Cavitation simulation method, system and computer readable storage medium

Abstract

The application provides a cavitation simulation method, system and computer readable storage medium. The cavitation simulation method comprises the following steps: obtaining macro flow characteristic data of a reference cavitation flow; simulating micro kinetic behavior of bubbles in the reference cavitation flow based on a microscale bubble dynamics model to obtain microscale bubble dynamics characteristics; constructing a training sample and training a machine learning model according to the macro flow characteristic data and the microscale bubble dynamics characteristics, and establishing a mapping relationship between the macro flow characteristics and the microscale bubble dynamics characteristics; embedding the trained machine learning model into a macroscale cavitation model, so that the macroscale cavitation model is corrected according to the microscale bubble dynamics characteristics output by the machine learning model to obtain a multiscale cavitation model; and solving a target cavitation flow according to the multiscale cavitation model to obtain a numerical simulation result of the target cavitation flow.

Description

Technical Field

[0001] This invention relates to fluid mechanics, and more particularly to a cavitation simulation method, system, and computer-readable storage medium. Background Technology

[0002] Cavitation is a common physical phenomenon in fluid machinery operation, and accurate numerical simulation of it is of great significance for the design and operation of fluid equipment. In computational fluid dynamics, cavitation models are usually used to describe the phase transition process between gas and liquid phases, and the generation and collapse of cavitation are introduced into the flow control equations through phase transition source terms.

[0003] However, the phase transition source terms of existing cavitation models rely on empirically given parameters, which are usually taken as fixed values ​​throughout the computational domain, making it difficult to reflect the differences in cavitation phase transitions between different spatial regions and at different times; under complex operating conditions, their prediction accuracy for cavitation processes often fails to meet the requirements. Summary of the Invention

[0004] The purpose of this invention is to provide a cavitation simulation method, system, and computer-readable storage medium.

[0005] One aspect of the present invention provides a method for cavitation simulation, comprising: acquiring macroscopic flow characteristic data of a reference cavitation flow; simulating the microscopic dynamic behavior of bubbles in the reference cavitation flow based on a microscopic bubble dynamics model to obtain microscopic bubble dynamic characteristics; constructing training samples and training a machine learning model based on the macroscopic flow characteristic data and the microscopic bubble dynamic characteristics to establish a mapping relationship between the macroscopic flow characteristics and the microscopic bubble dynamic characteristics; embedding the trained machine learning model into the macroscopic cavitation model, so that the macroscopic cavitation model is corrected according to the microscopic bubble dynamic characteristics output by the machine learning model to obtain a multi-scale cavitation model; and solving the target cavitation flow based on the multi-scale cavitation model to obtain the numerical simulation results of the target cavitation flow.

[0006] In some embodiments, the step of correcting the macroscale cavitation model based on the microscale bubble dynamics characteristics output by the machine learning model includes: using the microscale bubble dynamics characteristics to correct the phase transition source term of the macroscale cavitation model.

[0007] In some embodiments, the phase change source term includes an evaporation term characterizing bubble formation and a condensation term characterizing bubble collapse; the step of modifying the phase change source term of the macroscopic cavitation model using the microscopic bubble dynamics characteristics includes: modifying at least one of the evaporation term and the condensation term using the microscopic bubble dynamics characteristics.

[0008] In some embodiments, during the process of solving the target cavitation flow in computational fluid dynamics, the macroscopic flow characteristics obtained from the current solution are used as the input of the machine learning model, and the macroscopic cavitation model is modified according to the microscopic bubble dynamic characteristics output by the machine learning model, so that the multi-scale cavitation model is updated as the target cavitation flow evolves.

[0009] In some embodiments, the macroscopic flow characteristic data includes the pressure and velocity of the reference cavitation flow; obtaining the macroscopic flow characteristic data of the reference cavitation flow includes: collecting physical environment parameters around the bubbles in the reference cavitation flow through pressure sensors, temperature sensors, and flow velocity sensors; collecting experimental data of the reference cavitation flow; and extracting features from the experimental data to obtain the macroscopic flow characteristic data.

