Fluid mechanics control equation-oriented whole-process data processing method and system

Abstract

The application discloses a full-process data processing method and system for fluid mechanics control equations, and local servers are initialized and artificial intelligence agents are used to solve paths related to fluid mechanics control equations, so that the parameter space of a fluid mechanics problem is imported and disassembled; the artificial intelligence agent is used to identify the control equation corresponding to the fluid mechanics problem and batch match example files, and core parameters related to the control equation are extracted; the artificial intelligence agent is used to configure parameters of the control equation; the example file is converted in format and saved locally to the corresponding path; a solver of the control equation is called to perform solving processing, and the result of the solving processing is verified to determine the validity of the result. By constructing a special link for solving the fluid mechanics control equation, automation of the full-process link from problem parameter disassembly, control equation identification, special parameter configuration, format conversion and result verification is realized, and the example processing efficiency and solving accuracy are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of AI data processing, and in particular to a full-process data processing method and system for fluid mechanics control equations. BACKGROUND

[0002] Computational fluid dynamics (CFD) solves fluid mechanics control equations (such as continuity equation, momentum equation, energy equation, etc.) by computer and numerical method, realizes simulation analysis of fluid motion, heat transfer, mass transfer, etc., and is widely used in aerospace, automobile engineering, energy power, etc. The fluid mechanics control equation has the characteristics of strong nonlinearity, multi-physical field coupling, boundary condition sensitivity, etc., and its solution needs to rely on accurate grid division, reasonable selection of turbulence model, and adaptive numerical discrete format, and different types of fluid problems (incompressible flow, compressible flow, multi-directional flow, etc.) have significant differences in the shape and parameter requirements of the control equation.

[0003] The traditional computational fluid dynamics (CFD) processing has the following two core problems: first, the contradiction between fluid mechanics special requirements and manual operation, the construction of the example needs to manually adjust the key parameters (such as the setting of the viscosity coefficient of the momentum equation, the configuration of the thermal conductivity coefficient of the energy equation, etc.) according to the characteristics of the control equation, and the grid density needs to match the accuracy requirement of the equation solution (i.e. the grid in the boundary layer area needs to be encrypted to ensure the accuracy of the gradient calculation), manual operation is easy to cause the mismatch between the parameters and the equation type, and then affect the convergence of the solution; second, the generality process is out of touch with the solution of fluid mechanics, the existing example processing (such as file import, solver calling, etc.) is mostly computer general operation, and no special link is designed for the solution logic of fluid mechanics control equation, for example, when solving compressible flow, the Mach number, specific heat capacity, etc. Parameters need to be configured additionally, and the traditional process cannot automatically identify the equation type and match the corresponding parameters, which needs manual secondary modification, resulting in low processing efficiency.

[0004] The existing AI technology and the combination mode of the automation workflow tool cannot design a special link for the solution requirements of the fluid mechanics control equation, and still stay at the general file processing level, which leads to the technical problem that the general process cannot adapt to the special requirements of fluid mechanics. SUMMARY

[0005] Considering that the existing fluid mechanics control equation processing process cannot identify the control equation and match the corresponding network and numerical parameters, and cannot design a special link for the solution requirements of the fluid mechanics control equation, which leads to the problem that the general process cannot adapt to the special requirements of fluid mechanics. In view of the above problems, the present application is proposed to provide a full-process data processing method for fluid mechanics control equations to overcome the above problems or at least partially solve the above problems, which comprises:

[0006] initializing a path and an artificial intelligence agent related to fluid mechanics control equation solving on a local server, so as to import and disassemble a parameter space of a fluid mechanics problem;

[0007] identifying, by the artificial intelligence agent, a control equation corresponding to the fluid mechanics problem and a batch of matched example files, and extracting core parameters related to the control equation;

[0008] configuring parameters of the control equation by the artificial intelligence agent, and converting a format of the example files and saving the example files locally to a corresponding path;

[0009] calling a solver of the control equation to perform solving processing, and verifying a result of the solving processing, so as to determine validity of the result.

[0010] Optionally, initializing a path and an artificial intelligence agent related to fluid mechanics control equation solving on a local server, so as to import and disassemble a parameter space of a fluid mechanics problem, comprising:

[0011] deploying an automatic workflow platform on the local server, and setting a connection parameter of the artificial intelligence agent and a large language model in the automatic workflow platform;

[0012] initializing a path related to fluid mechanics control equation solving according to the automatic workflow platform, wherein the path is matched with a node of the automatic workflow platform;

[0013] importing a parameter space of a fluid mechanics problem to be solved by the node of the automatic workflow platform, and disassembling the parameter space into a plurality of executable steps according to a preset fluid mechanics knowledge graph by the artificial intelligence agent, wherein the plurality of executable steps comprise control equation selection, grid parameter configuration, boundary condition setting, and solving parameter definition.

[0014] Optionally, identifying, by the artificial intelligence agent, a control equation corresponding to the fluid mechanics problem and a batch of matched example files, and extracting core parameters related to the control equation, comprising:

[0015] identifying, by the artificial intelligence agent, a control equation type corresponding to the fluid mechanics problem, and batch-matching a corresponding example file in a local example warehouse according to the control equation type by a file screening node of the automatic workflow platform, and extracting core parameters related to the control equation from the example file;

[0016] configuring parameters of the control equation by the artificial intelligence agent, and converting a format of the example files and saving the example files locally to a corresponding path, comprising:

[0017] automatically modifying the calculation example file according to a solution requirement of the control equation; adjusting a grid parameter according to a gradient requirement of the control equation; and configuring a numerical parameter according to the control equation type;

[0018] converting the configured numerical parameter into a format supported by fluid mechanics solving software through a file processing node of the automated workflow platform, and saving the configured numerical parameter to a corresponding path according to the control equation type.

[0019] Optionally, adjusting the grid parameter according to the gradient requirement of the control equation comprises:

[0020] obtaining the control equation type corresponding to the fluid mechanics problem through the artificial intelligence agent;

[0021] analyzing a gradient source term corresponding to the control equation type according to the control equation type corresponding to the fluid mechanics problem, wherein the gradient source term refers to a physical quantity parameter term affecting a gradient change;

[0022] normalizing the physical quantity parameter term corresponding to the gradient source term corresponding to the control equation type to obtain a normalized physical quantity parameter term;

[0023] forming a gradient correlation vector corresponding to the control equation type by using the normalized physical quantity parameter term;

[0024] obtaining a gradient potential field Φ(x) by using the gradient correlation vector corresponding to the control equation type; wherein the gradient potential field Φ(x) is obtained by the following formula: wherein Φ(x) represents the gradient potential field; S(x) represents a relative intensity of the gradient source term, and S(x) =║X║2×(X0+║X║2) -1 , wherein║X║2 represents an L2 norm corresponding to the gradient correlation vector; X0 represents a reference value of the same dimension as the L2 norm corresponding to the gradient correlation vector; n represents a number of vector elements contained in the gradient correlation vector; λ i represents a weight value corresponding to the i-th vector element; X i represents a numerical value corresponding to the i-th vector element; G(x) represents a reinforcement factor, wherein the reinforcement factor is used to enhance the grid configuration intensity of the gradient vector in a key region, and the reinforcement factor G(x) is obtained by the following formula: wherein α represents a reinforcement intensity coefficient, and the value range of α is 0.5-1.3; β represents a stability period number, and the value range of β is 1-1.14; F(x) represents a divergence corresponding to the gradient correlation vector field; F0 represents a preset divergence reference value;

[0025] The current grid size is adjusted by using the gradient potential field Φ(x), and the adjusted gradient potential field Φ(x) is obtained by the following formula:

[0026] Wherein, represents the adjusted grid size, η represents the intensity adjustment coefficient, and the value is 0.7-1.5; Δ base represents the grid size before adjustment.

