Turbulence model generation method, turbulence simulation method, device, intelligent agent and equipment

Abstract

The disclosure provides a turbulent flow model generation method, a turbulent flow simulation method, a device, an intelligent agent and equipment, relates to the field of artificial intelligence, and in particular to the fields of large models, deep learning, intelligent agents and fluid dynamics. The specific implementation scheme is: inputting a model generation prompt word into a large model to generate a first turbulent flow model; simulating the first turbulent flow model based on a preset example to obtain simulation flow field data generated by the first turbulent flow model in simulating turbulent flow phenomena in a flow scene corresponding to the preset example; in response to a deviation between the simulation flow field data and real flow field data in the preset example being greater than a predetermined threshold, determining a reference turbulent flow model from a plurality of historical turbulent flow models generated by the large model, wherein the reference turbulent flow model is different from the model structure of the first turbulent flow model; inputting feedback information generated based on the deviation and the reference turbulent flow model into the large model to generate a second turbulent flow model.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence, and more particularly to the fields of large models, deep learning, intelligent agents, and fluid dynamics, and more particularly to a method for generating turbulence models, a method for turbulence simulation, an apparatus, an intelligent agent, and a device. Background Technology

[0002] With the development of computational fluid dynamics and artificial intelligence technologies, users can generate preliminary turbulence models simply by inputting their requirements. However, when faced with new flow scenarios or when model predictions deviate, it is difficult to correct or regenerate the model based on specific physical mechanisms and error characteristics. Summary of the Invention

[0003] This disclosure provides a method for generating turbulence models, a method for turbulence simulation, an apparatus, an intelligent agent, a device, a medium, and a product.

[0004] According to one aspect of this disclosure, a method for generating a turbulence model is provided, comprising: inputting model generation prompts into a large model to generate a first turbulence model; simulating the first turbulence model based on a preset example to obtain simulated flow field data of turbulence phenomena generated by the first turbulence model under the flow scenario corresponding to the preset example; in response to a deviation between the simulated flow field data and the actual flow field data in the preset example exceeding a predetermined threshold, determining a reference turbulence model from multiple historical turbulence models generated from the large model, wherein the reference turbulence model has a different model structure than the first turbulence model; and inputting feedback information generated based on the deviation and the reference turbulence model into the large model to generate a second turbulence model.

[0005] According to another aspect of this disclosure, a turbulence simulation method is provided, comprising: inputting target operating condition parameters into a target turbulence model and outputting target flow field data, wherein the target flow field data characterizes the turbulence features under the flow scenario indicated by the target operating condition parameters; wherein the target turbulence model is generated by the above method.

[0006] According to another aspect of this disclosure, a turbulence model generation apparatus is provided, comprising: a first generation module, configured to input model generation prompts into a large model to generate a first turbulence model; a model simulation module, configured to simulate the first turbulence model based on a preset example to obtain simulated flow field data of the first turbulence model simulating turbulence phenomena under the flow scenario corresponding to the preset example; a reference determination module, configured to determine a reference turbulence model from multiple historical turbulence models generated from the large model in response to a deviation between the simulated flow field data and the real flow field data in the preset example exceeding a predetermined threshold, wherein the reference turbulence model has a different model structure than the first turbulence model; and a second generation module, configured to input feedback information generated based on the deviation and the reference turbulence model into the large model to generate a second turbulence model.

[0007] According to another aspect of this disclosure, a turbulence simulation apparatus is provided, comprising: a turbulence simulation module for inputting target operating condition parameters into a target turbulence model and outputting target flow field data, wherein the target flow field data characterizes the turbulence features under the flow scenario indicated by the target operating condition parameters; wherein the target turbulence model is generated by the above method.

[0008] According to another aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described above.

[0009] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause a computer to perform the methods described above.

[0010] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the method described above.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0012] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0013] Figure 1 This illustration schematically shows an exemplary system architecture to which turbulence model generation methods, turbulence simulation methods, and apparatus can be applied according to embodiments of the present disclosure;

[0014] Figure 2 A flowchart illustrating a turbulence model generation method according to an embodiment of the present disclosure is shown schematically.

[0015] Figure 3 This diagram illustrates the architectural interaction of a turbulence model generation method according to an embodiment of the present disclosure.

[0016] Figure 4 This illustration schematically shows a model library update diagram of a turbulence model generation method according to an embodiment of the present disclosure;

[0017] Figure 5 A schematic diagram illustrating a system architecture of a turbulence model generation method according to another embodiment of the present disclosure is shown.

[0018] Figure 6A flowchart illustrating a turbulence simulation method according to an embodiment of the present disclosure is shown schematically;

[0019] Figure 7 The diagram illustrates an interactive schematic of a turbulence simulation method according to an embodiment of the present disclosure;

[0020] Figure 8 A block diagram of a turbulence model generation apparatus according to an embodiment of the present disclosure is shown schematically;

[0021] Figure 9 A block diagram of a turbulence simulation apparatus according to an embodiment of the present disclosure is shown schematically;

[0022] Figure 10 A schematic diagram illustrating the structure of an intelligent agent of artificial intelligence according to embodiments of the present disclosure is shown.

[0023] Figure 11 A block diagram of an electronic device suitable for implementing a turbulence model generation method and a turbulence simulation method according to embodiments of the present disclosure is shown schematically. Detailed Implementation

[0024] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0025] Figure 1 The illustration schematically depicts an exemplary system architecture to which turbulence model generation methods, turbulence simulation methods, and apparatus can be applied according to embodiments of the present disclosure.

[0026] It is important to note that Figure 1 The examples shown are merely examples of system architectures that can be applied to the embodiments of this disclosure, to help those skilled in the art understand the technical content of this disclosure, but do not mean that the embodiments of this disclosure cannot be used in other devices, systems, environments, or scenarios. For example, in another embodiment, an exemplary system architecture that can be applied to the turbulence model generation method, turbulence simulation method, and apparatus may include a terminal device, but the terminal device can implement the turbulence model generation method, turbulence simulation method, and apparatus provided by the embodiments of this disclosure without interacting with the server.

[0027] like Figure 1As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, and 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the terminal devices 101, 102, and 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0028] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (for example only).

[0029] Terminal devices 101, 102, and 103 can be various electronic devices with displays and web browsing capabilities, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0030] Server 105 can be a server that provides various services, such as a backend management server that supports the content browsed by users using terminal devices 101, 102, and 103 (for example only). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0031] It should be noted that the turbulence model generation method and turbulence simulation method provided in the embodiments of this disclosure can generally be executed by terminal devices 101, 102, or 103. Accordingly, the content processing device provided in the embodiments of this disclosure can also be disposed in terminal devices 101, 102, or 103.

