model can generate a key parameter file according to user input, and a parameter grid coupling tool can generate a finite element model file according to the key parameter file and a grid file.

Description

Technical Field

[0001] The present invention relates to the field of numerical simulation, and in particular to a finite element modeling system and method for hydraulic fracturing problems based on a large language model. Background Art

[0002] Finite element analysis technology plays an important role in hydraulic fracturing, mainly reflected in its ability to accurately simulate complex physical phenomena. This technology can carefully depict the dynamic process of interaction between rocks and fluids during fracturing, including the formation, expansion and stress distribution of cracks. By creating high-resolution two-dimensional and three-dimensional models, finite element analysis helps engineers understand the evolution of fracture networks and their impact on production, thereby optimizing fracturing design and improving oil and gas recovery. In addition, it can integrate field measurement data, improve the reliability and accuracy of the model, and provide a scientific basis for decision-making. At the same time, the visualization ability of the finite element method makes complex processes more intuitive and helps to quickly identify potential problems. Finite element analysis is not only a basic tool for hydraulic fracturing research, but also an important means to improve engineering efficiency and safety.

[0003] Existing technologies still have shortcomings in the automated generation of hydraulic fracturing finite element models, mainly reflected in the limited level of intelligence, which makes it difficult to effectively replace manual professional judgment. Finite element modeling involves complex geometric processing, meshing, material property settings, and boundary condition definitions, which require in-depth professional knowledge and rich practical experience. Existing technologies are difficult to complete automatically through intelligent means. In addition, the automatically generated models often fail to meet the actual engineering accuracy requirements in terms of quality and computational stability, and still require manual adjustment and optimization by professionals. Therefore, the automated generation of finite element models still faces great challenges in terms of intelligence and accuracy assurance.

[0004] Large language models have the ability to achieve incremental learning, and can continuously improve modeling effects by gradually updating data and knowledge to adapt to dynamically changing needs. At the same time, large language models demonstrate strong data understanding and generation capabilities in automated modeling, and can automatically complete complex modeling processes, significantly reducing manual intervention. In addition, large language models have high computational efficiency and can quickly generate the required models when resources are reasonably allocated, greatly reducing time costs. Therefore, the use of large language model technology for automated modeling has great potential and is expected to play an important role in improving modeling efficiency and accuracy in the future.

[0005] The current technology has insufficient application scope in the automated generation of finite element models. For example, patent US10451520B2 (“Method and system for automated finite element model generation”) proposes a method for automatically generating finite element models through software tools. Although this method can achieve certain results in the modeling of certain standard geometric shapes and simple structures, its intelligence level is still limited when dealing with complex geometries, nonlinear material properties, and multi-physical field coupling problems. Specifically, the automated modeling process in this patent has high requirements for input data, and usually requires a lot of manual intervention to ensure the accuracy and stability of the model. Therefore, although the patent demonstrates certain automation potential, in actual engineering applications, it is still difficult to completely replace the experience and judgment of professionals, resulting in its adaptability and application scope being limited in a wider range of complex engineering scenarios. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention provides a finite element modeling system and method for hydraulic fracturing problems based on a large language model.

[0007] The finite element modeling system for hydraulic fracturing problems based on a large language model in the present invention comprises: a client, a cloud service platform and a computing server;

[0008] The client is composed of an interactive interface and a display interface; wherein,

[0009] The interactive interface includes a user input unit and a system output unit;

[0010] The display interface is a window for the system to display the visual finite element model and operation results;

[0011] The cloud service platform includes a storage space, a first input terminal, a second input terminal, a third input terminal, a transmission terminal, a trained large language model, a grid tool, and a parameter grid coupling tool;

[0012] The first input end is an input port for receiving external data input and user input and inputting into the large language model;

[0013] The second input terminal is an input port for receiving a key parameter file and inputting it into a grid tool;

[0014] The third input terminal is an input port for receiving key parameter files and grid files and inputting them into the parameter grid coupling tool;

[0015] The transmission end is a transmission port for transmitting the finite element model file in the storage space of the cloud service platform to the storage unit of the computing server;

[0016] The trained large language model is mainly a large language model that generates key parameter files based on user input content;

[0017] The grid tool is a tool that is called by the cloud service platform to generate a grid file according to the key parameter file;

[0018] The parameter grid coupling tool is a tool that is called by the cloud service platform to generate a finite element model file based on the key parameter file and the grid file;

[0019] The computing server includes a storage unit, an operation unit and a visualization unit.