[0027] Optionally, the solver of the control equation is called for calculation processing, and the result of the calculation processing is verified to determine the validity of the result, including:

[0028] The exclusive calculation instruction is constructed according to the control equation type through the command line execution node of the automatic workflow platform, so as to call the control equation solver for calculation processing.

[0029] The calculation processing log is monitored in real time through the artificial intelligence agent, and the validity of the result of the calculation processing is determined according to the calculation processing log and the convergence characteristics of the control equation.

[0030] As an aspect of the present application, the embodiment of the present application also provides a full-process data processing system for fluid mechanics control equation, including:

[0031] An initialization module is configured to initialize a path and an artificial intelligence agent related to fluid mechanics control equation calculation on a local server, so as to import and disassemble a parameter space of a fluid mechanics problem.

[0032] An example file processing module is configured to identify a control equation corresponding to the fluid mechanics problem and batch match an example file through the artificial intelligence agent, and extract a core parameter related to the control equation.

[0033] A parameter configuration module is configured to perform parameter configuration on the control equation through the artificial intelligence agent.

[0034] A file conversion and saving module is configured to convert a format of the example file and save locally to a corresponding path.

[0035] A calculation and verification module is configured to call a solver of the control equation for calculation processing, and verify a result of the calculation processing, so as to determine the validity of the result.

[0036] Optionally, the initialization module is configured to initialize a path and an artificial intelligence agent related to fluid mechanics control equation calculation on a local server, so as to import and disassemble a parameter space of a fluid mechanics problem, including:

[0037] Deploy an automated workflow platform on a local server, and integrate an artificial intelligence agent into the automated workflow platform and configure the connection parameters between the artificial intelligence agent and the large language model;

[0038] Based on the automated workflow platform, initialize the path related to solving the fluid dynamics control equations; wherein the path matches the nodes of the automated workflow platform;

[0039] The parameter space of the fluid mechanics problem to be solved is imported through the nodes of the automated workflow platform. The artificial intelligence agent decomposes the parameter space into several executable steps according to the preset fluid mechanics knowledge graph. The several executable steps include the selection of control equations, configuration of mesh parameters, setting of boundary conditions, and definition of solution parameters.

[0040] Optionally, the example file processing module is used to identify the governing equations corresponding to the fluid mechanics problem and batch match example files through the artificial intelligence agent, and extract core parameters related to the governing equations, including:

[0041] The AI ​​agent identifies the type of control equations corresponding to the fluid mechanics problem. Based on the type of control equations, the AI ​​workflow platform uses a file filtering node to batch match the corresponding case files in the local case repository and extracts the core parameters related to the control equations from the case files.

[0042] The parameter configuration module is used to configure the parameters of the control equations through the artificial intelligence agent, including:

[0043] The AI ​​agent automatically modifies the example file according to the solution requirements of the control equations; adjusts the mesh parameters according to the gradient requirements of the control equations; and configures numerical parameters according to the type of control equation.

[0044] The file conversion and saving module is used to convert the format of the example file and save it locally to the corresponding path, including:

[0045] The configured numerical parameters are converted into a format supported by the fluid dynamics solution software through the file processing node of the automated workflow platform, and saved to the corresponding path according to the type of the governing equation.

[0046] Optionally, adjusting the mesh parameters according to the gradient requirements of the governing equations includes:

[0047] Retrieve the type of control equations corresponding to the fluid mechanics problem identified by the artificial intelligence agent;

[0048] Based on the type of governing equations corresponding to the fluid mechanics problem, the gradient source terms corresponding to the type of governing equations are analyzed, where the gradient source terms refer to the physical quantity parameter terms that affect the gradient change;

[0049] Normalize the physical quantity parameter terms corresponding to the gradient source terms of the control equation type to obtain the normalized physical quantity parameter terms.

[0050] The normalized physical quantity parameter terms are used to form a gradient correlation vector corresponding to the type of the governing equation;

[0051] The gradient potential field Φ(x) is obtained using the gradient correlation vector corresponding to the type of the governing equation; wherein, the gradient potential field Φ(x) is obtained by the following formula: Where Φ(x) represents the gradient potential field; S(x) represents the relative intensity of the gradient source term, and S(x) = ║X║2×(X0+║X║2) -1 ║X║2 represents the L2 norm of the gradient correlation vector; X0 represents a reference value with the same dimensions as the L2 norm of the gradient correlation vector; n represents the number of vector elements contained in the gradient correlation vector; λ i X represents the weight value corresponding to the i-th vector element; i Let G(x) represent the numerical value corresponding to the i-th vector element; G(x) represents the enhancement factor, which is used to enhance the grid configuration strength of the gradient vector in the key region, and the enhancement factor G(x) is obtained by the following formula: Where α represents the strengthening intensity coefficient, with a value ranging from 0.5 to 1.3; β represents the stability period, with a value ranging from 1 to 1.14; F(x) represents the divergence corresponding to the gradient correlation vector field; and F0 represents the preset divergence reference value.

[0052] The current mesh size is adjusted using the gradient potential field Φ(x), and the adjusted gradient potential field Φ(x) is obtained using the following formula:

[0053] Where Г represents the adjusted mesh size, η represents the strength adjustment coefficient, and its value ranges from 0.7 to 1.5; Δ base This indicates the grid size before adjustment.

[0054] Optionally, the solution and verification module is used to call the solver of the governing equations to perform solution processing and verify the results of the solution processing to determine the validity of the results, including:

[0055] Through the command line execution node of the automated workflow platform, a dedicated solution instruction is constructed according to the type of the control equation, thereby calling the control equation solver for solution processing;

[0056] The AI ​​agent monitors the solution processing log in real time, and judges the validity of the solution processing result based on the solution processing log and the convergence characteristics of the control equation.

[0057] The beneficial effects of the above-mentioned technical solutions provided in the embodiments of the present invention include at least the following:

[0058] This invention provides a full-process data processing method and system for fluid mechanics control equations. The system initializes paths and an AI agent related to solving fluid mechanics control equations on a local server, thereby importing and decomposing the parameter space of the fluid mechanics problem. The AI ​​agent identifies the control equations corresponding to the fluid mechanics problem and batch-matches example files, extracting core parameters related to the control equations. The AI ​​agent configures the parameters of the control equations. The example files are converted in format and saved locally to the corresponding path. The solver for the control equations is invoked for solution processing, and the results are verified to determine their validity. By constructing a dedicated link for solving fluid mechanics control equations, an automated full-process link is achieved, from problem parameter decomposition, control equation identification, dedicated parameter configuration, format conversion, and disassembly verification, improving the efficiency and accuracy of example processing.

[0059] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0060] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0061] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0062] Figure 1 This is a flowchart illustrating the full-process data processing method for fluid dynamics control equations provided in an embodiment of the present invention.

