Turbulence model parameter correction method suitable for automobile aerodynamics simulation calculation
By optimizing the turbulence model parameters through Latin hypercube sampling, Kriging surrogate model, and multi-island genetic algorithm, the problem of insufficient correction of turbulence model parameters in automotive aerodynamics simulation is solved, thereby improving the accuracy and adaptability of simulation results.
Patent Information
- Application Number
- CN202511984924.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-02-10
AI Technical Summary
In existing automotive aerodynamics simulation calculations, the lack of correction methods for turbulence model parameters leads to poor correlation between simulation results and wind tunnel results, especially in complex flow fields with high Reynolds numbers, resulting in insufficient accuracy.
The Latin hypercube sampling method is used to generate parameter samples. The parameters of the turbulence model are optimized by combining the Kriging surrogate model and the multi-island genetic algorithm. The parameters are then iteratively optimized through wind tunnel results, and finally a machine learning model is constructed to predict the optimal turbulence parameters.
It improves the correlation and accuracy between simulation results and wind tunnel results, enhances the precision of simulation prediction, and adapts to the aerodynamic simulation needs of different vehicle models.
Smart Images
Figure CN121503335A_ABST
Abstract
Description
Technical Field
[0001] This manual relates to the field of computational simulation technology, and in particular to a method for correcting parameters of turbulence models applicable to automotive aerodynamics simulation calculations. Background Technology
[0002] While my country has achieved remarkable success in the development of new energy vehicles, range anxiety remains a major factor hindering their further growth. Air resistance experienced by a car at high speeds is proportional to the square of its speed. At 120 km / h, 80% of the resistance comes from air resistance, and this resistance increases further with speed. A current drawback of new energy vehicles is their low range achievement rate at high speeds; therefore, reducing the drag coefficient is crucial for improving range.
[0003] Current automotive aerodynamics development primarily relies on wind tunnel testing and computational fluid dynamics (CFD) simulation. Wind tunnel testing suffers from resource constraints, long model preparation cycles, and high costs. CFD simulation, on the other hand, offers abundant resources, does not depend on physical models (only a 3D digital model is needed), and boasts low costs and short preparation cycles. Traditional automotive aerodynamics simulations, due to the need to consider computational cycles and costs, often rely on Reynolds-mean-stress models (such as Realizable k-epsilon and shear stress transmission k-ω models) for steady-state simulations. However, classic turbulence models contain numerous parameters derived from theoretical derivations, assumptions, and calibration through fluid dynamics experiments on common scenarios (uniform shear flow, grid turbulence attenuation, etc.). While these models offer high accuracy for simulating simple flow fields at low Reynolds numbers, they exhibit significant errors when simulating complex flow fields in high-Reynolds-number, complex-shaped vehicles. In common commercial software, the default coefficients are often the empirical coefficients of the classic turbulence model. Most car manufacturers use the default coefficients when performing whole vehicle aerodynamic simulations because they lack correction methods. This results in poor correlation between the simulation results and the wind tunnel results, meaning that the simulation results are not accurate enough. Summary of the Invention
[0004] This manual provides a method for correcting turbulence model parameters applicable to automotive aerodynamic simulation calculations, in order to address the problem of inaccurate simulation results in existing technologies.
[0005] This manual adopts the following technical solution: a turbulence model parameter correction method applicable to automotive aerodynamic simulation calculations, including the following steps: S1: Obtain wind tunnel results for multiple sample vehicles and establish a three-dimensional geometric model for each sample vehicle; For the preset turbulence model, determine the corresponding parameters to be corrected; S2: Based on the parameters to be corrected and the preset turbulence model, multiple sets of parameter samples are generated using the Latin hypercube sampling method; S3: Under the parameter samples of each group, perform aerodynamic simulation calculations on the three-dimensional geometric models of each sample vehicle to obtain the simulation results of each parameter sample; S4: For each of the sample vehicles, based on the simulation results of each parameter sample and the corresponding wind tunnel results, a surrogate model is constructed and iterated. When the surrogate model reaches the preset termination iteration condition, the optimal turbulence model parameters of each of the sample vehicles are determined.