[0032] Alternatively, the turbulence model generation method and turbulence simulation method provided in this disclosure embodiment can generally also be executed by server 105. Correspondingly, the content processing apparatus provided in this disclosure embodiment can generally be located in server 105. The turbulence model generation method and turbulence simulation method provided in this disclosure embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with terminal devices 101, 102, 103 and / or server 105. Correspondingly, the content processing apparatus provided in this disclosure embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with terminal devices 101, 102, 103 and / or server 105.

[0033] It should be understood that Figure 1The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0034] In the technical solution disclosed herein, the collection, storage, use, processing, transmission, provision, disclosure, and application of user personal information comply with the provisions of relevant laws and regulations, necessary confidentiality measures have been taken, and there is no violation of public order and good morals.

[0035] In the technical solution disclosed herein, the user's authorization or consent is obtained before acquiring or collecting the user's personal information.

[0036] Figure 2 A flowchart illustrating a method for generating a turbulence model according to an embodiment of the present disclosure is shown schematically.

[0037] like Figure 2 As shown, the method includes operations S210~S240.

[0038] When operating S210, input the model generation prompt into the large model to generate the first turbulence model.

[0039] In operation S220, the first turbulence model is simulated based on a preset example to obtain the simulated flow field data of the first turbulence model under the flow scenario corresponding to the preset example.

[0040] In operation S230, in response to the deviation between the simulated flow field data and the real flow field data in the preset example being greater than a predetermined threshold, a reference turbulence model is determined from multiple historical turbulence models generated by the large model, wherein the reference turbulence model has a different model structure than the first turbulence model.

[0041] In operation S240, the feedback information generated based on the deviation and the reference turbulence model are input into the large model to generate a second turbulence model.

[0042] According to embodiments of this disclosure, model generation prompts are constructed based on the turbulence model generation task. Specifically, information such as task description, preset physical constraints (e.g., positive definiteness, dimensional homogeneity, Galilean invariance, rotational invariance, triangular constraints), and baseline turbulence model are input into the large model to utilize the pattern recognition and knowledge transfer capabilities of the large model to generate a first turbulence model.

[0043] Subsequently, representative pre-defined examples are selected, which typically cover typical flow phenomena such as boundary layer separation, shock wave-boundary layer disturbance, or anisotropic turbulence. Numerical simulations are then performed on the first turbulence model based on these pre-defined examples, yielding the simulated flow field data predicted by the first turbulence model under specific flow scenarios. The simulation process can be solved using a Computational Fluid Dynamics (CFD) solver to obtain the distributions of key physical quantities such as velocity field, pressure field, and turbulent kinetic energy.

[0044] After obtaining the simulated flow field data, the deviation between the simulated flow field data and the real flow field data obtained through experiments or direct numerical simulation in the preset examples is determined. When the deviation exceeds a predetermined threshold, a reference turbulence model is selected from multiple historical turbulence models generated during the previous iterations of the large model. This reference turbulence model has a significantly different model structure from the current first turbulence model. This difference in model structure may be reflected in the mathematical form of the model equations, variable dependencies, or the expression of closing terms. The purpose is to introduce a completely new modeling approach and avoid getting trapped in local optima.

[0045] Simultaneously, based on the calculated deviations, structured feedback information is generated. This feedback information includes not only a quantitative description of the deviations, but also the spatial distribution characteristics of the deviations in the flow field and an analysis at the physical mechanism level, such as pointing out the systematic deviations of the first turbulence model when simulating turbulent dissipation rates.

[0046] The feedback information and the selected reference turbulence model are used as inputs and submitted to the large model again. The large model can integrate the model structure of the reference model and combine the correction direction in the feedback information to generate a more accurate second turbulence model. This second turbulence model usually shows better prediction ability in the verification of preset examples.

[0047] By inputting model generation prompts into a large model to generate a first turbulence model, and obtaining its simulated flow field data based on preset examples, a rapid assessment of the model's initial predictive capability was achieved, providing a quantifiable verification basis for subsequent iterative optimization. Analysis of the deviation between simulated and real flow field data, combined with the selection of reference turbulence models with different structures from historical turbulence models, provided a clear optimization direction and decision-making basis for guiding the large model to overcome its original modeling limitations. Feedback information and the reference turbulence model were input into the large model to generate a second turbulence model that integrates the advantages of multiple structures, achieving targeted correction and performance improvement of the first turbulence model's prediction deviations. This iterative model generation process effectively integrates the large model's content generation capabilities with deviation-based closed-loop feedback control, significantly improving the turbulence model's prediction accuracy in specific flow scenarios, development iteration efficiency, and adaptability to complex flow conditions, effectively saving computational resources and hardware costs.

[0048] According to an embodiment of this disclosure, a first turbulence model is simulated based on a preset calculation example to obtain simulated flow field data of the first turbulence model simulating turbulence phenomena under the flow scenario corresponding to the preset calculation example. This includes: compiling the mathematical expressions in the first turbulence model into simulation code; assigning values ​​to the corresponding variables in the simulation code according to the flow condition parameters in the preset calculation example to obtain target code; and running the target code to obtain simulated flow field data of the first turbulence model simulating turbulence phenomena under the flow scenario corresponding to the preset calculation example.

[0049] After obtaining the first turbulence model, a large-scale model is used to transform it into an executable simulation program to verify its predictive capabilities. In the specific implementation process, code generation techniques within the large-scale model can be used to compile the mathematical expressions in the first turbulence model, such as partial differential equations, turbulent viscosity calculation formulas, and various closure coefficients, into high-performance simulation code.

[0050] This process typically leverages symbolic computation and code template engines. Based on code examples from a baseline turbulence model, the mathematical expressions of the first turbulence model are directly mapped to low-level language code, generating interface modules adapted to CFD solvers. Automated compilation not only eliminates syntax errors and logical deviations that may be introduced by manual coding, but also improves code execution efficiency through compiler optimization, laying the foundation for subsequent large-scale parallel computing.

[0051] Subsequently, based on the flow parameters defined in the preset examples, such as inlet Mach number, Reynolds number, wall temperature, and boundary layer thickness, the corresponding variables in the generated simulation code are precisely assigned values ​​to form complete target code. This variable assignment ensures that the first turbulence model can perform customized simulations for specific engineering problems (such as airfoil flow around an aircraft, mixing in a gas turbine combustor, or the wake of a car rearview mirror), thereby accurately reflecting actual flow conditions.