[0063] Figure 2 This is a schematic diagram of the structure of a full-process data processing system for fluid dynamics control equations provided in an embodiment of the present invention. Detailed Implementation

[0064] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0065] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," "far," "near," "front," and "rear," etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the accompanying drawings and are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0066] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0067] Please see Figure 1 As shown, an embodiment of this application provides a full-process data processing method for fluid dynamics control equations. This full-process data processing method for fluid dynamics control equations includes:

[0068] The path and AI agent related to the solution of fluid dynamics control equations are initialized on the local server to import and decompose the parameter space of the fluid dynamics problem; the AI ​​agent mentioned above refers to AIagent, which acts as an AI intermediary to perform corresponding task processing functions on the local server.

[0069] By using an artificial intelligence agent to identify the control equations corresponding to fluid mechanics problems and batch matching example files, the core parameters related to the control equations are extracted.

[0070] The AI ​​agent configures the parameters of the control equations; converts the format of the example files and saves them locally to the corresponding paths;

[0071] The solver for the governing equations is invoked for calculation, and the results are verified to determine their validity. The beneficial effect of the above embodiment is that this full-process data processing method for fluid dynamics governing equations, by constructing a dedicated link for solving fluid dynamics governing equations, achieves an automated full-process link from problem parameter decomposition, governing equation identification, dedicated parameter configuration, format conversion, and disassembly verification, thereby improving the efficiency and accuracy of example processing.

[0072] In another embodiment, the path and AI agent related to solving the fluid dynamics governing equations are initialized on a local server to import and decompose the parameter space of the fluid dynamics problem, including:

[0073] Deploy an automated workflow platform on a local server, integrate an AI agent into the automated workflow platform, and configure the connection parameters between the AI ​​agent and the large language model.

[0074] Based on the automated workflow platform, initialize the paths related to solving the fluid dynamics control equations; where the paths match the nodes of the automated workflow platform.

[0075] The parameter space of the fluid mechanics problem to be solved is imported through the nodes of the automated workflow platform. The artificial intelligence agent decomposes the parameter space into several executable steps based on the preset fluid mechanics knowledge graph. These executable steps include the selection of governing equations, configuration of mesh parameters, setting of boundary conditions, and definition of solution parameters.

[0076] Among them, the automated workflow platform can be implemented as an n8n workflow platform, for example.

[0077] The beneficial effects of the above embodiments are that an automated workflow platform serves as the carrier of the automated workflow. This automated workflow platform integrates an artificial intelligence agent function module, capable of running the corresponding Large Language Model (LLM). In actual operation, the automated workflow platform is first deployed on a local server. The workflow platform integrates the artificial intelligence agent and configures its connection parameters with the LLM, enabling the artificial intelligence agent to connect to and invoke the LLM. The automated workflow platform contains multiple nodes matching the workflow. The solution of the fluid dynamics control equations involves multiple process steps. To ensure the workflow platform can adapt to the solution of the fluid dynamics control equations, paths related to the solution of the fluid dynamics control equations are initialized on the platform, ensuring that these paths are compatible with the platform's nodes.

[0078] Furthermore, fluid mechanics problems involve solving for different parameters. To ensure that all parameters can be adapted for processing, the parameter space of the fluid mechanics problem to be solved is imported through nodes in the automated workflow platform, allowing each platform node to undertake all tasks related to processing the parameter space. Specifically, an AI agent, based on a pre-defined fluid mechanics knowledge graph, breaks down the parameter space into executable steps such as selecting governing equations, configuring mesh parameters, setting boundary conditions, and defining solution parameters, ensuring that the AI ​​agent processes each executable step independently.

[0079] In another embodiment, adjusting the mesh parameters according to the gradient requirements of the governing equation includes:

[0080] Retrieve the type of control equations corresponding to the fluid mechanics problem identified by the artificial intelligence agent;

[0081] Based on the type of governing equations corresponding to the fluid mechanics problem, the gradient source terms corresponding to the type of governing equations are analyzed, where the gradient source terms refer to the physical quantity parameter terms that affect the gradient change;

[0082] Normalize the physical quantity parameter terms corresponding to the gradient source terms of the control equation type to obtain the normalized physical quantity parameter terms.

[0083] The normalized physical quantity parameter terms are used to form a gradient correlation vector corresponding to the type of the governing equation;

[0084] The gradient potential field Φ(x) is obtained using the gradient correlation vector corresponding to the type of the governing equation; wherein, the gradient potential field Φ(x) is obtained by the following formula: Where Φ(x) represents the gradient potential field; S(x) represents the relative intensity of the gradient source term, and S(x) = ║X║2×(X0+║X║2) -1 ║X║2 represents the L2 norm of the gradient correlation vector; X0 represents a reference value with the same dimensions as the L2 norm of the gradient correlation vector; n represents the number of vector elements contained in the gradient correlation vector; λ i X represents the weight value corresponding to the i-th vector element; i Let G(x) represent the numerical value corresponding to the i-th vector element; G(x) represents the enhancement factor, which is used to enhance the grid configuration strength of the gradient vector in the key region, and the enhancement factor G(x) is obtained by the following formula: Where α represents the strengthening intensity coefficient, with a value ranging from 0.5 to 1.3; β represents the stability period, with a value ranging from 1 to 1.14; F(x) represents the divergence corresponding to the gradient correlation vector field; and F0 represents the preset divergence reference value.

[0085] The current mesh size is adjusted using the gradient potential field Φ(x), and the adjusted gradient potential field Φ(x) is obtained using the following formula:

[0086] Where Г represents the adjusted mesh size, η represents the strength adjustment coefficient, and its value ranges from 0.7 to 1.5; Δ base This indicates the grid size before adjustment.

[0087] The beneficial effects of the above embodiments are achieved through an intelligent pre-configuration method that integrates physical knowledge and mathematical models. This generates an optimized mesh highly suited to specific flow physics and user needs from the very beginning of the computation, thus achieving a synergistic leap in efficiency, accuracy, and automation from the source. This overcomes the inherent limitations of traditional mesh adaptation techniques that rely on solver feedback and human experience. Simultaneously, by using AI to identify the type of control equations and analyze gradient source terms, it can selectively extract physical quantities affecting gradient changes (such as velocity gradients and pressure gradients), avoiding the insufficient adaptability of "general-purpose meshes" to different fluid problems, and allowing mesh adjustments to better fit the physical essence of specific problems. The L2 norm and divergence of gradient correlation vectors significantly increase in "critical regions where gradient changes are drastic." Furthermore, combined with the enhancing effect of the strengthening factor, the gradient potential field increases in these regions. Ultimately, the mesh size is reduced through the mesh adjustment formula, improving the computational resolution in critical regions and reducing numerical errors. Moreover, the mesh parameters are directly mapped and generated by analytical formulas, avoiding repeated trial-and-error based on error estimation or gradient threshold determination and global / local mesh reconstruction operations, significantly reducing computational resource consumption purely for mesh adaptation. Through mechanisms such as "gradient potential field" and "local aggregation degree", it is possible to proactively identify and enhance the encryption of key regions with drastic changes in physical quantities, such as boundary layer, shock wave, and shear layer, without relying on any instantaneous solution, thereby improving the ability to distinguish complex flow phenomena.