[0010] In some embodiments, before feature extraction from the experimental data, the method further includes: cleaning the experimental data, which includes removing noise, invalid values, and duplicate values, and smoothing or removing outliers.

[0011] In some embodiments, the machine learning model is a regression model or a neural network model; the neural network model includes a convolutional neural network or an artificial neural network.

[0012] In some embodiments, before embedding the trained machine learning model into a macro-scale cavitation model, the method further includes: validating the accuracy of the trained machine learning model using a validation dataset independent of the training samples.

[0013] In some embodiments, when solving the target cavitation flow in computational fluid dynamics, the region in the target cavitation flow where cavitation occurs is meshed; the mesh refinement satisfies a mesh refinement criterion, which is the ratio of the volume of the refined fine mesh to the volume of the original coarse mesh.

[0014] Another aspect of the present invention provides a cavitation simulation system, comprising: a data acquisition module for acquiring macroscopic flow characteristic data of a reference cavitation flow; a microscopic simulation module for simulating the microscopic dynamic behavior of bubbles in the reference cavitation flow based on a microscopic-scale bubble dynamics model to obtain microscopic-scale bubble dynamics characteristics; a model training module for constructing training samples and training a machine learning model based on the macroscopic flow characteristic data and the microscopic-scale bubble dynamics characteristics to establish a mapping relationship between the macroscopic flow characteristics and the microscopic-scale bubble dynamics characteristics; a model construction module for embedding the trained machine learning model into a macroscopic-scale cavitation model, so that the macroscopic-scale cavitation model is corrected according to the microscopic-scale bubble dynamics characteristics output by the machine learning model to obtain a multi-scale cavitation model; and a solution module for solving a target cavitation flow in computational fluid dynamics based on the multi-scale cavitation model to obtain numerical simulation results of the target cavitation flow.

[0015] Another aspect of the present invention provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the cavitation simulation method as described in any of the above embodiments.

[0016] The cavitation simulation method of this invention simulates the microscopic dynamic behavior of bubbles in a reference cavitation flow using a microscale bubble dynamics model to obtain microscale bubble dynamic characteristics. A machine learning model is trained based on the macroscopic flow characteristics of the reference cavitation flow and these microscale bubble dynamic characteristics, establishing a mapping relationship between them. The trained machine learning model is then embedded into a macroscale cavitation model, which is modified based on the microscale bubble dynamic characteristics output by the machine learning model to form a multiscale cavitation model. This introduces microscale bubble dynamic information, typically missed by a single macroscale cavitation model, into the macroscale cavitation model. Thus, when numerically solving the target cavitation flow using the multiscale cavitation model, the numerical simulation results incorporate microscale bubble dynamic information, improving the prediction accuracy of the cavitation flow. Furthermore, since the solution for the target cavitation flow relies solely on the trained multiscale cavitation model for prediction, without requiring re-collection of experimental data or microscopic simulation of the target cavitation flow itself, it can be applied to target cavitation flows lacking experimental measurement conditions, possessing both versatility and engineering feasibility. Attached Figure Description

[0017] The above and other features, properties and advantages of the present invention will become more apparent from the following description taken in conjunction with the accompanying drawings and embodiments, wherein: Figure 1 This is a flowchart illustrating an embodiment of the cavitation simulation method according to the present invention; Figure 2 This is a flowchart illustrating another embodiment of the cavitation simulation method according to the present invention; Figure 3 This is a schematic diagram of the components of the visualization tool of the present invention. Detailed Implementation

[0018] Reference will now be made in detail to embodiments of the invention, one or more examples of which are illustrated in the accompanying drawings. Each example is provided to explain the invention and not to limit it. In fact, it will be apparent to those skilled in the art that various modifications and variations may be made to the invention without departing from the scope or spirit thereof. For example, a feature shown or described as part of one embodiment may be used with another embodiment to produce yet another embodiment. Therefore, the invention is intended to cover these modifications and variations that fall within the scope of the appended claims and their equivalents.