[0052] Running the target code in a CFD solver performs numerical iterative solutions, yielding simulated flow field data predicted by the first turbulence model under a specified flow scenario. This data encompasses key physical quantities such as the velocity vector field, pressure distribution, turbulent kinetic energy, and its dissipation rate. By automating the simulation, the turbulence model can be rapidly validated and compared under various operating conditions, significantly shortening the model evaluation cycle.

[0053] Figure 3 The diagram illustrates the architectural interaction of a turbulence model generation method according to an embodiment of the present disclosure.

[0054] like Figure 3 As shown, the model generation prompts are input into the large model 301, which generates a first turbulence model and its target code based on the prompts. The target code is then sent to the simulator 302, where it runs under the preset flow parameters and outputs simulated flow field data. The simulated flow field data is sent to the deviation analysis module 303, which calculates the deviation from the actual flow field data and generates physically meaningful feedback information based on preset mapping rules.

[0055] Meanwhile, historical turbulence models stored in model library 304 according to their model structure are used to select a reference turbulence model, which has a different model structure from the first turbulence model. The reference turbulence model and feedback information are used together as input to the large model 301, which then drives the large model 301 to generate a more accurate second turbulence model.

[0056] Through the closed-loop iterative mechanism generated by the above model, adaptive optimization of turbulence model development is realized, which significantly improves the prediction accuracy of turbulence model for complex flow phenomena, while greatly shortening the research and development cycle, reducing dependence on large-scale computing resources, and effectively reducing hardware costs and time consumption.

[0057] According to embodiments of this disclosure, the turbulence model generation method further includes: in response to an error occurring in the running target code, inputting the acquired error information into a large model to correct the simulation code based on the error information to obtain corrected code, wherein the error information includes at least one of compilation error information and solution error information; and re-simulating based on the corrected code.

[0058] During the execution of the target code, if an error occurs, the relevant error information is automatically captured and parsed. This error information includes compilation errors, such as syntax errors, type mismatches, or linking errors, as well as solution errors, such as numerical divergence, matrix singularity, convergence failure, or library function call exceptions.

[0059] After obtaining the error message, it is structured and then input into the large model. Leveraging its deep understanding of programming language syntax, numerical computation methods, and the physical equations of turbulence models, the large model can accurately pinpoint the root cause of defects in the code and automatically generate correction schemes.

[0060] For example, when an error message indicates that a turbulent transport equation has experienced numerical overflow under specific boundary conditions, the large model can identify whether the problem stems from an improper time step setting, an unstable difference scheme, or an error in the source term processing. It then adjusts the code logic or numerical parameters accordingly to generate corrected code. The simulation process is then restarted based on this corrected code, automatically completing the compilation and solution process.

[0061] This mechanism achieves a fully automated closed loop from error reporting to repair, reducing reliance on manual debugging and significantly improving the efficiency of model verification and iteration. Simultaneously, by diagnosing and repairing complex errors through large models, it reduces the unnecessary consumption of computational resources caused by code errors, thereby effectively reducing hardware resource consumption and time costs while ensuring the continuity of simulation tasks.

[0062] According to embodiments of this disclosure, determining a reference turbulence model from multiple historical turbulence models generated from a large model includes: determining a target library from multiple model libraries classified and stored according to model structure, wherein each model library stores historical turbulence models with the same model structure, and the model structure corresponding to the target library is different from the model structure of the first turbulence model; and determining a reference turbulence model from multiple historical turbulence models based on the deviations of each historical turbulence model in the target library.

[0063] To select suitable reference turbulence models from multiple historical turbulence models generated by a large model, all generated historical turbulence models are classified and stored according to the mathematical characteristics of their structures, constructing multiple structured model libraries. The historical turbulence models stored within each model library have the same model structure, such as two-equation models based on the eddy viscosity assumption, or Reynolds stress transport equation models, while different model libraries correspond to completely different modeling theoretical frameworks.

[0064] When it is necessary to determine a reference turbulence model for the current primary turbulence model, these model libraries are traversed to identify a target library that differs fundamentally from the primary turbulence model in terms of model structure. This difference in model structure ensures that the introduced reference turbulence model carries a completely new modeling approach, providing a broader scope for exploration in subsequent model fusion and optimization.

[0065] After determining the target library, a bias evaluation is performed on each historical turbulence model stored in the library. These biases are quantitative indicators calculated between the historical turbulence models and real flow field data during previous iterations when they were validated using preset examples. By comparing these biases, the prediction accuracy and error distribution characteristics of each historical turbulence model under specific flow scenarios can be accurately determined.

[0066] Based on bias, historical turbulence models that perform best or have specific advantages in the target library are selected as reference turbulence models. For example, a historical turbulence model with the smallest bias in predicting separated flows might be chosen, or a historical turbulence model that performs robustly in the near-wall region might be selected. The specific selection strategy depends on the error characteristics of the current first turbulence model and the optimization objective. This mechanism ensures that the selection process of the reference turbulence model has a clear data-driven basis, rather than blindly trying different approaches, thereby significantly improving the directionality and efficiency of iterative optimization.

[0067] In practical applications, such as the simulation of complex internal flows of aero engines, this screening method that combines structural differentiation with performance optimization can quickly identify the most valuable turbulence model, effectively guide the large model to generate a more accurate second turbulence model, reduce a large number of invalid trial calculations, significantly reduce the consumption of computing resources while ensuring model accuracy, and accelerate the entire turbulence model development process.

[0068] Furthermore, if it is detected that all historical turbulence models stored in a model library produce deviations exceeding a predetermined threshold when validated against preset examples, this means that the historical turbulence models with this specific model structure can no longer meet the accuracy requirements for this flow scenario. In this case, to reduce the storage space occupied by invalid historical turbulence models and their interference with the selection direction in subsequent iterations, all historical turbulence models with excessive deviations in the model library are removed at once. This operation not only frees up storage resources but, more importantly, eliminates the possibility of inferior historical turbulence models being selected as reference turbulence models again, ensuring that the model library always maintains a high-quality candidate set.

[0069] After cleaning, the process moves to model libraries with different model structures. New turbulence models meeting the accuracy requirements are selected from the historical turbulence models stored in these libraries and added to the emptied model libraries. The selection process is still based on the deviation performance of each historical turbulence model in preset examples, prioritizing models with high prediction accuracy or specific advantages. This cross-model library migration and filling ensures that each model library can continuously evolve and always contain the most representative turbulence models of the current iteration.