[0088] In another embodiment, an artificial intelligence agent is used to identify the governing equations and batch-matched case files corresponding to the fluid mechanics problem, and to extract the core parameters related to the governing equations, including:

[0089] The AI ​​agent identifies the type of control equations for fluid mechanics problems. Based on the control equation type, the automated workflow platform uses a file filtering node to batch match the corresponding case files in the local case repository and extracts the core parameters related to the control equations from the case files.

[0090] The AI ​​agent configures the parameters of the governing equations; it converts the format of the example files and saves them locally to the corresponding paths, including:

[0091] The AI ​​agent automatically modifies the example files according to the solution requirements of the control equations; adjusts the mesh parameters according to the gradient requirements of the control equations; and configures the numerical parameters according to the type of control equation.

[0092] The configured numerical parameters are converted into a format supported by the fluid dynamics solution software through the file processing node of the automated workflow platform, and saved to the corresponding path according to the type of governing equation.

[0093] Among them, the fluid dynamics solution software can be implemented as VirtualFlow software, for example.

[0094] The beneficial effects of the above embodiments are that different types (forms) of fluid dynamics control equations correspond to different example files during the solution process. To ensure that each type of fluid dynamics control equation has accurate example file support, an artificial intelligence agent uses the decomposed control equation type and the platform's file filtering node to batch match the corresponding example files in the local example repository, extracting the core parameters related to the control equations from the example files. The core parameters differ for different types of control equations; for example, the core parameter for the momentum equation is the viscosity coefficient, and the core parameter for the energy equation is the thermal conductivity coefficient. Extracting core parameters in this way can meet the solution requirements of different types of control equations.

[0095] Furthermore, different types of governing equations require different example files. To ensure effective processing of these example files, an AI agent is used to automatically modify them based on the solution requirements of the governing equations; adjust mesh parameters according to the gradient requirements of the governing equations; configure numerical parameters according to the type of governing equation; and convert the configured numerical parameters into a format supported by the fluid dynamics solution software through the file processing node of the automated workflow platform, saving them to the corresponding path according to the type of governing equation. By converting the example files into their formats, compatibility between the example files and the corresponding software can be ensured, thus improving the efficiency of example file processing.

[0096] In another embodiment, the solver of the governing equations is invoked to perform the solution process, and the results of the solution process are verified to determine the validity of the results, including:

[0097] The command-line execution node of the automated workflow platform constructs a dedicated solution instruction based on the type of control equation, thereby calling the control equation solver for solution processing; the control equation solver can be implemented as, for example, the transatmb solver.

[0098] The validity of the solution processing results is judged by real-time monitoring of the solution processing logs and the convergence characteristics of the control equations based on the solution processing logs and the convergence characteristics of the control equations.

[0099] The beneficial effects of the above embodiments, in practical operation, are that, taking the type of control equation as a reference, the platform's command-line execution node first constructs solution instructions that match the control equation, ensuring that the solution instructions can calculate and process each solution step of the control equation, and then calls the control equation solver for solution processing, achieving efficient solution of the control equation. Furthermore, an artificial intelligence agent monitors the solution processing log in real time. Based on the solution processing log and the convergence characteristics of the control equation, if the solution processing of the control equation meets the convergence requirements, the solution processing result is determined to be valid; otherwise, the solution processing result is determined to be invalid, and the control equation is re-solved until the convergence requirements are met.

[0100] This invention takes the CFD example of "flow around an incompressible two-dimensional cylinder" as an example, aiming to achieve full automation from problem parameter input to solution verification. The fluid dynamics governing equations are the incompressible Navier-Stokes equations (continuity equation + momentum equation, excluding energy equation). The core requirements include:

[0101] Mesh: The boundary layer region of the cylindrical surface is meshed with finer mesh (minimum element size 0.005m), while the far-field region has a sparser mesh (element size 0.1m).

[0102] Numerical parameters: The finite volume method is used to discretize the equations, and the SIMPLE algorithm is used to solve the pressure-velocity coupling. The convergence criterion is that the residual is less than 1e-3.

[0103] Boundary conditions: velocity inlet (flow velocity 10 m / s), pressure outlet (static pressure 0 Pa), and no-slip boundary on the cylindrical surface.

[0104] The process for handling the "flow around an incompressible two-dimensional cylinder" is as follows:

[0105] (1) Deployment and environment configuration of automated workflow platform

[0106] Install n8n by executing npm install-gn8n on the local server, start the platform by n8n start --port 5678, and access http: / / serverIP:5678 to enter the visual interface;

[0107] AI Agent Integration and Large Language Model LLM Configuration:

[0108] Add an "AI Agent" module to the automated workflow platform, configure the connection parameters with the target LLM (deepseek / deepseek-r1-0528-qwen3-8b); the API address is https: / / api.openrouter.ai / v1 / chat / completions, the API key is obtained through the OpenRouter platform, and the request timeout is set to 60 seconds;

[0109] Path initialization:

[0110] Local test case repository path:

[0111] "C:\Users\18274\OneDrive\Desktop\CFD_Case\Incompressible";

[0112] Path to save the modified example file:

[0113] "C:\Users\18274\OneDrive\Desktop\CFD_Case\Processed\Incompressible\Cylinder_Flow";

[0114] Solution result output path:

[0115] The file is located at “C:\Users\18274\OneDrive\Desktop\CFD_Case\Result\Incompressible\Cylinder_Flow” and is stored in categories such as “Residual Files”, “Flow Field Data Files”, and “Convergence Reports”.

[0116] (2) Parametric space diagram and artificial intelligence agent decomposition of fluid mechanics problems

[0117] Parameter space import:

[0118] Add a "Form Input" node to the automated workflow platform and design a dedicated input interface for fluid mechanics problems. The association logic between the fields and the governing equations is shown in Table 1 below, ensuring that the input parameters directly serve the subsequent equation identification and configuration.

[0119] Table 1;

[0120] ;

[0121] Once the above parameters are entered, the node automatically converts the parameter space into a JSON format data stream, as follows:

[0122] json

[0123] {

[0124] "problem_type":"Flow around an external cylinder in two dimensions",

[0125] "fluid_type":"Incompressible air",

[0126] "governing_equation":"Incompressible Navier-Stokes equations",

[0127] "physical_parameters":{"velocity_inlet":10,"density":1.225,"viscosity":1.789e-5},

[0128] "solution_requirement":{"residual_criterion":1e-6,"output_data":"velocity distribution"}

[0129] }

[0130] AI Agent Parameter Space Decomposition

[0131] The AI ​​agent receives the above JSON data stream and breaks down the parameter space into four executable steps strongly associated with the incompressible Navier-Stokes equations, ensuring that each step serves the equation solution requirements:

[0132] json

[0133] {

[0134] "step1":"Governing equation identification and matching: Confirm that the target equation is an incompressible Navier-Stokes equation, and filter .stt / .inp files whose names contain "incompressible" or "2D_Cylinder".

[0135] "step2":"Mesh parameter configuration: Based on the requirements of the incompressible flow boundary layer, the cylindrical surface mesh minimum_cellsize ≤ 0.005m, and the far-field maximum_cellsize ≥ 0.1m, to improve the stability of the equation discretization."