[0006] S5: Obtain the geometric feature parameters of each of the sample vehicles, and train the pre-set machine learning model based on the geometric feature parameters of each of the sample vehicles and the corresponding optimal turbulence model parameters to construct a prediction model; S6: Obtain the geometric feature parameters of the vehicle under test, and input the geometric feature parameters of the vehicle under test into the prediction model to predict the optimal turbulence model parameters of the vehicle under test.
[0007] Based on the above technical means, wind tunnel test results of multiple sample vehicles are obtained by combining wind tunnel tests. Furthermore, the test conditions for the parameters to be optimized in the turbulence model are expanded based on Latin hypercube sampling. The parameters to be optimized in the turbulence model are calibrated and optimized, which improves the consistency between simulation results and wind tunnel results. Compared with traditional simulation results, the correlation between simulation results and wind tunnel results is stronger, thereby improving the accuracy of simulation prediction results.
[0008] In this invention, by selecting multiple sample vehicles and extracting representative geometric feature parameters, the trained prediction model has strong generalization ability and can adapt to the aerodynamic simulation needs of different vehicle models.
[0009] Furthermore, the preset turbulence model includes the Realizable k-epsilon model and the SST k-ω model.
[0010] Based on the above technical means, in the specific implementation process, the Realizable k-epsilon model is robust in simulating large-scale separation and free shear flow, while the SST k-ω model is more advantageous in handling reverse pressure gradients, near-wall flow and transition flow. At the same time, selecting these two models for parameter correction can improve the simulation accuracy of different key flow characteristics of vehicles in complex flow fields.
[0011] Furthermore, the parameters to be corrected in the Realizable k-epsilon model include: turbulent viscosity coefficient, dissipation rate equation constant, turbulent kinetic energy Prandtl number, and dissipation rate Prandtl number.
[0012] Furthermore, the parameters to be corrected in the k-ω model of shear stress transmission include: turbulence coefficient, first turbulence Prandtl number, second turbulence Prandtl number, inner layer constant, outer layer constant, and correction factor.
[0013] Furthermore, step S4 specifically includes the following steps: S41: For each of the sample vehicles, a Kriging surrogate model is constructed with the parameter sample as input and the error between the corresponding simulation result and the corresponding wind tunnel result as output. S42: Based on the Kriging surrogate model, the multi-island genetic algorithm is used to optimize and iterate the parameters to be corrected, and the error between the simulation results and the wind tunnel results is obtained; S43: When the optimization iteration of the Kriging proxy model reaches the convergence condition, the parameter to be corrected is the optimal turbulence model parameter of the corresponding sample vehicle.
[0014] Based on the aforementioned technical means, the Kriging model can accurately predict the objective function, i.e., the error between the simulation results and the wind tunnel results, based on a limited number of initial simulation samples. This avoids the need for extensive and time-consuming direct computational fluid dynamics simulations during the optimization process. Furthermore, by combining the global search capability of the multi-island genetic algorithm, it can efficiently find the optimal solution in a high-dimensional parameter space, enhancing the algorithm's ability to escape local optima and approach the global optimum. This transforms a process that might otherwise require thousands of direct simulations into an intelligent optimization process centered on a surrogate model with controllable computational costs, thereby reducing the corresponding development costs.
[0015] Furthermore, the simulation results include the simulated drag coefficient and the simulated lift coefficient, and the wind tunnel results include the wind tunnel drag coefficient and the wind tunnel lift coefficient.
[0016] Furthermore, the convergence condition in S43 is: the absolute value of the error between the simulated drag coefficient and the corresponding wind tunnel drag coefficient is less than a preset value, and the absolute value of the error between the simulated lift coefficient and the corresponding wind tunnel lift coefficient is less than a preset value.
[0017] Based on the above technical means, by judging the error between the simulated drag coefficient and the corresponding wind tunnel drag coefficient, and judging the error between the simulated lift coefficient and the corresponding wind tunnel lift coefficient, that is, the convergence condition must simultaneously satisfy that the absolute values of both errors are less than the preset values, the one-sided behavior of sacrificing one coefficient in order to match one coefficient in the optimization process is prevented, and the optimization process is avoided from going on indefinitely or getting stuck in overfitting the training samples, so that the optimal parameters have better generalization ability.
[0018] Furthermore, the geometric feature parameters of the sample vehicle in S5 include: parameters representing the rear shape of the sample vehicle, parameters representing the chassis flatness of the sample vehicle, the rear upturn angle of the sample vehicle, the body length of the sample vehicle, and the ground clearance of the sample vehicle.