[0070] In practical applications, as simulation tasks progress, some model structures may be phased out in batches because they cannot adapt to new flow phenomena. New turbulence models will be added from other model libraries, keeping the entire model library system in a state of dynamic optimization. This provides high-quality references for subsequent model generation tasks of large models, thereby improving model accuracy and effectively reducing invalid iterations caused by model library fixation, saving computing resources and time costs.

[0071] Figure 4 The diagram illustrates a model library update of a turbulence model generation method according to an embodiment of the present disclosure.

[0072] like Figure 4 As shown, by constructing multiple independent model libraries, the historical turbulence models are classified, stored, and dynamically maintained. Each model library corresponds to a model structure. In the initial state, the model library corresponding to model structure 1 stores multiple historical turbulence models with the same mathematical expression but different specific parameters or correction terms, such as historical turbulence model 1-1, historical turbulence model 1-2, and historical turbulence model 1-3. Similarly, the model library corresponding to model structure 2 stores historical turbulence model 2-1, historical turbulence model 2-2, and historical turbulence model 2-3, etc.

[0073] As the iteration process progresses, based on the deviations obtained by each first turbulence model under preset calculations, a periodic update operation is performed independently on each model library. When the deviations of all historical turbulence models in a certain model library exceed a predetermined threshold, for example, when all historical turbulence models in the model library corresponding to model structure 1 fail to meet the accuracy requirements, the historical turbulence models in that model library are removed in batches. At the same time, historical turbulence models that meet the requirements are selected from other model libraries with excellent accuracy performance, and these are used as the basis for generating new turbulence models. These models are then moved into the emptied model library, thus completing the update of that model library.

[0074] For example, after the model library corresponding to model structure 1 is cleared, the historical turbulence model 2-1 from the model library corresponding to model structure 2 is introduced, and new turbulence models are iteratively generated based on this historical turbulence model 2-1, such as historical turbulence model 2-1-1, historical turbulence model 2-1-2, historical turbulence model 2-1-3, etc. Meanwhile, the model library corresponding to model structure 2 retains or continues to update the original models based on its own evaluation results.

[0075] This periodic update mechanism ensures that the model library corresponding to each model structure always retains the historical turbulence model with the highest reference value and best prediction accuracy in the current iteration stage, reducing interference with the subsequent selection process of reference turbulence models. At the same time, through cross-library model migration, it enables the indirect fusion and iterative evolution of the advantageous features of different model structures.

[0076] In practical applications, such as long-term R&D projects like aero-engine combustor design or hypersonic vehicle aerodynamic analysis, this mechanism can continuously provide high-quality reference candidates for turbulence model generation, significantly improving the directionality and efficiency of iterative optimization. While ensuring model accuracy, it greatly saves computational resources and time costs caused by invalid trial calculations.

[0077] According to an embodiment of this disclosure, the turbulence model generation method further includes: analyzing the deviation based on a preset mapping rule to determine the deviation interval in which the deviation is located, and determining the corresponding feedback information based on the deviation interval; wherein the preset mapping rule includes multiple deviation intervals and feedback information corresponding to each deviation interval, and the feedback information is used to describe the physical reasons for the deviation of the first turbulence model in the flow scenario.

[0078] In the preset mapping rules, the continuously changing deviations are discretized and classified into multiple deviation intervals with clear physical meanings. Each deviation interval is not a simple numerical range, but corresponds to a typical failure mode commonly encountered by the turbulence model in a specific flow scenario.

[0079] For example, one deviation interval can be defined as "the turbulence model over-predicts the turbulence dissipation rate, resulting in an undersized separated bubble," and another deviation interval can be defined as "the turbulence model fails to capture the Reynolds stress anisotropy, resulting in distorted secondary flow intensity." These deviation intervals and their corresponding physical causes are pre-constructed based on fluid mechanics theory and a large amount of prior knowledge.

[0080] Once the deviation of the current first turbulence model in the preset example is obtained, the deviation is matched to the corresponding deviation interval according to the preset mapping rules. This process may involve comprehensive consideration of multiple dimensions of the deviation, such as not only considering the average deviation magnitude, but also analyzing the spatial distribution characteristics of the deviation in the flow field, or the local deviation peaks of key physical quantities (such as wall friction drag and pressure coefficient).

[0081] After determining the deviation range, feedback information pre-bound to that range is extracted from the preset mapping rules. This feedback information, in the form of natural language or structured data, accurately describes the physical reasons for the deviation of the first turbulence model in the current flow scenario. For example, it indicates that the first turbulence model underestimates the attenuation effect of turbulence fluctuations in the adverse pressure gradient region. Generating feedback information in this way effectively incorporates diagnostic opinions from the physical mechanism analysis, providing corrective guidance for subsequent optimization of the larger model.

[0082] In practical applications, such as simulating boundary layer transition for hypersonic vehicles, this physics-based feedback helps large models understand whether the prediction bias is due to insufficient correction of compressibility effects or a lack of capture of crossflow instabilities, thus guiding the generation of more targeted correction schemes. This effectively improves the level of model iterative optimization, reduces blind attempts, and makes the correction process of turbulence models more efficient and accurate.

[0083] According to embodiments of this disclosure, the turbulence model generation method further includes: when there are multiple first turbulence models, converting each first turbulence model into a mathematical proof language based on the mathematical expressions in each first turbulence model and the physical description of each parameter in the mathematical expressions; converting preset physical constraints into verification logic for each parameter; performing reasoning verification on the mathematical proof language of each first turbulence model based on the verification logic to obtain physical verification results; and determining candidate turbulence models that satisfy preset physical constraints from multiple first turbulence models based on the physical verification results, so as to use the candidate turbulence models as the first turbulence models for simulation.

[0084] In the initial generation stage of the large model, when multiple first turbulence models are output, physical consistency verification will be performed on these first turbulence models to reduce the direct input of first turbulence models that do not conform to basic physical laws into subsequent simulation calculations.

[0085] Specifically, for each first turbulence model, its mathematical expression and the physical descriptions corresponding to each parameter in the mathematical expression (such as turbulent kinetic energy, dissipation rate, Reynolds stress components, etc.) are extracted, and symbolic computation and formal language tools are used to transform this information into mathematical proof language that can be processed by the automatic reasoning engine.

[0086] At the same time, the preset physical constraints, such as the entropy increase characteristic required by the second law of thermodynamics, the dimensional homogeneity that the turbulence model should satisfy, the symmetric positive definiteness of the Reynolds stress tensor, and the realizability conditions (such as the non-negativity of the turbulence time scale), are transformed into verification logic for each parameter in the first turbulence model.