[0136] Step 3: Boundary conditions and numerical parameter settings: velocity inlet (10 m / s), pressure outlet (0 Pa). The SIMPLE algorithm is used to solve the pressure-velocity coupling problem, with ≥1000 iterations.

[0137] Step 4: "Solution verification rules: Monitor the residuals of the continuity equation and the momentum equations in the x / y directions. Both must be reduced to below 1e-6 and stabilized for 10 consecutive steps to ensure the validity of the equation solution."

[0138] }

[0139] Decomposition Logic: By recognizing the "governing_equation" field, the AI ​​agent automatically matches the solution logic for incompressible Navier-Stokes equations.

[0140] (3) Identification of governing equations and batch matching of example files (dedicated filtering)

[0141] Add a "custom code node (function1)" to the automated workflow platform to receive the decomposition results output by the AI ​​agent, extract "control equation type = incompressible NS equation", and generate targeted filtering rules:

[0142] File type filtering: Only keep .stt mesh files and .inp solver input files.

[0143] Name tag filtering: The file name must contain "incompressible" or "2D_Cylinder".

[0144] Mesh parameter pre-verification: Read the "minimal_cellsize" field in the .stt file and keep only files ≤0.01m (initially meeting the requirements for incompressible flow-around boundary layer refinement, which will be further optimized later).

[0145] Batch reading of example files and path processing

[0146] Add a "File Reading" node, configure the above filtering rules, scan the local test case repository "C:\Users\18274\OneDrive\Desktop\CFD_Case\Incompressible", enable batch read mode, and obtain all files that meet the conditions at once.

[0147] Write a JS script to handle file paths using a "custom code node (function1)":

[0148] Path extraction: Detect the input data field, convert it to a string, split it by newline character, filter out empty lines and invalid paths, and obtain the valid absolute file path;

[0149] Standardization process: Double quotes at the beginning and end of the path are removed by regular expression matching, backslashes are replaced with forward slashes, whitespace characters are removed, and finally the output is in the format of "Item + key-value pair" (e.g., [Item0{"Path":"C:\Users\18274\OneDrive\Desktop\CFD_Case\Incompressible\Cylinder_2D.incompressible.stt"},]), ensuring that each case corresponds to a unique ID and path;

[0150] Add a "LoopOverItems" node to execute recursive functionality, traverse all key-value pairs of items, extract information such as "example file name" and "example path" using the built-in command {{$json.source_dir}}, store it in the form of data (as shown in Table 2 below), and extract text information from the data using the "ExtractfromFile" node, set the input binary field to "data" and the target output field to "text", and pass it to the artificial intelligence agent node.

[0151] Table 2;

[0152] ;

[0153] (4) AI agent’s dedicated parameter configuration for control equations

[0154] Specific prompts for constructing control equations:

[0155] In the "AI Agent" node of the n8n workflow platform, based on the decomposition results of step 2 and the file pre-extraction information of step 3, parameter modification prompts for incompressible Navier-Stokes equations are constructed to ensure that each modification matches the equation solution requirements: {{$json.data}}

[0156] Given that the target fluid dynamics governing equations are the "incompressible Navier-Stokes equations (continuity + momentum)," and the example file to be processed is a two-dimensional flow around an external cylinder (.stt mesh file and .inp solver input file), modify the parameters according to the following rules. Only the complete modified file content will be returned without explanation:

[0157] 1. Modify the .stt file (to adapt to the requirements of incompressible flow around boundary layer calculation):

[0158] Under -X_DIRECTION and Y_DIRECTION, "minimal_cellsize" is adjusted to 0.005m and "maximal_cellsize" is adjusted to 0.1m;

[0159] - Increase the number of grid cells "cornerpoint_number" on the cylindrical surface by 20;

[0160] - Keep the mesh orthogonality parameter "orthogonality_ratio" unchanged;

[0161] 2. Modify the .inp file (to match the solution logic for incompressible Navier-Stokes equations):

[0162] - The "EQUATION_TYPE" field is confirmed to be "INCOMPRESSIBLE", and "ENERGY_EQUATION:OFF" is added;

[0163] - The "NUMERICAL_METHOD" field is set to "FINITE_VOLUME_SIMPLE";

[0164] - "BOUNDARY_CONDITION" field configuration:

[0165] -Entrance boundary (INLET): "VELOCITY:1000";

[0166] -Outlet boundary: "PRESSURE:0";

[0167] - Cylindrical surface (WALL): "NO_SLIP";

[0168] - "SOLVER_SETTING" field settings:

[0169] -"ITERATION_STEPS:1500";

[0170] -"RESIDUAL_CRITERION:1e-6";

[0171] -"VISCOSITY:1.789e-5".

[0172] Among them, {{$json.data}} is a built-in command of the node, which is used to reference the transformed example text data in data to ensure that the prompt words are accurately associated with the file to be processed.

[0173] Data preprocessing and large language model LLM call:

[0174] Add a "custom code node (function2)" to split the batch of case data into a single processing unit (such as the current processing of the "Cylinder_2D" case), and add a unique identifier: case_id:Cylinder_2D_Incompressible_20240520 and a control equation type label "equation_type:incompressible_NS" to facilitate the matching and traceability of subsequent solution results with equation types;

[0175] The JSON.parse function parses the input string data, generates a structured data object, adds "caseId" and "equation_type" attributes and assigns values, and returns the processed object;

[0176] The AI ​​agent node automatically combines the preprocessed example data with the above prompts and sends it to the target LLM interface via a POST request. An example request format is as follows:

[0177] json

[0178] {

[0179] "model":"deepseek / deepseek-r1-0528-qwen3-8b",

[0180] "messages":[

[0181] {"role":"system","content":"You are the expert in configuring parameters for the fluid dynamics control equations. You must strictly modify the example files according to the solution requirements of the incompressible Navier-Stokes equations to ensure that the parameters completely match the equation assumptions and physical scenarios."}

[0182] {"role":"user","content":{"prompt":"[the above-mentioned specific prompt word]","file_data":"[the original text count of the Cylinder_2D.incompressible.stt and .inp files]"}}

[0183] ],

[0184] "timeout":60

[0185] }

[0186] Results cleaning and data validation:

[0187] Receive the response results returned by the Large Language Model (LLM). An example response is shown below:

[0188] json

[0189] [Item0{

[0190] "result":{

[0191] "stt_data":"[Modified complete .stt file content, minimal_cellsize=0.005m, cornerpoint_number increased by 20]",

[0192] "inp_data":"[The modified complete .inp file content, ENERGY_EQUATION:OFF, ITERATION_STEPS=1500]"

[0193] },

[0194] Note: "Parameters have been modified according to the requirements of the incompressible Navier-Stokes equations."

[0195] }]

[0196] Cleaning results after adding "Custom code node (function3)":

[0197] Text extraction: Extract the "stt_data" and "inp_data" fields from the structured data and assign them to temporary variables;

[0198] Invalid information filtering: Match and remove descriptive text (such as the content of the "note" field) using regular expressions (such as / ^note: / gm), retaining the plain file data.

[0199] Fluid dynamics parameter verification:

[0200] Check if the .inp file contains "ENERGY_EQUATION:OFF" and "EQUATION_TYPE:INCOMPRESSIBLE" to ensure it matches the incompressible NS equations;

[0201] Check if the "minimal_cellsize" value in the .stt file is ≤ 0.005m and if the "cornerpoint_number" value is increased as required to ensure that the grid parameters meet the discretization requirements of the equations.