[0019] Furthermore, the parameter characterizing the rear shape of the sample vehicle is the angle between the rear windshield and the horizontal direction; the parameter characterizing the chassis flatness of the sample vehicle is the projected area coverage of the underbody protection plate.
[0020] Furthermore, the machine learning model pre-set in S5 is an ensemble learning model of the extreme gradient boosting algorithm.
[0021] Furthermore, the machine learning model pre-defined in S5 is determined to be an ensemble learning model of the extreme gradient boosting algorithm, which can efficiently learn the complex mapping relationship between vehicle styling features and optimal simulation parameters, thereby ensuring that the final prediction model has strong generalization ability.
[0022] The above-mentioned technical solutions adopted in this specification can achieve the following beneficial effects: 1. By combining wind tunnel tests of multiple sample vehicles to obtain corresponding wind tunnel results data for the sample vehicles, and by expanding the test conditions for the parameters to be optimized in the turbulence model based on Latin hypercube sampling, the parameters to be optimized in the turbulence model are calibrated and optimized, which improves the consistency between simulation results and wind tunnel results. Compared with traditional simulation results, the correlation between simulation results and wind tunnel results is stronger, thereby improving the accuracy of simulation prediction results.
[0023] 2. In this invention, by selecting multiple sample vehicles and extracting representative geometric feature parameters, the trained prediction model has a strong generalization ability and can adapt to the aerodynamic simulation requirements of different vehicle models. Attached Figure Description
[0024] The accompanying drawings, which are included to provide a further understanding of this specification and form part of this specification, illustrate exemplary embodiments and are used to explain this specification, but do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall process of this embodiment; Figure 2 This is a schematic diagram illustrating the process of constructing and iterating a proxy model in this embodiment.
[0025] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. To better illustrate this embodiment, some components in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings. The same or similar reference numerals correspond to the same or similar components. The terms describing positional relationships in the drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. Detailed Implementation
[0026] To make the objectives, technical solutions, and advantages of this specification clearer, the technical solutions of this specification will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments in this specification without creative effort are within the scope of protection of this application.
[0027] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention.
[0028] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0029] In the embodiments of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined with "first" and "second" may explicitly or implicitly include one or more of that feature.
[0030] In the embodiments of this application, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral part; it can be a direct connection or an indirect connection through an intermediate medium.
[0031] In embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0032] The technical solutions provided in the various embodiments of this specification are described in detail below with reference to the accompanying drawings.
[0033] like Figure 1 As shown, this embodiment provides a method for correcting turbulence model parameters suitable for automotive aerodynamic simulation calculations, including the following steps: S1: Obtain wind tunnel results for multiple sample vehicles and establish a three-dimensional geometric model for each sample vehicle; determine the corresponding parameters to be corrected for the preset turbulence model. In this preferred embodiment, representative models available on the market (including sedans, SUVs, and MPVs, and power types including pure electric, range-extended, plug-in hybrid, and gasoline) are selected for wind tunnel testing. The number of models is 100 (including 36 sedans, 42 SUVs, 22 MPVs, 47 pure electric vehicles, 18 range-extended vehicles, 11 plug-in hybrid vehicles, and 24 gasoline vehicles).
[0034] In other embodiments, different types of vehicles and the number of different types of vehicles can be selected.
[0035] In this embodiment, the simulation results include the simulated drag coefficient and the simulated lift coefficient.
[0036] In this embodiment, the preset turbulence models include the Realizable k-epsilon model and the Shear Stress Transport (SST k-ω) model. The Realizable k-epsilon model is robust in simulating large-scale separation and free shear flow, while the SST k-ω model is more advantageous in handling adverse pressure gradients, near-wall flow, and transition flow. By selecting these two models and modifying their parameters, the simulation accuracy of different key flow characteristics of vehicles in complex flow fields can be improved.
[0037] In this preferred embodiment, the parameters to be corrected in the Realizable k-epsilon model include: turbulent viscosity coefficient C. μ Dissipation rate equation constant C 2eThe turbulent kinetic energy Prandtl number Sigma_e and the dissipation rate Prandtl number Sigma_k; the turbulence coefficient kappa, the first turbulence Prandtl number Sigma_w1, the second turbulence Prandtl number Sigma_w2, the inner layer constant β1, the outer layer constant β2, and the correction factor β * .