[0020] The finite element modeling method for hydraulic fracturing problems based on a large language model in the present invention includes:

[0021] The user inputs the key description of the finite element model that he wants to generate in the user input unit, and the API interface is consistent with the API interface of the trained large language model;

[0022] The cloud service platform's large language model processes the description entered by the user and generates a key parameter file; the mesh tool reads the key parameter file and generates a mesh file; the parameter mesh coupling tool processes the key parameter file and mesh file and generates a finite element model file;

[0023] The computing server runs the finite element model file, and the finite element model and the running results are presented using the visualization unit. The system output unit prompts the user that the modeling and running have been completed, and the visual finite element model and the running results are displayed on the display interface;

[0024] The user determines whether it meets expectations; if it does not meet expectations, the user needs to re-enter the relevant description; if it meets expectations, the user can choose to output the finite element model and running results.

[0025] Beneficial effects of the present invention:

[0026] This application proposes a finite element modeling system for hydraulic fracturing problems based on a large language model, and trains a large language model. The large language model can generate key parameter files based on user input, and the parameter grid coupling tool can generate finite element model files based on the key parameter files and grid files.

[0027] The computing server of the present application is capable of running the finite element modeling files generated by the large language model, allowing the user and the system to continuously interact until an ideal finite element model is obtained, thereby realizing efficient acquisition of the finite element model. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 System and each file architecture diagram;

[0029] Figure 2 Diagram of the process of training a large language model;

[0030] Figure 3 Diagram of the training parameter grid coupling tool process;

[0031] Figure 4 Process diagram from user input to display of finite element model and operation results;

[0032] Figure 5 Classification diagram of sub-problems of hydraulic fracturing problem;

[0033] Figure 6 Flowchart of training a large language model;

[0034] Figure 7 Example diagram of key parameter files;

[0035] Figure 8 Users utilize the system to model flow charts. DETAILED DESCRIPTION

[0036] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings.

[0037] This application provides a finite element modeling system for hydraulic fracturing problems based on a large language model, such as Figure 1 As shown, the system consists of a client, a cloud service platform, and a computing server;

[0038] The client is composed of an interactive interface and a display interface; the interactive interface includes a user input unit and a system output unit; the user input unit is a window for users to input content in a composite form of text, pictures, and vector graphics; the system output unit is a window for the system to prompt the user; the display interface is a window for the system to display the visualized finite element model and operation results;

[0039] The cloud service platform includes a storage space, an input terminal 1, an input terminal 2, an input terminal 3, a transmission terminal, a large language model that has been trained, a grid tool, and a parameter grid coupling tool.

[0040] The storage space is used to store the database and key parameter files generated by the trained large language model, the mesh files generated by the mesh tool, and the finite element model files of the parameter mesh coupling tool.

[0041] Furthermore, the database stores all data that appears during the training of the large language model in the form of "keyword-data value", which facilitates the subsequent processing of user input content by the trained large language model to generate key parameter files.

[0042] Furthermore, the key parameter file is the result of a trained large language model processing user input content, and the parameter file format in the file is "keyword-data value".

[0043] Furthermore, the grid file is generated by the cloud service platform by calling the grid tool according to the key parameter file.

[0044] Furthermore, the finite element model file is generated by coupling the key parameter file and the mesh file through the parameter mesh coupling tool, and the finite element model file can be directly run by the running unit to generate the result.

[0045] The input terminal 1 is an input port for receiving external data input and user input and inputting into the large language model; the input terminal 2 is an input port for receiving key parameter files and inputting into the grid tool; the input terminal 3 is an input port for receiving key parameter files and grid files and inputting into the parameter grid coupling tool.

[0046] The transmission end is a transmission port for transmitting the finite element model file in the storage space of the cloud service platform to the storage unit of the computing server; the trained large language model is mainly a large language model that generates a key parameter file based on the user input content; the grid tool is a tool that is called by the cloud service platform to generate a grid file based on the key parameter file; the parameter grid coupling tool is a tool that is called by the cloud service platform to generate a finite element model file based on the key parameter file and the grid file.