[0202] If the verification passes, a standardized data stream is generated; if it fails, the process is returned to the AI ​​agent for modification until the control equations are met.

[0203] (5) Example file format conversion and local saving

[0204] Format conversion for governing equation adaptation:

[0205] Add a "ConverttoFile" node to receive the standardized data stream output by the AI ​​agent and convert it according to the format requirements of the VirtualFlow software (whose core solver is transatmb):

[0206] .stt file: Writes the modified grid parameters (such as "minimal_cellsize" and "cornerpoint_number") into specified modules such as "GRIDPROPERTIES", "X_DIRECTION", and "Y_DIRECTION".

[0207] .inp files: Arrange numerical parameters and boundary conditions according to the transatmb command format. For example, the "BOUNDARY_CONDITION" module should be written in the order of "boundary name-type-parameter" to avoid solver parsing errors.

[0208] Saved by category according to governing equations:

[0209] Add a "File Write" node and configure the save path to "C:\Users\18274\OneDrive\Desktop\CFD_Case\Processed\Incompressible\Cylinder_Flow" set in step (1) above;

[0210] The file naming rule is set to "case_id_equation type_modification time.file extension". In this example, the generated file is:

[0211] The files “Cylinder_2D_Incompressible_20240520_incompressible_NS.stt” and “Cylinder_2D_Incompressible_20240520_incompressible_NS.inp” are used to ensure that the files are strongly associated with the control equation type and the example ID, thus avoiding overwriting.

[0212] The node automatically generates a save log, recording the file save path, file name, modification information, and save time.

[0213] (6) Solver call and solution verification of the governing equations

[0214] Configuration of environment variables specific to the control equations:

[0215] Add a "Command Line Execution" node (cmd1) and set environment variables related to the VirtualFlow software to ensure that the solver can recognize the solution dependencies of the incompressible Navier-Stokes equations:

[0216] cmd

[0217] setVIRTUALFLOW_HOME=D:\VirtualFlow\v3.0

[0218] The solver installation directory contains transatmbDP.exe (the solver program) and VirtualFlowCmd.exe (the command-line application).

[0219] setPATH=%PATH%;%VIRTUALFLOW_HOME%\bin

[0220] Add the solver executable directory to the system PATH to ensure it can be accessed from any directory. Dedicated solver instruction construction and execution:

[0221] Add a "Command Line Execution" node (cmd2), extract the full path of the modified .inp file ("C:\Users\18274\OneDrive\Desktop\CFD_Case\Processed\Incompressible\Cylinder_Flow\Cylinder_2D_Incompressible_20240520_incompressible_NS.inp") from the save log in step 5, and construct the transatmb solver call instructions based on the solution requirements of the incompressible NS equations:

[0222] cmd

[0223] "D:\VirtualFlow\v3.0\bin\VirtualFlowCmd.exe"^

[0224] -log"C:\Users\18274\OneDrive\Desktop\CFD_Case\Result\Incompressible\Cylinder_Flow\convergence.log"^

[0225] Output residual convergence logs for subsequent verification.

[0226] -PArgs="-n2"^

[0227] The solver core count is set to 2 to accommodate incompressible streaming parallel computing requirements, balancing efficiency and resource consumption.

[0228] -PFile"C:\Users\18274\OneDrive\Desktop\CFD_Case\Processed\Incompressible\Cylinder_Flow\Cylinder_2D_Incompressible_20240520_incompressible_NS.inp"^Input file path.

[0229] Solution verification and result processing:

[0230] The node executes the above instructions, starts the transatmb solver, and captures console output and log files in real time.

[0231] Add a "Custom Code Node (function4)" to write the fluid dynamics solution verification logic:

[0232] Residual extraction: Extract residual data of the continuity equation and the momentum equations (Momentum-X, Momentum-Y) in the x and y directions from the log;

[0233] Convergence determination: When the residuals of the three equations are all ≤1e-6 and remain stable for 10 consecutive iterations, the solution is determined to be "successful" and "Solution converged successfully" is output; otherwise, return to step four to reconfigure.

[0234] After successful solution, the result files are stored in "C:\Users\18274\OneDrive\Desktop\CFD_Case\Result\Incompressible\Cylinder_Flow" according to the "Governing Equation Type + Problem Type". The AI ​​agent generates an "Incompressible Navier-Stokes Equation Case Processing Report", which includes parameter modification records, solution convergence curves, etc.

[0235] If there are batch case files, the "LoopOverItems" node repeats the above steps (3)-(6) in the order of "case_id" until all incompressible stream case files are processed, thus achieving a closed loop for the entire process.

[0236] Please see Figure 2 As shown, an embodiment of this application provides a full-process data processing system for fluid dynamics control equations. This full-process data processing system for fluid dynamics control equations includes:

[0237] The initialization module is used to initialize the paths and artificial intelligence agents related to the solution of fluid dynamics control equations on the local server, thereby importing and decomposing the parameter space of fluid dynamics problems.

[0238] The example file processing module is used to identify the control equations corresponding to fluid mechanics problems and batch match example files through artificial intelligence agents, and extract the core parameters related to the control equations.

[0239] The parameter configuration module is used to configure the parameters of the control equations through an artificial intelligence agent;

[0240] The file conversion and saving module is used to convert the format of the example files and save them locally to the corresponding path.

[0241] The solution and verification module is used to call the solver of the governing equations to perform the solution process and verify the results to determine the validity of the results.

[0242] The beneficial effects of the above embodiments are that the full-process data processing system for fluid dynamics control equations improves the efficiency of example processing and the accuracy of solution by constructing a dedicated link for solving fluid dynamics control equations, realizing an automated full-process link from problem parameter decomposition, control equation identification, dedicated parameter configuration, format conversion, and disassembly verification.

[0243] In another embodiment, the initialization module is used to initialize the paths and AI agents related to the solution of the fluid dynamics governing equations on a local server, thereby importing and decomposing the parameter space of the fluid dynamics problem, including:

[0244] Deploy an automated workflow platform on a local server, integrate an AI agent into the automated workflow platform, and configure the connection parameters between the AI ​​agent and the large language model.

[0245] Based on the automated workflow platform, initialize the paths related to solving the fluid dynamics control equations; where the paths match the nodes of the automated workflow platform.

[0246] The parameter space of the fluid mechanics problem to be solved is imported through the nodes of the automated workflow platform. The artificial intelligence agent decomposes the parameter space into several executable steps based on the preset fluid mechanics knowledge graph. These executable steps include the selection of governing equations, configuration of mesh parameters, setting of boundary conditions, and definition of solution parameters.

[0247] In another embodiment, the case study file processing module is used to identify the governing equations corresponding to the fluid mechanics problem and batch match case study files through an artificial intelligence agent, extracting core parameters related to the governing equations, including:

[0248] The AI ​​agent identifies the type of control equations for fluid mechanics problems. Based on the control equation type, the automated workflow platform uses a file filtering node to batch match the corresponding case files in the local case repository and extracts the core parameters related to the control equations from the case files.

[0249] The parameter configuration module is used to configure the parameters of the control equations through an artificial intelligence agent, including:

[0250] The AI ​​agent automatically modifies the example files according to the solution requirements of the control equations; adjusts the mesh parameters according to the gradient requirements of the control equations; and configures the numerical parameters according to the type of control equation.