[0038] In this embodiment, C μ This parameter is the turbulent viscosity ratio, which determines the relationship between turbulent viscosity, turbulent kinetic energy, and dissipation rate. It directly affects the dissipation and diffusion of turbulent energy and is the most crucial constant in the model; C 2e The parameter Σ_k controls the generation term of the turbulent dissipation rate, which is the rate at which turbulent kinetic energy is converted into thermal energy. This parameter has a significant impact on the simulation accuracy of complex flows such as flow separation and recirculation. Sigma_k and Sigma_ε are Prandtl numbers for turbulent kinetic energy and dissipation rate, respectively. Simply put, they represent the efficiency of turbulence in diffusion and dissipation processes, determining the spatial transfer and dissipation characteristics of turbulent energy. These parameters work together to allow the Realizable k-epsilon model to more accurately predict turbulent behavior when simulating complex flows such as boundary layers and recirculation.
[0039] In this embodiment, kappa: the von Kármán constant, determines the relationship between turbulent viscosity and wall distance, affecting the simulation accuracy of boundary layer flow; its standard value is typically 0.41. Sigma_w1 and Sigma_w2: Prandtl numbers of the specific dissipation rate ω, controlling the diffusion and dissipation characteristics of ω; their standard values are 2 and 1.168, respectively. β1 and β2: control the dissipation term of ω, affecting the rate of turbulent energy dissipation; their standard values are 0.075 and 0.0828, respectively. β*: a correction factor used to adjust the generation term of ω; its standard value is typically 0.09. These parameters work together to allow the SST k-ω model to perform well in both near-wall regions and free shear flows.
[0040] like Figure 1 As shown, in this embodiment, S2: Based on the parameters to be corrected and the preset turbulence model, multiple sets of parameter samples are generated using the Latin hypercube sampling method; In this preferred embodiment, 40 samples are generated for the four parameters in the Realizable k-epsilon turbulence model using the Latin hypercube sampling method, and 60 samples are generated for the six parameters in the SST k-ω turbulence model using the Latin hypercube sampling method.
[0041] like Figure 1As shown, in this embodiment, S3: Under each set of parameter samples, aerodynamic simulation calculations are performed on the three-dimensional geometric models of each sample vehicle to obtain the simulation results of each parameter sample.
[0042] In this embodiment, simulation calculations are performed on the sample vehicles using 40 sets of samples from the Realizable k-epsilon turbulence model and 60 sets of samples from the SST k-ω turbulence model. A total of 100 simulation cases are completed for each vehicle, and the simulated drag coefficient and simulated lift coefficient of each case are obtained (i.e., each vehicle will generate 40 sets of simulated drag coefficients and simulated lift coefficients from the Realizable k-epsilon turbulence model and 60 sets of simulated drag coefficients and simulated lift coefficients from the SST k-ω turbulence model). The simulation calculations are completed for 100 vehicles, and the total number of simulation cases will exceed 10,000.
[0043] like Figure 1 As shown, in this embodiment, S4: For each of the sample vehicles, based on the simulation results of each parameter sample and the corresponding wind tunnel results, a surrogate model is constructed and iterated. When the surrogate model reaches the preset termination iteration condition, the optimal turbulence model parameters of each of the sample vehicles are determined.
[0044] like Figure 2 As shown, in this embodiment, S4 specifically includes the following steps: S41: For each of the sample vehicles, a Kriging surrogate model is constructed using the parameter sample as input and the error between the corresponding simulation result and the corresponding wind tunnel result as output.
[0045] S42: Based on the Kriging surrogate model, the multi-island genetic algorithm is used to optimize and iterate the parameters to be corrected, and the error between the simulation results and the wind tunnel results is obtained.
[0046] S43: When the optimization iteration of the Kriging proxy model reaches the convergence condition, the parameter to be corrected is the optimal turbulence model parameter of the corresponding sample vehicle.
[0047] In this embodiment, the absolute value of the error between the simulated drag coefficient and the corresponding wind tunnel drag coefficient is less than a preset value, and the absolute value of the error between the simulated lift coefficient and the corresponding wind tunnel lift coefficient is less than a preset value.