[0019] As used herein, the terms “first,” “second,” and “third” are used interchangeably to distinguish one component from another and are not intended to indicate the location or importance of individual components.

[0020] As used in this article, the term "cavitation" refers to the phenomenon in which bubbles are generated and subsequently collapse when the local pressure inside a fluid drops below its saturated vapor pressure.

[0021] The term "computational fluid dynamics" (CFD) refers to the technique of simulating fluid motion by solving flow control equations using numerical methods.

[0022] The term "macroscale cavitation model" refers to a cavitation model that describes the gas-liquid phase transition on a continuous medium scale through phase transition source terms, which include evaporation terms characterizing bubble formation and condensation terms characterizing bubble collapse.

[0023] The term "microscale bubble dynamics" refers to the description of the formation, growth, and collapse behavior of bubbles at the bubble scale.

[0024] The terms “Molecular Dynamics” (MD) and “Direct Numerical Simulation” (DNS) are alternative methods for microscale simulations.

[0025] The terms "Convolutional Neural Network" (CNN) and "Artificial Neural Network" (ANN) are optional implementations of machine learning models.

[0026] The cavitation simulation method of this invention obtains microscale bubble dynamic characteristics by simulating the microscopic dynamic behavior of bubbles at a microscale. A machine learning model is then used to establish a mapping relationship between macroscopic flow characteristics and microscale bubble dynamic characteristics. The trained machine learning model is then embedded into a macroscopic cavitation model, which is modified based on the microscale bubble dynamic characteristics output by the machine learning model to form a multi-scale cavitation model. This introduces microscale bubble formation, growth, and collapse information, which is usually missed by a single macroscopic cavitation model, into the numerical solution of cavitation flow. This allows the numerical simulation results to reflect microscale bubble dynamics, thereby improving the prediction accuracy of cavitation flow. Furthermore, since the microscale information is provided through prediction via the machine learning model, direct microscopic calculations for each bubble are avoided in the macroscopic solution, thus improving prediction accuracy while also considering computational feasibility.

[0027] Figure 1 The flowchart of a cavitation simulation method according to an embodiment of the present invention is shown. Figure 1 As shown, the method includes steps S100 to S500.

[0028] In step S100, macroscopic flow characteristic data of the reference cavitation flow are acquired. In one embodiment, the macroscopic flow characteristic data includes the pressure and velocity of the reference cavitation flow.

[0029] Optionally, the macroscopic flow characteristic data are derived from experimental data collected for reference cavitation flow and obtained through feature extraction.

[0030] Furthermore, the experimental data includes parameters for bubble formation, growth, and collapse. This data was collected by measuring the physical environment parameters surrounding the bubbles in the reference cavitation flow using pressure, temperature, and flow rate sensors. Temperature is one of the core parameters, determining the saturated vapor pressure and influencing a range of parameters such as density, viscosity, surface tension, and sound velocity, ensuring the accuracy of cavitation number calculations and microbubble dynamics.

[0031] This provides real and comprehensive training samples for the machine learning model, which helps improve the reliability of subsequent mapping relationships. Optionally, before feature extraction from the experimental data, data cleaning is also performed. Data cleaning includes removing noise, invalid values, and duplicate values, and smoothing or removing outliers, thereby improving the quality of the training data and thus improving the prediction accuracy of the machine learning model.

[0032] In step S200, the micro-scale bubble dynamics model is used to simulate the micro-dynamic behavior of bubbles in the reference cavitation flow to obtain the micro-scale bubble dynamics characteristics, including bubble formation, growth and collapse.

[0033] In one embodiment, the formation and motion of the bubbles are described by the Navier-Stokes equations (NS), which are expressed as follows:

[0034] Where ρ is the fluid density; u is the velocity field vector; u represents the gradient operator. Acting on the velocity field vector u; t is time; p is the pressure field; p represents the gradient operator The force is applied to a pressure field p; μ is the dynamic viscosity; and f is the volume force, such as gravity. This formula is used to describe the motion of fluids to predict the formation and trajectory of bubbles.