[0087] Subsequently, the theorem prover or constraint solver is invoked to perform rigorous reasoning verification on the mathematical proof language of each first turbulence model based on the verification logic. The system automatically determines whether the first turbulence model satisfies the preset physical constraints under all possible flow states and finally outputs the physical verification results.

[0088] Based on the physical verification results, candidate turbulence models that have passed the verification are selected from multiple first turbulence models as the first turbulence models, and subsequent simulations are performed on the first turbulence models.

[0089] In practical applications, such as in the design of aerospace engine combustion chambers or the simulation of complex flows inside turbomachinery, where the reliability of turbulence models is critical, this mechanism can eliminate a large number of turbulence models with hidden physical contradictions before simulation, reducing invalid simulation iterations caused by inherent defects in turbulence models, thereby saving computational resources and time costs.

[0090] Meanwhile, by ensuring that the first turbulence model participating in the simulation has a physical basis, the first turbulence model can exhibit high accuracy and generalization ability when predicting simulated flow field data, which significantly improves the development efficiency and quality of turbulence models.

[0091] According to an embodiment of this disclosure, the turbulence model generation method further includes: in response to a physical verification result indicating that multiple first turbulence models do not meet preset physical constraints, regenerating the first turbulence model based on the physical verification result and model generation prompts.

[0092] When the physical verification results show that all the first turbulence models generated from the large model fail to meet the preset physical constraints, an in-depth analysis of the physical verification results generated during the physical verification process is performed. These results typically include detailed information on which preset physical constraints each first turbulence model violated. For example, it might indicate that the turbulent kinetic energy generation term of a certain first turbulence model could lead to negative turbulent kinetic energy in certain flow regions, or that the Reynolds stress tensor failed to maintain positive definiteness, or that there are inconsistencies in the dimensional analysis of the model equations.

[0093] By combining the physical verification results with the original model generation prompts, an enhanced input instruction with explicit correction guidance is formed. This input instruction not only reiterates the flow scenario and modeling requirements, but also clearly informs the large model of the first turbulence model generated previously on which preset physical constraints are flawed, and guides it to avoid the same error paths.

[0094] Subsequently, the enhanced prompts are re-inputted into the large model, driving it to search and generate new first turbulence models in a completely new solution space. These new first turbulence models incorporate strategies to avoid the preset physical constraints during the generation process, thus having a higher probability of satisfying the preset physical constraints.

[0095] In practical applications, such as the design of aero-engine combustors or the development of thermal protection systems for hypersonic vehicles, where the physical reliability of the model is extremely demanding, this mechanism can effectively reduce the waste of subsequent simulation resources caused by the inherent physical defects of turbulence models, ensuring that each simulation iteration is based on a solid physical foundation.

[0096] Meanwhile, by providing real-time feedback of the failures in physical verification results to the large model, the hit rate and iteration efficiency of model generation are effectively improved, and the conversion cycle from requirements to effective models is shortened. This significantly saves hardware computing costs while ensuring the high accuracy of the final model.

[0097] Figure 5 A schematic diagram of a system architecture for a turbulence model generation method according to another embodiment of the present disclosure is shown.

[0098] like Figure 5 As shown, during the automated generation and optimization of turbulence models, knowledge from literature can be extracted through the Model Context Protocol (MCP) service to construct model generation prompts. These prompts are then input into the Large Language Model (LLM) to initially generate several candidate turbulence models, such as turbulence model 1, turbulence model 2, turbulence model 3, and turbulence model 4.

[0099] Subsequently, the mathematical proof tool is invoked through the MCP service to verify the physical constraints of the generated multiple turbulence models, ensuring that they satisfy physical constraints such as the second law of thermodynamics and dimensional homogeneity. Only the verified turbulence models (such as turbulence model 1 and turbulence model 3) are retained and proceed to the next processing stage. If none of them satisfy the physical constraints, the physical verification results are fed back to the larger model, and a new turbulence model is generated. By introducing mathematical proof tools to pre-select turbulence models that do not meet the physical laws, a large number of invalid simulation trials can be reduced, significantly saving computational resources and time costs.

[0100] The selected turbulence model, baseline turbulence model, and task description are combined and input into LLM again to generate executable turbulence model code (such as turbulence model 1 code and turbulence model 3 code). The generated turbulence model code is then input into the CFD solver for compilation and calculation. If compilation or calculation errors occur, the error information is fed back to LLM via the MCP service to drive code correction. If the calculation is successful, the simulated flow field data is output and model verification is performed.

[0101] During the validation process, the performance of each turbulence model is recorded by the memory management module, and poorly performing models are periodically cleaned up based on model scores to ensure that high-performance models are always retained in the model library. The model library is stored according to model structure, such as Model Library 1, Model Library 2, and Model Library 3. These models can be used as reference turbulence models in subsequent iterations, or optional intervention operations can be performed under expert guidance.

[0102] The periodic update mechanism for the memory management module and model library ensures the quality and diversity of turbulence models, guiding LLM to generate more accurate turbulence models. Ultimately, in practical applications (such as aircraft aerodynamic design and turbomachinery optimization), high-confidence flow field prediction results can be obtained quickly, significantly improving the accuracy and speed of R&D iterations while reducing hardware investment.

[0103] Figure 6 A flowchart illustrating a turbulence simulation method according to an embodiment of the present disclosure is shown schematically.

[0104] like Figure 6 As shown, the method includes operation S610.

[0105] In operation S610, the target operating condition parameters are input into the target turbulence model, and the target flow field data is output. The target flow field data characterizes the turbulence features under the flow scenario indicated by the target operating condition parameters. The target turbulence model is generated by the turbulence model generation method.

[0106] According to embodiments of this disclosure, after a validated and optimized target turbulence model is generated using a turbulence model generation method, the target turbulence model can be deployed to actual engineering application scenarios to predict flow characteristics under specific working conditions.

[0107] Specifically, the target operating parameters are structured, which typically include geometric configuration, boundary conditions, inlet velocity, Reynolds number, Mach number, angle of attack, and thermodynamic state, thus fully describing the physical environment and initial state of the flow scenario to be solved.

[0108] Subsequently, the target operating condition parameters are input into the target turbulence model, which has been compiled into code or encapsulated in the CFD solver. The mathematical expressions inside the target turbulence model will be automatically assigned and calculated based on the input target operating condition parameters.