[0047] The computing service platform includes a storage unit, an operation unit and a visualization unit; the storage unit is mainly a unit for storing finite element model files, where the finite element model files are transmitted from the storage space by the transmission end of the cloud service platform; the operation unit is a unit for running the finite element model files; the visualization unit is a unit for visualizing the finite element model and the operation results.

[0048] like Figure 1 As shown, the present application provides a finite element modeling system for hydraulic fracturing problems based on a large language model, which mainly includes two-step implementation routes, numbered ① and ② respectively, wherein the blue part is the system component, the yellow part is the file, and the purple part is the large language model that has completed training.

[0049] In one embodiment, the first step is to implement the route and train the large language model built into the system. The training process is mainly carried out on the cloud service platform: Figure 2 As shown in , a large amount of external data and human input content are first used as input content to train the large language model. During the training process, the main goal is to generate a database and key parameter files in the form of "keyword-data value"; Figure 3As shown, the mesh tool is called by the cloud service platform to generate a mesh file according to the key parameter file generated by the trained large language model, and the parameter mesh coupling tool is called by the cloud service platform to generate a finite element model file that can be run on the running unit according to the coupling of the key parameter file and the mesh file;

[0050] In one embodiment, the second step implements the route from user input to the computing server to the display interface to display the finite element model and the running results: Figure 4 As shown, input terminal 1 receives the content input by the user in the user input unit, and the trained large language model generates a key parameter file; input terminal 2 receives the key parameter file, and the mesh tool generates a mesh file; input terminal 3 receives the key parameter file and the mesh file, and the parameter mesh coupling tool generates a finite element model file; the finite element model file is transmitted to the storage unit of the computing server by the transmission terminal; the running unit runs the finite element model file, and the visualization unit processes the finite element model and the running results, and displays them to the user on the display interface.

[0051] like Figure 5 As shown, the hydraulic fracturing problem in this application includes many sub-problems, including fracture propagation mechanism, proppant selection, induced earthquake, wellbore damage, fracturing fluid optimization, pressure control, three-dimensional fracturing network evolution, microseismic interpretation, fracturing effect evaluation, backflow and other sub-problems. The subsequent sub-problem database will be constructed mainly based on the classification of sub-problems.

[0052] like Figure 6 As shown in the figure, the training process of the large language model for hydraulic fracturing problems includes the following steps. This example takes the "crack propagation mechanism", a sub-problem of the hydraulic fracturing problem, as an example:

[0053] A. Data preparation:

[0054] A-1. Data collection: Select a variety of data sources as external data, such as web crawlers, public data sets, digital books, etc., to ensure that the amount of data is large enough and as much as possible related to the "crack extension mechanism";

[0055] Further including: at the same time, human input of relevant content also needs to be used as input content in order to train a large language model that can generate key parameter files;

[0056] A-2. Data cleaning: Remove information irrelevant to the “crack extension mechanism” and remove duplicate content.

[0057] B. Data preprocessing:

[0058] B-1. Formatting: Convert documents and texts into a unified format that is readable by the large language model;

[0059] B-2. Construct a vocabulary: The vocabulary should include as many keywords as possible in the "crack propagation mechanism" problem, such as geometry keywords, material property keywords, initial condition keywords, boundary condition keywords, etc.

[0060] It further includes: B-3, word frequency filtering: some words expressing the same meaning are classified as the same keyword to reduce the size of the vocabulary; for example: "intersection angle" and "approach angle" are both classified as "intersection angle".

[0061] C. Training model and configuration settings:

[0062] C-1. Model architecture: Choose a suitable training model. You can use the existing pre-trained model as a basis to reduce training time and resources.

[0063] C-2. Set hyperparameters: determine the learning rate and batch size, and set the effective number of training rounds.

[0064] D. Model training:

[0065] D-1. Training goal: The goal is to generate a database and key parameter file in the form of "keyword-data value" in the "crack propagation mechanism" problem; the database contains all the keywords and related data values ​​of "crack propagation mechanism"; the key parameter file contains sub-problem names, geometric keywords, attribute keywords, grid keywords, etc.