[0035] In one embodiment, the pressure changes of the bubble are described using the Rayleigh-Plesset (RP) equation, which describes the radial motion of a spherical bubble in an incompressible viscous fluid to predict bubble growth and collapse:

[0036] Where R is the bubble radius, The velocity of the bubble wall. The acceleration of the bubble wall motion, The density of the liquid phase is... The internal pressure of the bubble, This represents the local pressure at the bubble, representing the macroscopic flow field surrounding the bubble. The surface tension coefficient of the liquid. This refers to the dynamic viscosity of the liquid.

[0037] Optionally, the microscopic dynamics of bubbles can be simulated using molecular dynamics (MD) or direct numerical simulation (DNS) methods to capture the details of bubble formation and motion at the microscale, thereby providing more reliable microscale bubble dynamics features for machine learning models.

[0038] In step S300, training samples are constructed and a machine learning model is trained based on macroscopic flow characteristic data and microscopic bubble dynamics characteristics to establish a mapping relationship between macroscopic flow characteristics and microscopic bubble dynamics characteristics. In one embodiment, the machine learning model is a regression model, and its expression is:

[0039] Where y is the target variable, including cavitation intensity; X is the feature variable, including flow velocity and pressure; f(X) is the regression model; and ε is the error term. This formula establishes the relationship between input features and cavitation phenomena to predict cavitation intensity. The regression model can be an f(X) nonlinear regression model, which is a nonparametric regression, i.e., a neural network that needs to be trained.

[0040] As an extended implementation method, the machine learning model can also be a neural network model, including convolutional neural networks (CNN) or artificial neural networks (ANN), that is, using CNN, ANN or deep neural networks to replace the regression model to predict cavitation behavior.

[0041] Understandably, regression models are simple to implement and have low computational cost, while neural network models can fit the more complex nonlinear mapping relationship between macroscopic flow characteristics and microscopic bubble dynamics characteristics, thereby further improving prediction accuracy under complex working conditions and making the method adaptable to different operating conditions and fluid characteristics.

[0042] Furthermore, before using the trained machine learning model in subsequent steps, the accuracy of the trained machine learning model can be verified using a validation dataset independent of the training samples, thereby evaluating the model's generalization ability and avoiding overfitting.

[0043] In step S400, the trained machine learning model is embedded into the macro-scale cavitation model, so that the macro-scale cavitation model is corrected according to the micro-scale bubble dynamics characteristics output by the machine learning model, thus obtaining a multi-scale cavitation model.

[0044] In one embodiment, the correction includes: modifying the phase change source terms of the macroscale cavitation model using microscale bubble dynamics characteristics. The phase change source terms include an evaporation term characterizing bubble formation and a condensation term characterizing bubble collapse, and at least one of the evaporation and condensation terms is modified. Thus, the macroscale cavitation model can incorporate microscale bubble dynamics information, thereby improving the accuracy of the description of the phase change process.

[0045] In step S500, the target cavitation flow in computational fluid dynamics is solved according to the multi-scale cavitation model to obtain the numerical simulation results of the target cavitation flow.

[0046] In one embodiment, parallel computing can be accelerated using a multi-core processor or a graphics processing unit (GPU), thereby incorporating microscale information while maintaining solution efficiency.

[0047] Optionally, during the solution process for the target cavitation flow in step S500, the currently obtained macroscopic flow characteristics are used as input to the machine learning model, and the macroscopic cavitation model is corrected in real time based on the microscopic bubble dynamics characteristics output by the machine learning model, so that the multi-scale cavitation model is updated as the target cavitation flow evolves. In this way, the multi-scale cavitation model can adapt to the temporal and spatial changes of cavitation phase transition, thereby further improving the prediction accuracy of cavitation flow under unsteady operating conditions.

[0048] The update frequency can be selected to be every N time steps, where N = bubble relaxation time / Δt, and Δt is on the order of Rayleigh-Plesset characteristic time.