[0251] The file conversion and saving module is used to convert the format of the example files and save them locally to the corresponding path, including:

[0252] The configured numerical parameters are converted into a format supported by the fluid dynamics solution software through the file processing node of the automated workflow platform, and saved to the corresponding path according to the type of governing equation.

[0253] In another embodiment, adjusting the mesh parameters according to the gradient requirements of the governing equation includes:

[0254] Retrieve the type of control equations corresponding to the fluid mechanics problem identified by the artificial intelligence agent;

[0255] Based on the type of governing equations corresponding to the fluid mechanics problem, the gradient source terms corresponding to the type of governing equations are analyzed, where the gradient source terms refer to the physical quantity parameter terms that affect the gradient change;

[0256] Normalize the physical quantity parameter terms corresponding to the gradient source terms of the control equation type to obtain the normalized physical quantity parameter terms.

[0257] The normalized physical quantity parameter terms are used to form a gradient correlation vector corresponding to the type of the governing equation;

[0258] The gradient potential field Φ(x) is obtained using the gradient correlation vector corresponding to the type of the governing equation; wherein, the gradient potential field Φ(x) is obtained by the following formula: Where Φ(x) represents the gradient potential field; S(x) represents the relative intensity of the gradient source term, and S(x) = ║X║2×(X0+║X║2) -1 ║X║2 represents the L2 norm of the gradient correlation vector; X0 represents a reference value with the same dimensions as the L2 norm of the gradient correlation vector; n represents the number of vector elements contained in the gradient correlation vector; λ i X represents the weight value corresponding to the i-th vector element; iLet G(x) represent the numerical value corresponding to the i-th vector element; G(x) represents the enhancement factor, which is used to enhance the grid configuration strength of the gradient vector in the key region, and the enhancement factor G(x) is obtained by the following formula: Where α represents the strengthening intensity coefficient, with a value ranging from 0.5 to 1.3; β represents the stability period, with a value ranging from 1 to 1.14; F(x) represents the divergence corresponding to the gradient correlation vector field; and F0 represents the preset divergence reference value.

[0259] The current mesh size is adjusted using the gradient potential field Φ(x), and the adjusted gradient potential field Φ(x) is obtained using the following formula:

[0260] Where Г represents the adjusted mesh size, η represents the strength adjustment coefficient, and its value ranges from 0.7 to 1.5; Δ base This indicates the grid size before adjustment.

[0261] The beneficial effects of the above embodiments are achieved through an intelligent pre-configuration method that integrates physical knowledge and mathematical models. This generates an optimized mesh highly suited to specific flow physics and user needs from the very beginning of the computation, thus achieving a synergistic leap in efficiency, accuracy, and automation from the source. This overcomes the inherent limitations of traditional mesh adaptation techniques that rely on solver feedback and human experience. Simultaneously, by using AI to identify the type of control equations and analyze gradient source terms, it can selectively extract physical quantities affecting gradient changes (such as velocity gradients and pressure gradients), avoiding the insufficient adaptability of "general-purpose meshes" to different fluid problems, and allowing mesh adjustments to better fit the physical essence of specific problems. The L2 norm and divergence of gradient correlation vectors significantly increase in "critical regions where gradient changes are drastic." Furthermore, combined with the enhancing effect of the strengthening factor, the gradient potential field increases in these regions. Ultimately, the mesh size is reduced through the mesh adjustment formula, improving the computational resolution in critical regions and reducing numerical errors. Moreover, the mesh parameters are directly mapped and generated by analytical formulas, avoiding repeated trial-and-error based on error estimation or gradient threshold determination and global / local mesh reconstruction operations, significantly reducing computational resource consumption purely for mesh adaptation. Through mechanisms such as "gradient potential field" and "local aggregation degree", it is possible to proactively identify and enhance the encryption of key regions with drastic changes in physical quantities, such as boundary layer, shock wave, and shear layer, without relying on any instantaneous solution, thereby improving the ability to distinguish complex flow phenomena.

[0262] In another embodiment, the solution and verification module is used to call the solver of the governing equations to perform solution processing and verify the results of the solution processing to determine the validity of the results, including:

[0263] By using the command-line execution node of the automated workflow platform, a dedicated solution instruction is constructed based on the type of control equation, which is then used to call the control equation solver for solution processing.

[0264] The validity of the solution processing results is judged by real-time monitoring of the solution processing logs and the convergence characteristics of the control equations based on the solution processing logs and the convergence characteristics of the control equations.

[0265] The operation and effect of the full-process data processing system for fluid dynamics control equations of the present invention are consistent with the above-mentioned full-process data processing method for fluid dynamics control equations, and the full-process data processing system for fluid dynamics control equations will not be described again here.

[0266] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. This disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A full-process data processing method for fluid dynamics control equations, characterized in that, include: Initialize the paths and AI agents related to the solution of fluid dynamics control equations on the local server to import and decompose the parameter space of the fluid dynamics problem. The AI ​​agent identifies the control equations and batch matching example files corresponding to the fluid mechanics problem, and extracts the core parameters related to the control equations. The AI ​​agent configures the parameters of the control equations; converts the format of the example file and saves it locally to the corresponding path; The solver of the governing equations is invoked to perform the solution process, and the results of the solution process are verified to determine the validity of the results. This includes initializing the paths and AI agents related to solving the fluid dynamics governing equations on the local server, thereby importing and decomposing the parameter space of the fluid dynamics problem, including: Deploy an automated workflow platform on a local server, and integrate an artificial intelligence agent into the automated workflow platform and configure the connection parameters between the artificial intelligence agent and the large language model; Based on the automated workflow platform, initialize the path related to solving the fluid dynamics control equations; wherein the path matches the nodes of the automated workflow platform; The parameter space of the fluid mechanics problem to be solved is imported through the nodes of the automated workflow platform. The artificial intelligence agent decomposes the parameter space into several executable steps according to the preset fluid mechanics knowledge graph. The several executable steps include the selection of control equations, configuration of mesh parameters, setting of boundary conditions, and definition of solution parameters. The adjustment of mesh parameters according to the gradient requirements of the governing equations includes: Retrieve the type of control equations corresponding to the fluid mechanics problem identified by the artificial intelligence agent; Based on the type of governing equations corresponding to the fluid mechanics problem, the gradient source terms corresponding to the type of governing equations are analyzed, where the gradient source terms refer to the physical quantity parameter terms that affect the gradient change; Normalize the physical quantity parameter terms corresponding to the gradient source terms of the control equation type to obtain the normalized physical quantity parameter terms. The normalized physical quantity parameter terms are used to form a gradient correlation vector corresponding to the type of the governing equation; The gradient potential field Φ(x) is obtained using the gradient correlation vector corresponding to the type of the governing equation; wherein, the gradient potential field Φ(x) is obtained by the following formula: ; Where Φ(x) represents the gradient potential field; S(x) represents the relative intensity of the gradient source term, and S(x) = ║X║2×(X0+║X║2) -1 ║X║2 represents the L2 norm of the gradient correlation vector; X0 represents a reference value with the same dimensions as the L2 norm of the gradient correlation vector; n represents the number of vector elements contained in the gradient correlation vector; λ i X represents the weight value corresponding to the i-th vector element; i Let G(x) represent the numerical value corresponding to the i-th vector element; G(x) represents the enhancement factor, which is used to enhance the grid configuration strength of the gradient vector in the key region, and the enhancement factor G(x) is obtained by the following formula: ； Where α represents the strengthening intensity coefficient, with a value ranging from 0.5 to 1.3; β represents the stability period, with a value ranging from 1 to 1.14; F(x) represents the divergence corresponding to the gradient correlation vector field; and F0 represents the preset divergence reference value. The current mesh size is adjusted using the gradient potential field Φ(x), and the adjusted gradient potential field Φ(x) is obtained using the following formula: ； Where Г represents the adjusted mesh size, η represents the strength adjustment coefficient, and its value ranges from 0.7 to 1.5; Δ base This indicates the grid size before adjustment.