[0109] Through the numerical solution process, the target turbulence model finally outputs the target flow field data, which specifically includes key physical quantities such as velocity vector field, pressure distribution, temperature field, turbulent kinetic energy and its dissipation rate, and wall friction drag coefficient. It can comprehensively reflect the statistical characteristics and transient structure of turbulent motion under this working condition.

[0110] For example, in typical applications such as aircraft wing flow analysis, gas turbine blade cooling design, or automotive external aerodynamic optimization, this process enables engineers to quickly obtain high-fidelity flow field information to evaluate aerodynamic performance, identify separation zones, or predict heat load distribution.

[0111] Since the target turbulence model itself has achieved high-precision prediction capabilities for specific flow scenarios through iterative optimization, it can significantly improve the accuracy of flow field data while maintaining computational efficiency during actual working condition predictions. This reduces the dependence on dense grids or computing resources and effectively lowers hardware investment costs and energy consumption.

[0112] According to embodiments of this disclosure, the turbulence simulation method further includes: in response to receiving simulation feedback information for target flow field data, if the simulation feedback information indicates that the simulation results need to be adjusted, determining a candidate turbulence model from multiple historical turbulence models, wherein the model structure of the candidate turbulence model is different from the model structure of the target turbulence model; inputting target operating condition parameters into the candidate turbulence model, and outputting simulated flow field data.

[0113] After generating the target flow field data, if simulation feedback indicates that the current simulation results need adjustment, the adjustment requirements in the simulation feedback are analyzed. This simulation feedback may stem from engineers' questions about specific flow characteristics, such as doubts about overly conservative separation zone predictions, shock wave location deviations, or unreasonable heat load distribution. It may also originate from the automated post-processing module's detection of exceeding limits for key aerodynamic parameters.

[0114] From multiple model libraries categorized by model structure, a candidate turbulence model is selected that differs in structure from the current target turbulence model. This structural difference ensures that the selected candidate turbulence model can examine the same flow problem from different physical modeling perspectives, such as switching from the eddy viscosity assumption to the Reynolds stress transport framework, or from linear constitutive relations to nonlinear modified modes.

[0115] After determining the candidate turbulence model, the original target operating condition parameters, including geometric configuration, boundary conditions, and incoming flow parameters, are completely input into the candidate turbulence model, and the appropriate solver module is called to perform simulation calculations, ultimately outputting new simulated flow field data.

[0116] In practical applications, such as high-risk scenarios like aircraft icing airworthiness certification or hypersonic inlet design, this mechanism enables engineers to quickly obtain flow predictions from multiple independent turbulence model perspectives, and improve the reliability of conclusions through cross-validation.

[0117] Meanwhile, the switching process of turbulence models relies on existing model libraries and automated simulation, which can significantly improve the efficiency of comparative analysis of multiple turbulence models and reduce design iterations caused by the limitations of a single model, thereby effectively saving computing resources while ensuring the comprehensiveness of the analysis.

[0118] Figure 7 An interactive schematic diagram of a turbulence simulation method according to an embodiment of the present disclosure is shown.

[0119] like Figure 7 As shown, when the user inputs the target operating condition parameters, the large model calls the target turbulence model obtained through the aforementioned generation method to perform simulation calculations, outputting target flow field data that can characterize the flow characteristics under that operating condition, such as key physical quantities like pressure distribution, velocity field, or turbulent kinetic energy. This process enables the rapid deployment of the optimized high-precision model to specific application scenarios.

[0120] Subsequently, if the user provides simulation feedback indicating that the current results require further adjustment or verification, the large model will select a candidate turbulence model from the model library with a different model structure than the target turbulence model, and then call this candidate turbulence model again to simulate the same target operating parameters, outputting entirely new simulated flow field data. The entire calling process achieves human-computer collaboration through an interactive box, enabling users to easily trigger multi-model comparative analysis.

[0121] In practical applications, such as aircraft aerodynamic shape iteration or automobile aerodynamic performance evaluation, this mechanism enables researchers to quickly obtain cross-validation results from different modeling perspectives based on the predictions of a single turbulence model, significantly improving the comprehensiveness of the understanding of complex flow phenomena and the reliability of the conclusions.

[0122] Meanwhile, since the selection of alternative turbulence models relies on existing model libraries and automated simulation processes, the efficiency of comparing and selecting multiple schemes is greatly improved, and the design risks caused by the limitations of a single model are effectively reduced. Thus, while ensuring the accuracy of analysis, R&D time and computing resource costs are significantly saved.

[0123] Figure 8 A block diagram of a turbulence model generation apparatus according to an embodiment of the present disclosure is shown schematically.

[0124] like Figure 8 As shown, the turbulence model generation device 800 includes a first generation module 810, a model simulation module 820, a reference determination module 830, and a second generation module 840.

[0125] The first generation module 810 is used to input model generation prompts into the large model to generate the first turbulence model.

[0126] The model simulation module 820 is used to simulate the first turbulence model based on a preset example, and obtain the simulation flow field data of the first turbulence model under the flow scenario corresponding to the preset example to simulate turbulence phenomena.

[0127] The reference determination module 830 is used to determine a reference turbulence model from multiple historical turbulence models generated by the large model in response to the deviation between the simulated flow field data and the real flow field data in the preset example being greater than a predetermined threshold. The reference turbulence model has a different model structure than the first turbulence model.

[0128] The second generation module 840 is used to input the feedback information generated based on the deviation and the reference turbulence model into the large model to generate a second turbulence model.

[0129] According to embodiments of this disclosure, the model simulation module 820 includes a code compilation submodule, a variable assignment submodule, and a code execution submodule.

[0130] The code compilation submodule is used to compile the mathematical expressions in the first turbulence model into simulation code.

[0131] The variable assignment submodule is used to assign values ​​to the corresponding variables in the simulation code based on the flow condition parameters in the preset example, so as to obtain the target code.

[0132] The code execution submodule is used to run the target code and obtain the simulation flow field data generated by the first turbulence model simulating turbulence phenomena under the flow scenario corresponding to the preset example.

[0133] According to embodiments of this disclosure, the turbulence model generation apparatus 800 includes a code correction module and a code simulation module.

[0134] The code correction module is used to respond to errors occurring when running the target code, input the acquired error information into the large model, and correct the simulation code based on the error information to obtain the corrected code. The error information includes at least one of compilation error information and solution error information.

[0135] The code simulation module is used to re-simulate based on the corrected code.

[0136] According to embodiments of this disclosure, the reference determination module 830 includes a target library determination submodule and a reference determination submodule.