[0049] Furthermore, when solving for the target cavitation flow, the mesh can be refined in the region where cavitation occurs. The mesh refinement satisfies the mesh refinement criterion R, which is the ratio of the volume Vf of the refined fine mesh to the volume Vc of the original coarse mesh, i.e.:

[0050] Where Vf is the fine mesh volume and Vc is the coarse mesh volume; in areas with complex cavitation phenomena, such as areas with concentrated bubbles, a finer mesh is used.

[0051] In one embodiment, the region where cavitation occurs is characterized by the volume fraction of bubbles, expressed as follows:

[0052] Where α is the volume fraction of the bubbles, Vb is the bubble volume, and Vt is the total fluid volume, this formula describes the distribution of bubbles in the fluid to characterize the cavitation phenomenon. Thus, by refining the mesh locally only in the complex regions where cavitation occurs, the computational accuracy of the cavitation region can be improved without significantly increasing the global mesh size and computational cost, balancing accuracy and computational efficiency.

[0053] like Figure 2 As shown, another embodiment of the cavitation simulation method of the present invention includes steps S1 to S6: In step S1, experimental data is collected, data is cleaned, and features are extracted to obtain feature data.

[0054] In step S2, the coupling relationship between the micro model and the macro model is established, and the micro data is embedded into the macro model to enhance the accuracy of the overall model.

[0055] In step S3, a machine learning model is used to predict cavitation behavior, the feature data is used to train the model, the model parameters are optimized to obtain a multi-scale model, and the accuracy of the multi-scale model is verified using an independent dataset.

[0056] In step S4, the multi-scale model is applied to OpenFOAM and a data-driven machine learning prediction module is embedded, utilizing multi-core processors or GPUs for parallel computation.

[0057] In step S5, the prediction accuracy of the model is analyzed by comparing experimental data with numerical simulation results, the model performance is evaluated using statistical methods, bubble behavior and pressure change indicators are analyzed, and sensitivity analysis is performed.

[0058] In step S6, visualization tools are used to display bubble motion and pressure field distribution, forming a visual analysis.

[0059] refer to Figure 3 In one embodiment, after obtaining the numerical simulation results of the target cavitation flow, a visualization tool 200 is used to display the bubble motion and pressure field distribution, forming a visual analysis. For example... Figure 3 As shown, the visualization tool 200 includes a ParaView visualization unit 210 and a VisIt visualization unit 220. Optionally, the visualization analysis index is represented by the aforementioned volume fraction α to visually demonstrate the cavitation phenomenon. As an extension, the visualization tool 200 can also be implemented using other visualization tools besides ParaView and VisIt.

[0060] Furthermore, the predictive accuracy of the model can be analyzed by comparing experimental data with numerical simulation results. Statistical methods can be used to evaluate model performance, analyze bubble behavior and pressure change indicators, and perform sensitivity analysis. This facilitates intuitive analysis of cavitation phenomena and quantitative evaluation of the predictive performance of multi-scale cavitation models.

[0061] The cavitation flow examples in this invention can be selected from typical examples in published literature, including but not limited to NACA0015 hydrofoil, E779A submarine propeller, low specific speed centrifugal pump, and venturi tube.

[0062] Understandably, one embodiment of the present invention also provides a cavitation simulation system, which includes a data acquisition module, a microscopic simulation module, a model training module, a model building module, and a solution module. Specifically, the data acquisition module is used to acquire macroscopic flow characteristic data of a reference cavitation flow; the microscopic simulation module is used to simulate the microscopic dynamic behavior of bubbles in the reference cavitation flow based on a microscopic-scale bubble dynamics model to obtain microscopic-scale bubble dynamic characteristics; the model training module is used to construct training samples and train a machine learning model based on the macroscopic flow characteristic data and the microscopic-scale bubble dynamic characteristics to establish a mapping relationship between the macroscopic flow characteristics and the microscopic-scale bubble dynamic characteristics; the model building module is used to embed the trained machine learning model into the macroscopic-scale cavitation model, so that the macroscopic-scale cavitation model is corrected according to the microscopic-scale bubble dynamic characteristics output by the machine learning model, resulting in a multi-scale cavitation model; the solution module is used to solve the target cavitation flow in computational fluid dynamics based on the multi-scale cavitation model to obtain the numerical simulation results of the target cavitation flow. All of the above modules can run on a multi-core processor or GPU for parallel computing. Therefore, this system can incorporate microscopic bubble dynamics information into the macroscopic cavitation solution, improve the prediction accuracy of cavitation flow, and is easy to deploy and run in software form.