[0048] In this embodiment, the Kriging model can accurately predict the objective function based on a limited number of initial simulation samples, i.e., the error between the simulation results and the wind tunnel results, thereby avoiding a large amount of time-consuming computational fluid dynamics direct simulation calculations during the optimization process. At the same time, it is combined with the global search capability of the multi-island genetic algorithm, which can efficiently find the optimal solution in the high-dimensional parameter space, enhancing the algorithm's ability to escape local optima and approach the global optimal solution. The process that may have required tens of thousands of direct simulations is transformed into an intelligent optimization process with controllable computational costs centered on the surrogate model, reducing the corresponding development costs.
[0049] In this embodiment, the preset value is preferably 1%.
[0050] like Figure 1 As shown, in this embodiment, S5: Obtain the geometric feature parameters of each sample vehicle, and train the pre-set machine learning model according to the geometric feature parameters of each sample vehicle and the corresponding optimal turbulence model parameters to construct a prediction model.
[0051] In this preferred embodiment, the pre-set machine learning model is an ensemble learning model of the extreme gradient boosting algorithm. Determining that the pre-set machine learning model in S5 is an ensemble learning model of the extreme gradient boosting algorithm enables efficient learning of the complex mapping relationship between vehicle styling features and optimal simulation parameters, thereby ensuring that the final constructed prediction model has strong generalization ability.
[0052] The study extracted distinctive features from the styling of these 100 vehicles, revealing that rear styling, chassis flatness, rear upturn angle, vehicle length, and ground clearance significantly impact the accuracy of drag coefficient and lift coefficient calculations. Currently, passenger vehicles on the market are categorized by rear styling into sloping rear (stepback sedans, fastback sedans) and boxy rear (SUVs, MPVs). These two types are quantified using the angle between the rear windshield and the horizontal direction. Based on powertrain type, vehicles can be divided into four categories: pure electric, range-extended, plug-in hybrid, and gasoline. The biggest difference between these four powertrain types lies in the flatness of the underbody. Pure electric vehicles use a full underbody protection plate to achieve a lower drag coefficient, resulting in an extremely flat bottom. Range-extended and plug-in hybrid vehicles use a hybrid powertrain, with both a flat battery pack and an exhaust system (the area below the exhaust pipe is not covered by a protection plate to consider heat dissipation). Therefore, the flatness of the underbody of range-extended and plug-in hybrid vehicles is inferior to that of pure electric vehicles. Gasoline vehicles do not have a power battery and are not sensitive to wind resistance. In some mid-to-low-end models, OEMs consider cost factors and use the least amount of protection plate, resulting in the worst flatness. To quantify flatness, the underbody protection plate coverage rate is used as a descriptive parameter (underbody protection plate coverage rate = projected area of flat protection plate in the Z direction / projected area of the whole vehicle in the Z direction * 100%). Based on daily development experience, the three parameters of rear overhang angle (the angle between the rear overhang and the horizontal direction), body length, and the height of the lowest point of the body (excluding the wheels) from the ground have a certain impact on the calculation accuracy. Therefore, the above five parameters are selected as typical vehicle styling features.
[0053] In this embodiment, the geometric feature parameters of the sample vehicle include: parameters representing the rear shape of the sample vehicle, parameters representing the chassis flatness of the sample vehicle, the rear upturn angle of the sample vehicle, the body length of the sample vehicle, and the ground clearance of the sample vehicle; wherein, the parameter representing the rear shape of the sample vehicle is the angle between the rear windshield of the sample vehicle and the horizontal direction; the parameter representing the chassis flatness of the sample vehicle is the projected area coverage of the bottom guard plate of the sample vehicle.
[0054] In other embodiments, other parameters can be selected as geometric feature parameters of the sample vehicle.
[0055] like Figure 1 As shown, in this embodiment, S6: Obtain the geometric feature parameters of the vehicle under test, and input the geometric feature parameters of the vehicle under test into the prediction model to predict the optimal turbulence model parameters of the vehicle under test.
[0056] In this embodiment, when characteristic parameters of vehicles other than the 100 vehicles under test are input, the prediction model automatically predicts the optimal turbulence model parameters for the vehicles under test and performs simulation calculations. Finally, simulation calculations were performed using 10 vehicles under test, and the results show that the simulation error of the vehicles under test is significantly improved compared to the traditional fixed-parameter simulation error.