[0066] D-2. Evaluation and adjustment: Manually evaluate the training results and adjust incorrect training results until the correct key parameter file is generated.

[0067] E. Model deployment:

[0068] E-1. Platform selection: The deployment platform is the cloud service platform;

[0069] E-2. API interface: Select the appropriate API interface to support real-time reasoning.

[0070] The key parameter file in this example includes the key parameters that describe a physical model. Here we take the "crack propagation mechanism" problem as an example:

[0071] Human input: A linear elastic fracture mechanics two-dimensional model (LEFM) is established with a length of 100m and a width of 50m. The initial crack shape is a rectangle and the position is on the left. The material elastic modulus is 1.5GPa, the Poisson's ratio is 0.18, the formation pressure is 9.81MPa, and the fracture criterion based on the stress intensity factor is adopted.

[0072] Key parameter files: can be improved in combination with the database, including sub-problem names, geometry keywords, attribute keywords, grid keywords, etc. Figure 7 As shown:

[0073] Sub-problem name: “Crack propagation mechanism—1”;

[0074] Geometry keywords: "2D - 1", "Length - 100m", "Width - 50m", "Rectangle - 1", "Left side - 1";

[0075] Attribute keywords: "formation pressure - 9.81MPa", "elastic modulus - 1.5GPa", "Poisson's ratio - 0.18", "stress intensity factor - 1";

[0076] Mesh keywords: "triangular element-1" etc.

[0077] After the above steps, the database of the sub-problem "crack extension mechanism" is obtained, which assists in generating the key parameter file next time; the large language model has been trained, and the content is input in the user input unit, and the generated result is the key parameter file; the database of other sub-problems of the hydraulic fracturing problem can also be created by this process, and the large language models of other sub-problems can also be trained by this process.

[0078] Furthermore, the parameter mesh coupling tool generates a finite element model file according to the coupling of the key parameter file and the mesh file. The finite element model file includes multiple parts: control statements, nodes, unit definitions, initial values, boundary conditions, material definitions, working conditions, time steps, etc.

[0079] like Figure 8 As shown, users can use this system to perform finite element modeling of hydraulic fracturing problems, including the following steps:

[0080] Step 1: The user inputs the key description of the finite element model that he wants to generate in the user input unit. The API interface is consistent with the API interface of the trained large language model.

[0081] Step 2: After the large language model of the cloud service platform processes the description input by the user, a key parameter file is generated, and the mesh tool reads the key parameter file and generates a mesh file; after the parameter mesh coupling tool processes the key parameter file and the mesh file, a finite element model file is generated;

[0082] The computing server runs the finite element model file, and the finite element model and the running results are presented using the visualization unit. The system output unit prompts the user that the modeling and running have been completed, and the visual finite element model and the running results are displayed on the display interface;

[0083] Step 3: The user determines whether it meets expectations; if it does not meet expectations, the user needs to re-enter the relevant description; if it meets expectations, the user can choose to output the finite element model and running results.

[0084] The contents described in the embodiments of this specification are merely an enumeration of implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A finite element modeling system for hydraulic fracturing problems based on a large language model, comprising a client, a cloud service platform and a computing server, characterized in that: The client is composed of an interactive interface and a display interface; wherein, The interactive interface includes a user input unit and a system output unit; The display interface is a window for the system to display the visual finite element model and operation results; The cloud service platform includes a storage space, a first input terminal, a second input terminal, a third input terminal, a transmission terminal, a trained large language model, a grid tool, and a parameter grid coupling tool; The first input end is an input port for receiving external data input and user input and inputting into the large language model; The second input terminal is an input port for receiving a key parameter file and inputting it into a grid tool; The third input terminal is an input port for receiving key parameter files and grid files and inputting them into the parameter grid coupling tool; The transmission end is a transmission port for transmitting the finite element model file in the storage space of the cloud service platform to the storage unit of the computing server; The trained large language model is mainly a large language model that generates key parameter files based on user input content; The grid tool is a tool that is called by the cloud service platform to generate a grid file according to the key parameter file; The parameter grid coupling tool is a tool that is called by the cloud service platform to generate a finite element model file based on the key parameter file and the grid file; The computing server includes a storage unit, an operation unit and a visualization unit.