[0063] Understandably, one embodiment of the present invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the cavitation simulation method described in any of the above embodiments. Optionally, the computer-readable storage medium includes a read-only memory, a random access memory, a disk, or an optical disk, etc., which are media capable of storing program code, thereby facilitating the distribution, deployment, and reuse of the method.

[0064] This invention significantly improves the prediction accuracy of cavitation models by integrating multi-scale modeling and data-driven methods. This enables researchers to more accurately understand the dynamic behavior of bubbles and their impact on fluid flow, thereby optimizing equipment design. The improved model also demonstrates excellent computational efficiency, reducing computation time. This not only improves research efficiency but also reduces the consumption of computational resources, allowing more simulations to be completed in a shorter time.

[0065] This invention utilizes machine learning technology, enabling the model to adapt to different operating conditions and fluid characteristics, thus making it practical in a wider range of applications. This provides a flexible tool for further research on cavitation phenomena; by accurately predicting cavitation, researchers can identify potential problems during the design phase, optimize the design of pumps and other fluid equipment, thereby reducing equipment damage and maintenance costs, and improving their service life and operating efficiency.

[0066] When the cavitation simulation method of the present invention is implemented as a computer program, it can also be stored as an article of manufacture in a computer-readable storage medium. For example, computer-readable storage media may include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic stripes), optical discs (e.g., compact discs (CDs), digital multifunction discs (DVDs)), smart cards, and flash memory devices (e.g., electrically erasable programmable read-only memory (EPROM), cards, sticks, key drives). Furthermore, the various storage media described herein can represent one or more devices and / or other machine-readable media for storing information. The term "machine-readable medium" may include, but is not limited to, wireless channels and various other media (and / or storage media) capable of storing, containing, and / or carrying code and / or instructions and / or data.

[0067] It should be understood that the embodiments described above are merely illustrative. The embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or any combination thereof. For hardware implementation, the processor may be implemented within one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, and / or other electronic units designed to perform the functions described herein, or combinations thereof.

[0068] Some aspects of this application can be executed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The aforementioned hardware or software may be referred to as a "data block," "module," "engine," "unit," "component," or "system." The processor may be one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DAPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, or combinations thereof. Furthermore, aspects of this application may manifest as computer products residing in one or more computer-readable media, including computer-readable program code. For example, computer-readable media may include, but are not limited to, magnetic storage devices (e.g., hard disks, floppy disks, magnetic tapes, etc.), optical discs (e.g., compressed CDs, digital multifunction DVDs, etc.), smart cards, and flash memory devices (e.g., cards, sticks, key drives, etc.).

[0069] A computer-readable medium may contain a propagated data signal containing computer program code, for example, on baseband or as part of a carrier wave. This propagated signal may take various forms, including electromagnetic, optical, and so on, or suitable combinations thereof. A computer-readable medium can be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, apparatus, or device to enable communication, propagation, or transmission of a program for use. The program code located on the computer-readable medium can be propagated through any suitable medium, including radio, cable, fiber optic cable, radio frequency signals, or similar media, or any combination of the above media.

[0070] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Any variations and modifications can be made by those skilled in the art without departing from the spirit and scope of the invention. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention, without departing from the scope of the invention, fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for cavitation simulation, characterized in that, include: Obtain macroscopic flow characteristic data of the reference cavitation flow; Based on a microscale bubble dynamics model, the microscale dynamic behavior of bubbles in the reference cavitation flow is simulated to obtain microscale bubble dynamic characteristics. Based on the macroscopic flow characteristic data and the microscopic bubble dynamics characteristics, training samples are constructed and a machine learning model is trained to establish a mapping relationship between macroscopic flow characteristics and microscopic bubble dynamics characteristics. The trained machine learning model is embedded into the macro-scale cavitation model, so that the macro-scale cavitation model is corrected according to the micro-scale bubble dynamics characteristics output by the machine learning model, thus obtaining a multi-scale cavitation model. The target cavitation flow is solved based on the multi-scale cavitation model to obtain the numerical simulation results of the target cavitation flow.