[0057] In summary, this embodiment combines wind tunnel tests of multiple sample vehicles to obtain corresponding wind tunnel results data. Furthermore, it expands the test conditions for the parameters to be optimized in the turbulence model based on Latin hypercube sampling, calibrating and optimizing these parameters. This improves the consistency between simulation results and wind tunnel results. Compared to traditional simulation results, the correlation between simulation and wind tunnel results is stronger, thereby improving the accuracy of simulation predictions. By selecting multiple sample vehicles and extracting representative geometric feature parameters, the trained prediction model has strong generalization ability and can adapt to the aerodynamic simulation needs of different vehicle models.
[0058] The above embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention.
Claims
1. A method for correcting turbulence model parameters applicable to automotive aerodynamic simulation calculations, characterized in that, Includes the following steps: S1: Obtain wind tunnel results for multiple sample vehicles and establish a three-dimensional geometric model for each sample vehicle; For the preset turbulence model, determine the corresponding parameters to be corrected; S2: Based on the parameters to be corrected and the preset turbulence model, multiple sets of parameter samples are generated using the Latin hypercube sampling method; S3: Under the parameter samples of each group, perform aerodynamic simulation calculations on the three-dimensional geometric models of each sample vehicle to obtain the simulation results of each parameter sample; S4: For each of the sample vehicles, based on the simulation results of each parameter sample and the corresponding wind tunnel results, a surrogate model is constructed and iterated. When the surrogate model reaches the preset termination iteration condition, the optimal turbulence model parameters of each of the sample vehicles are determined. S5: Obtain the geometric feature parameters of each of the sample vehicles, and train the pre-set machine learning model based on the geometric feature parameters of each of the sample vehicles and the corresponding optimal turbulence model parameters to construct a prediction model; S6: Obtain the geometric feature parameters of the vehicle under test, and input the geometric feature parameters of the vehicle under test into the prediction model to predict the optimal turbulence model parameters of the vehicle under test.
2. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 1, characterized in that, The preset turbulence models include the Realizable k-epsilon model and the SST k-ω model.
3. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 2, characterized in that, The parameters to be corrected in the Realizable k-epsilon model include: turbulent viscosity coefficient, dissipation rate equation constant, turbulent kinetic energy Prandtl number, and dissipation rate Prandtl number.
4. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 2, characterized in that, The parameters to be corrected in the k-ω model of shear stress transmission include: turbulence coefficient, first turbulence Prandtl number, second turbulence Prandtl number, inner layer constant, outer layer constant, and correction factor.
5. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41: For each of the sample vehicles, a Kriging surrogate model is constructed with the parameter sample as input and the error between the corresponding simulation result and the corresponding wind tunnel result as output. S42: Based on the Kriging surrogate model, the multi-island genetic algorithm is used to optimize and iterate the parameters to be corrected, and the error between the simulation results and the wind tunnel results is obtained; S43: When the optimization iteration of the Kriging proxy model reaches the convergence condition, the parameter to be corrected is the optimal turbulence model parameter of the corresponding sample vehicle.
6. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 5, characterized in that, The simulation results include the simulated drag coefficient and the simulated lift coefficient, and the wind tunnel results include the wind tunnel drag coefficient and the wind tunnel lift coefficient.
7. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 6, characterized in that, The convergence condition in S43 is that the absolute value of the error between the simulated drag coefficient and the corresponding wind tunnel drag coefficient is less than a preset value, and the absolute value of the error between the simulated lift coefficient and the corresponding wind tunnel lift coefficient is less than a preset value.
8. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 1, characterized in that, The geometric feature parameters of the sample vehicle in S5 include: parameters representing the rear shape of the sample vehicle, parameters representing the chassis flatness of the sample vehicle, the rear upturn angle of the sample vehicle, the body length of the sample vehicle, and the ground clearance of the sample vehicle.
9. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 8, characterized in that, The parameter characterizing the rear profile of the sample vehicle is the angle between the rear windshield and the horizontal direction; the parameter characterizing the chassis flatness of the sample vehicle is the projected area coverage of the underbody protection plate.
10. The method for correcting turbulence model parameters for automotive aerodynamics simulation calculations according to claim 1, characterized in that, The machine learning model pre-set in S5 is an ensemble learning model of the extreme gradient boosting algorithm.