[0137] The target library determination submodule is used to determine the target library from multiple model libraries stored according to model structure. Each model library stores historical turbulence models with the same model structure, and the model structure corresponding to the target library is different from the model structure of the first turbulence model.

[0138] The reference determination submodule is used to determine a reference turbulence model from multiple historical turbulence models based on the deviations of each historical turbulence model in the target library.

[0139] According to embodiments of this disclosure, the turbulence model generation apparatus 800 includes a feedback determination module.

[0140] The feedback determination module is used to analyze the deviation based on preset mapping rules, determine the deviation range in which the deviation is located, and determine the corresponding feedback information based on the deviation range.

[0141] The preset mapping rules include multiple deviation intervals and feedback information corresponding to each deviation interval. The feedback information is used to describe the physical reasons for the deviation of the first turbulence model in the flow scenario.

[0142] According to embodiments of this disclosure, the turbulence model generation device 800 includes a mathematical transformation module, a condition transformation module, a physical verification module, and a model screening module.

[0143] The mathematical transformation module is used to transform each first turbulence model into a mathematical proof language based on the mathematical expressions in each first turbulence model and the physical description of each parameter in the mathematical expressions, when there are multiple first turbulence models.

[0144] The condition conversion module is used to convert preset physical constraints into verification logic for each parameter.

[0145] The physical verification module is used to perform reasoning verification on the mathematical proof language of each first turbulence model based on the verification logic, and obtain the physical verification results.

[0146] The model selection module is used to determine candidate turbulence models that meet preset physical constraints from multiple first turbulence models based on physical verification results, so as to use the candidate turbulence models as the first turbulence models for simulation.

[0147] According to embodiments of this disclosure, the turbulence model generation apparatus 800 includes a regeneration module.

[0148] The regeneration module is used to regenerate the first turbulence model based on the physical verification results and model generation prompts when multiple first turbulence models do not meet the preset physical constraints.

[0149] Figure 9 A block diagram of a turbulence simulation apparatus according to an embodiment of the present disclosure is shown schematically.

[0150] like Figure 9 As shown, the turbulence simulation device 900 includes a turbulence simulation module 910.

[0151] The turbulence simulation module is used to input target operating condition parameters into the target turbulence model and output target flow field data, wherein the target flow field data characterizes the turbulence features under the flow scenario indicated by the target operating condition parameters; wherein the target turbulence model is generated by the aforementioned turbulence model generation method.

[0152] Figure 10 A schematic block diagram of a large-scale intelligent agent according to an embodiment of the present disclosure is shown.

[0153] In embodiments of this disclosure, the von Neumann architecture in modern computer theory is inspired, such as... Figure 10 As shown, the AI ​​agent 1000 may include five core modules: input module 1010, processing module 1020 and output module 1030.

[0154] The input module 1010 is responsible for receiving or sensing information such as queries, requests, instructions, signals, or data from the outside world (e.g., users or the external environment), and converting it into a format that the AI ​​agent 1000 can understand and process. The input module 1010 is the primary link for the AI ​​agent 1000 to interact with the outside world. It enables the AI ​​agent 1000 to efficiently and accurately obtain the necessary "sensory" information from the outside world and respond to this information.

[0155] In the example, input module 1010 can input the prompt words generated by the model described above.

[0156] In embodiments of this disclosure, the processing module 1020 may include a control module 1021, a storage module 1022, and a computation module 1023. The processing module 1020 is configured to determine a target task based on input information received by the input module 1010, determine a target large-scale model based on the target task, and obtain output information by invoking a turbulence model generation method using the target large-scale model. In the example, the target large-scale model includes at least one of a visual language large-scale model, a video generation large-scale model, and a video evaluation large-scale model.

[0157] The control module 1021 is the core support for the AI ​​agent 1000's ability to handle complex tasks. The control module 1021 can execute the turbulence model generation method described above.

[0158] In the example, the control module 1021 will continuously interact with the storage module 1022, the arithmetic module 1023, and / or the output module 1030 during operation. However, it should be noted that in the embodiments of this disclosure, the control module 1021 initiates communication with the storage module 1022, the arithmetic module 1023, and / or the output module 1030 as a single initiator, and there is no communication coupling between the storage module 1022, the arithmetic module 1023, and the output module 1030.

[0159] In the example, the performance of the control module 1021 is closely related to the large model on which the AI ​​agent 1000 is based. To fully leverage the capabilities of the large language model, the internal structure of the control module 1021 can be designed to be highly configurable and scalable to handle various types of tasks and requirements in real-world scenarios.

[0160] Storage module 1022 can be responsible for storing the large visual language model, the large video generation model, and the large video evaluation model. The aforementioned large visual language model, large video generation model, and large video evaluation model can be included in storage module 1022.

[0161] In the example, after receiving the model generation prompt, the AI ​​agent 1000 can trigger the code generation process to obtain a second turbulence model and feed it back to the control module 1021. Then, the control module 1021 can pass the feedback second turbulence model to the output module 1030.

[0162] The computation module 1023 can be viewed as a predefined tool library. Tools for timing alignment, as described above, can be included in the computation module 1023.

[0163] In the example, when the AI ​​agent 1000 needs to process data, it can invoke relevant tools from the computing module 1023 and feed them back to the control module 1021. The control module 1021 can then use the fed-back tools to process the relevant data. It's understandable that while large language models have excellent language understanding and generation capabilities, like humans, the tasks they can solve without any tools are very limited. When the AI ​​agent 1000 is given the ability to invoke tools, it can perform tasks such as timing alignment using tools designed for timing alignment.

[0164] Output module 1030 can output the second turbulence model described above.

[0165] The AI ​​agent 1000 according to the embodiments of this disclosure can simply and effectively improve the level of intelligence, as well as enhance flexibility and versatility.

[0166] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0167] According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described above.

[0168] According to embodiments of the present disclosure, a non-transitory computer-readable storage medium stores computer instructions, wherein the computer instructions are used to cause a computer to perform the methods described above.

[0169] According to an embodiment of this disclosure, a computer program product includes a computer program that, when executed by a processor, implements the method described above.

[0170] Figure 11 A block diagram of an electronic device suitable for implementing a turbulence model generation method and a turbulence simulation method according to embodiments of the present disclosure is shown schematically.

[0171] Electronic devices are intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0172] like Figure 11 As shown, device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in read-only memory (ROM) 1102 or a computer program loaded into random access memory (RAM) 1103 from storage unit 1108. The RAM 1103 may also store various programs and data required for the operation of device 1100. The computing unit 1101, ROM 1102, and RAM 1103 are interconnected via bus 1104. Input / output (I / O) interface 1105 is also connected to bus 1104.