2. The finite element modeling system for hydraulic fracturing problems based on a large language model according to claim 1, characterized in that: The user input unit is a window for users to input content in a composite form of text, pictures, or vector graphics; The system output unit is a window for the system to prompt the user.

3. The finite element modeling system for hydraulic fracturing problems based on a large language model according to claim 1, characterized in that: The storage space is a space for storing the database and key parameter files generated by the large language model that has completed training, the mesh files generated by the mesh tool, and the finite element model files of the parameter mesh coupling tool.

4. The finite element modeling system for hydraulic fracturing problems based on a large language model according to claim 3, characterized in that: The database stores all data that appear during the training of the large language model, and the data format is "keyword-data value"; The key parameter file is the result of the trained large language model processing the user input content, and the parameter file format in the file is "keyword-data value"; The grid file is generated by the cloud service platform calling the grid tool according to the key parameter file; The finite element model file is generated by coupling the key parameter file and the mesh file with the parameter mesh coupling tool, and the finite element model file can be directly run by the running unit to generate the result.

5. The finite element modeling system for hydraulic fracturing problems based on a large language model according to claim 1, characterized in that: The storage unit is mainly a unit for storing finite element model files; The running unit is a unit for running a finite element model file; The visualization unit is a unit for visualizing the finite element model and the operation results.

6. A finite element modeling method for hydraulic fracturing problems based on a large language model, using the system described in any one of claims 1 to 5, characterized in that The method comprises the following steps: The user inputs the key description of the finite element model that he wants to generate in the user input unit, and the API interface is consistent with the API interface of the trained large language model; The cloud service platform's large language model processes the user's input description and generates a key parameter file. The grid tool reads the key parameter file and generates a grid file. The parameter mesh coupling tool processes the key parameter files and mesh files to generate the finite element model files; The computing server runs the finite element model file, and the finite element model and the running results are presented using a visualization unit. The system output unit prompts the user that the modeling and running have been completed, and the visual finite element model and the running results are displayed on the display interface; Whether the user judges whether it meets expectations; If it does not meet the user's expectations, the user needs to re-enter the relevant description; if it meets the expectations, the user can choose to output the finite element model and running results.

7. The finite element modeling method for hydraulic fracturing problems based on a large language model according to claim 6 is characterized by: The hydraulic fracturing problems include fracture propagation mechanism, proppant selection, induced earthquake, wellbore damage, fracturing fluid optimization, pressure control, three-dimensional fracturing network evolution, microseismic interpretation, fracturing effect evaluation and backflow. X is used to represent one of the problems. The large language model training process of problem X is: A. Data preparation: A-1. Data collection: Select a variety of data sources as external data and ensure that the data volume is large enough; A-2. Data cleaning: remove information irrelevant to question X and remove duplicate content; B. Data preprocessing: B-1. Formatting: Convert documents and texts into a unified format that is readable by the large language model; B-2. Build a vocabulary: The vocabulary should include as many keywords as possible in question X; C. Training model and configuration settings: C-1. Model architecture: Select a training model. You can use an existing pre-trained model as a basis to reduce training time and resources. C-2. Set hyperparameters: determine the learning rate and batch size, and set the effective number of training rounds; D. Model training: D-1. Training goal: The goal is to generate a database and key parameter file in the form of "keyword-data value" in problem X; the database contains all the keywords and related data values ​​of problem X; the key parameter file contains the problem name, geometry keywords, attribute keywords and grid keywords; D-2. Evaluation and adjustment: Evaluate the training results and adjust incorrect training results until the correct key parameter file is generated; E. Model deployment: E-1. Platform selection: The deployment platform is the cloud service platform; E-2, API interface: Select the API interface to support real-time reasoning.

8. The finite element modeling method for hydraulic fracturing problems based on a large language model according to claim 7, characterized in that: It also includes B-3, word frequency filtering: some words expressing the same meaning are classified as the same keyword to reduce the size of the vocabulary.

9. The finite element modeling method for hydraulic fracturing problems based on a large language model according to claim 7, characterized in that: The parameter mesh coupling tool generates a finite element model file according to the coupling of the key parameter file and the mesh file. The finite element model file includes control statements, nodes, unit definitions, initial values, boundary conditions, material definitions, working conditions and time steps.

Images