2. The full-process data processing method for fluid dynamics control equations as described in claim 1, characterized in that: The AI ​​agent identifies the governing equations and batch-matched example files corresponding to the fluid mechanics problem, and extracts the core parameters related to the governing equations, including: The AI ​​agent identifies the type of control equations corresponding to the fluid mechanics problem. Based on the type of control equations, the AI ​​workflow platform uses a file filtering node to batch match the corresponding case files in the local case repository and extracts the core parameters related to the control equations from the case files. The AI ​​agent configures the parameters of the control equations; the example file is converted to a different format and saved locally to the corresponding path, including: The AI ​​agent automatically modifies the example file according to the solution requirements of the control equations; adjusts the mesh parameters according to the gradient requirements of the control equations; and configures numerical parameters according to the type of control equation. The configured numerical parameters are converted into a format supported by the fluid dynamics solution software through the file processing node of the automated workflow platform, and saved to the corresponding path according to the type of the governing equation.

3. The full-process data processing method for fluid dynamics control equations as described in claim 2, characterized in that: The solver of the governing equations is invoked for solution processing, and the results of the solution processing are verified to determine the validity of the results, including: Through the command line execution node of the automated workflow platform, a dedicated solution instruction is constructed according to the type of the control equation, thereby calling the control equation solver for solution processing; The AI ​​agent monitors the solution processing log in real time, and judges the validity of the solution processing result based on the solution processing log and the convergence characteristics of the control equation.

4. A full-process data processing system for fluid dynamics control equations, characterized in that: include: The initialization module is used to initialize the paths and artificial intelligence agents related to the solution of fluid dynamics control equations on the local server, thereby importing and decomposing the parameter space of fluid dynamics problems. The example file processing module is used to identify the control equations corresponding to the fluid mechanics problem and batch match example files through the artificial intelligence agent, and extract the core parameters related to the control equations; The parameter configuration module is used to configure the parameters of the control equations through the artificial intelligence agent; The file conversion and saving module is used to convert the format of the example file and save it locally to the corresponding path. The solution and verification module is used to call the solver of the governing equations to perform solution processing and verify the results of the solution processing to determine the validity of the results. The initialization module is used to initialize the paths and AI agents related to the solution of fluid dynamics control equations on the local server, thereby importing and decomposing the parameter space of the fluid dynamics problem, including: Deploy an automated workflow platform on a local server, and integrate an artificial intelligence agent into the automated workflow platform and configure the connection parameters between the artificial intelligence agent and the large language model; Based on the automated workflow platform, initialize the path related to solving the fluid dynamics control equations; wherein the path matches the nodes of the automated workflow platform; The parameter space of the fluid mechanics problem to be solved is imported through the nodes of the automated workflow platform. The artificial intelligence agent decomposes the parameter space into several executable steps according to the preset fluid mechanics knowledge graph. The several executable steps include the selection of control equations, configuration of mesh parameters, setting of boundary conditions, and definition of solution parameters. The adjustment of mesh parameters according to the gradient requirements of the governing equations includes: Retrieve the type of control equations corresponding to the fluid mechanics problem identified by the artificial intelligence agent; Based on the type of governing equations corresponding to the fluid mechanics problem, the gradient source terms corresponding to the type of governing equations are analyzed, where the gradient source terms refer to the physical quantity parameter terms that affect the gradient change; Normalize the physical quantity parameter terms corresponding to the gradient source terms of the control equation type to obtain the normalized physical quantity parameter terms. The normalized physical quantity parameter terms are used to form a gradient correlation vector corresponding to the type of the governing equation; The gradient potential field Φ(x) is obtained using the gradient correlation vector corresponding to the type of the governing equation; wherein, the gradient potential field Φ(x) is obtained by the following formula: ; Where Φ(x) represents the gradient potential field; S(x) represents the relative intensity of the gradient source term, and S(x) = ║X║2×(X0+║X║2) -1 ║X║2 represents the L2 norm of the gradient correlation vector; X0 represents a reference value with the same dimensions as the L2 norm of the gradient correlation vector; n represents the number of vector elements contained in the gradient correlation vector; λ i X represents the weight value corresponding to the i-th vector element; i Let G(x) represent the numerical value corresponding to the i-th vector element; G(x) represents the enhancement factor, which is used to enhance the grid configuration strength of the gradient vector in the key region, and the enhancement factor G(x) is obtained by the following formula: ； Where α represents the strengthening intensity coefficient, with a value ranging from 0.5 to 1.3; β represents the stability period, with a value ranging from 1 to 1.14; F(x) represents the divergence corresponding to the gradient correlation vector field; and F0 represents the preset divergence reference value. The current mesh size is adjusted using the gradient potential field Φ(x), and the adjusted gradient potential field Φ(x) is obtained using the following formula: ； Where Г represents the adjusted mesh size, η represents the strength adjustment coefficient, and its value ranges from 0.7 to 1.5; Δ base This indicates the grid size before adjustment.

5. The full-process data processing system for fluid dynamics control equations as described in claim 4, characterized in that: The example file processing module is used to identify the governing equations corresponding to the fluid mechanics problem and batch match example files through the artificial intelligence agent, and extract the core parameters related to the governing equations, including: The AI ​​agent identifies the type of control equations corresponding to the fluid mechanics problem. Based on the type of control equations, the AI ​​workflow platform uses a file filtering node to batch match the corresponding case files in the local case repository and extracts the core parameters related to the control equations from the case files. The parameter configuration module is used to configure the parameters of the control equations through the artificial intelligence agent, including: The AI ​​agent automatically modifies the example file according to the solution requirements of the control equations; adjusts the mesh parameters according to the gradient requirements of the control equations; and configures numerical parameters according to the type of control equation. The file conversion and saving module is used to convert the format of the example file and save it locally to the corresponding path, including: The configured numerical parameters are converted into a format supported by the fluid dynamics solution software through the file processing node of the automated workflow platform, and saved to the corresponding path according to the type of the governing equation.

6. The full-process data processing system for fluid dynamics control equations as described in claim 5, characterized in that: The solution and verification module is used to call the solver of the governing equations to perform solution processing and to verify the results of the solution processing, thereby determining the validity of the results, including: Through the command line execution node of the automated workflow platform, a dedicated solution instruction is constructed according to the type of the control equation, thereby calling the control equation solver for solution processing; The AI ​​agent monitors the solution processing log in real time, and judges the validity of the solution processing result based on the solution processing log and the convergence characteristics of the control equation.

Images