[0173] Multiple components in device 1100 are connected to input / output (I / O) interface 1105, including: input unit 1106, such as a keyboard, mouse, etc.; output unit 1107, such as various types of displays, speakers, etc.; storage unit 1108, such as a disk, optical disk, etc.; and communication unit 1109, such as a network card, modem, wireless transceiver, etc. Communication unit 1109 allows device 1100 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0174] The computing unit 1101 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1101 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1101 performs the various methods and processes described above, such as turbulence model generation methods and turbulence simulation methods. For example, in some embodiments, the turbulence model generation methods and turbulence simulation methods can be implemented as computer software programs tangibly contained in a machine-readable medium, such as storage unit 1108. In some embodiments, part or all of the computer program can be loaded and / or installed on device 1100 via ROM 1102 and / or communication unit 1109. When the computer program is loaded into RAM 1103 and executed by the computing unit 1101, one or more steps of the turbulence model generation methods and turbulence simulation methods described above can be performed. Alternatively, in other embodiments, the computing unit 1101 may be configured to perform turbulence model generation methods and turbulence simulation methods by any other suitable means (e.g., by means of firmware).

[0175] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0176] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0177] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0178] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0179] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0180] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, distributed system servers, or servers incorporating blockchain technology.

[0181] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0182] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for generating a turbulence model, comprising: Input the model generation prompts into the large model to generate the first turbulence model; The first turbulence model is simulated based on a preset example to obtain the simulated flow field data of the first turbulence model under the flow scenario corresponding to the preset example. In response to the deviation between the simulated flow field data and the real flow field data in the preset example being greater than a predetermined threshold, a reference turbulence model is determined from multiple historical turbulence models generated by the large model, wherein the reference turbulence model has a different model structure than the first turbulence model; The feedback information generated based on the deviation and the reference turbulence model are input into the large model to generate a second turbulence model.

2. The method according to claim 1, wherein, The simulation of the first turbulence model based on a preset calculation example yields simulated flow field data of the first turbulence model simulating turbulence phenomena under the flow scenario corresponding to the preset calculation example, including: Compile the mathematical expressions in the first turbulence model into simulation code; Based on the flow condition parameters in the preset example, the corresponding variables in the simulation code are assigned values ​​to obtain the target code; Running the target code yields the simulation flow field data generated by the first turbulence model simulating turbulence phenomena in the flow scenario corresponding to the preset example.

3. The method according to claim 2, further comprising: In response to an error occurring while running the target code, the acquired error information is input into the large model to correct the simulation code based on the error information, thereby obtaining corrected code. The error information includes at least one of compilation error information and solution error information. The simulation was then performed again based on the corrected code.

4. The method according to claim 1, wherein, The process of determining a reference turbulence model from multiple historical turbulence models generated from the large model includes: A target library is determined from multiple model libraries classified and stored according to model structure, wherein each model library stores historical turbulence models with the same model structure, and the model structure corresponding to the target library is different from the model structure of the first turbulence model. Based on the deviations of the historical turbulence models in the target library, the reference turbulence model is determined from the multiple historical turbulence models.

5. The method according to claim 1, further comprising: Based on preset mapping rules, the deviation is analyzed to determine the deviation range in which the deviation is located, and the corresponding feedback information is determined based on the deviation range. The preset mapping rule includes multiple deviation intervals and feedback information corresponding to each deviation interval. The feedback information is used to describe the physical reasons why the first turbulence model generates the deviation in the flow scenario.

6. The method according to claim 1, further comprising: When there are multiple first turbulence models, each first turbulence model is transformed into a mathematical proof language based on the mathematical expressions in each first turbulence model and the physical description of each parameter in the mathematical expressions. The preset physical constraints are transformed into verification logic for each of the parameters. Based on the verification logic, the mathematical proof language of each of the first turbulence models is used for reasoning verification to obtain physical verification results; Based on the physical verification results, candidate turbulence models that satisfy the preset physical constraints are determined from multiple first turbulence models, and the candidate turbulence models are used as the first turbulence models for simulation.

7. The method according to claim 6, further comprising: In response to the physical verification results indicating that multiple first turbulence models do not meet the preset physical constraints, the first turbulence model is regenerated based on the physical verification results and the model generation prompts.

8. A turbulence simulation method, comprising: The target operating condition parameters are input into the target turbulence model, and the target flow field data is output. The target flow field data characterizes the turbulence features under the flow scenario indicated by the target operating condition parameters. The target turbulence model is generated by the method described in any one of claims 1 to 7.

9. The method according to claim 8, further comprising: In response to receiving simulation feedback information for the target flow field data, if the simulation feedback information indicates that the simulation results need to be adjusted, a candidate turbulence model is determined from multiple historical turbulence models, wherein the model structure of the candidate turbulence model is different from that of the target turbulence model. The target operating condition parameters are input into the alternative turbulence model, and the simulated flow field data is output.

10. A turbulence model generation device, comprising: The first generation module is used to input model generation prompts into the large model and generate the first turbulence model. The model simulation module is used to simulate the first turbulence model based on a preset example to obtain the simulation flow field data generated by the first turbulence model simulating turbulence phenomena under the flow scenario corresponding to the preset example. A reference determination module is used to determine a reference turbulence model from multiple historical turbulence models generated by the large model in response to the deviation between the simulated flow field data and the real flow field data in the preset example being greater than a predetermined threshold. The reference turbulence model has a different model structure than the first turbulence model. The second generation module is used to input the feedback information generated based on the deviation and the reference turbulence model into the large model to generate a second turbulence model.

11. A turbulence simulation device, comprising: The turbulence simulation module is used to input target operating condition parameters into the target turbulence model and output target flow field data, wherein the target flow field data characterizes the turbulence features under the flow scenario indicated by the target operating condition parameters; The target turbulence model is generated by the method described in any one of claims 1 to 7.

12. An intelligent agent for model generation, comprising: The input module is used to receive prompts generated by the model; The processing module is configured to determine the target task based on the model-generated prompts received by the input module, determine the large model based on the target task, and obtain the second turbulence model by calling the large model to execute the method of any one of claims 1 to 7. The output module is used to output the second turbulence model obtained by the processing module.

13. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 9.

14. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 9.

15. A computer program product comprising a computer program stored on at least one of a readable storage medium and an electronic device, the computer program implementing the method according to any one of claims 1 to 9 when executed by a processor.

Images