2. The cavitation simulation method as described in claim 1, characterized in that, The process of correcting the macroscale cavitation model based on the microscale bubble dynamics characteristics output by the machine learning model includes: The phase transition source term of the macroscopic cavitation model is modified using the microscopic bubble dynamics characteristics.

3. The cavitation simulation method as described in claim 2, characterized in that, The phase change source term includes an evaporation term characterizing bubble generation and a condensation term characterizing bubble collapse; The step of modifying the phase transition source term of the macroscopic cavitation model using the microscopic bubble dynamics characteristics includes: At least one of the evaporation and condensation terms is modified using the microscale bubble dynamics characteristics.

4. The cavitation simulation method as described in claim 1, characterized in that, In the process of solving the target cavitation flow in computational fluid dynamics, the macroscopic flow characteristics obtained from the current solution are used as the input of the machine learning model, and the macroscopic cavitation model is modified according to the microscopic bubble dynamic characteristics output by the machine learning model, so that the multi-scale cavitation model is updated as the target cavitation flow evolves.

5. The cavitation simulation method as described in claim 1, characterized in that, The macroscopic flow characteristic data includes the pressure and velocity of the reference cavitation flow; The acquisition of macroscopic flow characteristic data of the reference cavitation flow includes: The physical environment parameters around the bubbles in the reference cavitation flow are collected by pressure sensors, temperature sensors, and flow rate sensors. Experimental data of the reference cavitation flow were collected, and features were extracted from the experimental data to obtain the macroscopic flow characteristic data.

6. The cavitation simulation method as described in claim 5, characterized in that, Before feature extraction from the experimental data, the following steps are also included: The experimental data is cleaned, including removing noise, invalid values, and duplicate values, and smoothing or removing outliers.

7. The cavitation simulation method according to any one of claims 1-6, characterized in that, The machine learning model is a regression model or a neural network model; The neural network model includes convolutional neural networks or artificial neural networks.

8. The cavitation simulation method according to any one of claims 1-6, characterized in that, Before embedding the trained machine learning model into the macro-scale cavitation model, the method further includes: The accuracy of the trained machine learning model is validated using a validation dataset independent of the training samples.

9. The cavitation simulation method according to any one of claims 1-6, characterized in that, When solving the target cavitation flow in computational fluid dynamics, the mesh of the region where cavitation occurs in the target cavitation flow is refined. The mesh refinement meets the mesh refinement standard, which is the ratio of the volume of the refined fine mesh to the volume of the original coarse mesh.

10. A cavitation simulation system, characterized in that, include: The data acquisition module is used to acquire macroscopic flow characteristic data of the reference cavitation flow; The microscopic simulation module is used to simulate the microscopic dynamic behavior of bubbles in the reference cavitation flow based on a microscopic bubble dynamics model, and obtain microscopic bubble dynamic characteristics. The model training module is used to construct training samples and train a machine learning model based on the macroscopic flow characteristic data and the microscopic bubble dynamics characteristics, so as to establish a mapping relationship between the macroscopic flow characteristics and the microscopic bubble dynamics characteristics. The model building module is used to embed the trained machine learning model into the macro-scale cavitation model, so that the macro-scale cavitation model is corrected according to the micro-scale bubble dynamics characteristics output by the machine learning model, and a multi-scale cavitation model is obtained. The solver module is used to solve the target cavitation flow in computational fluid dynamics based on the multi-scale cavitation model, and obtain the numerical simulation results of the target cavitation flow.

11. A computer-readable storage medium, characterized in that, The system stores computer instructions, which, when executed by a processor, implement the cavitation simulation method as described in any one of claims 1-9.

Images