CN117171873B

CN117171873B - Vehicle aerodynamic optimization method and device and vehicle

Info

Publication number: CN117171873B
Application number: CN202311031715.9A
Authority: CN
Inventors: 苏永雷
Original assignee: Xiaomi Automobile Technology Co Ltd
Current assignee: Xiaomi Automobile Technology Co Ltd
Priority date: 2023-08-16
Filing date: 2023-08-16
Publication date: 2024-05-14
Anticipated expiration: 2043-08-16
Also published as: CN117171873A

Abstract

The present disclosure proposes a vehicle aerodynamic optimization method, device and vehicle, the method comprising: constructing a high-precision analysis model and a medium-precision analysis model of a vehicle, wherein the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle, and a low-precision analysis model is generated according to the medium-precision analysis model to determine a sample point set, wherein the sample point set comprises: the high-precision sample point is used for constructing a multi-credibility approximation model according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model and the low-precision analysis model and carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model, so that the obtained multi-credibility approximation model can give consideration to the prediction performance and the calculation cost in the aerodynamic analysis process of the vehicle, and the aerodynamic optimization effect on the vehicle can be effectively improved.

Description

Vehicle aerodynamic optimization method and device and vehicle

Technical Field

The disclosure relates to the technical field of automobiles, and in particular relates to a vehicle aerodynamic optimization method and device and a vehicle.

Background

Along with the continuous development of automobile technology, in the early stage of automobile product development, the aerodynamic development flow starts to gradually change from the traditional 'empirical design' to the 'integrated optimization' technical route, namely, the multiple sample automatic update and scheme optimization of the scheme are performed by comprehensively utilizing the parameterized deformation technology, the experimental design method, the approximate model method and the data optimization algorithm.

In the related art, high time costs are required to perform aerodynamic analysis of a vehicle, and an aerodynamic optimization effect for the vehicle is poor.

Disclosure of Invention

To overcome the problems in the related art, the present disclosure provides a vehicle aerodynamic optimization method, apparatus, vehicle, non-transitory computer-readable storage medium storing computer instructions, and computer program product.

An embodiment of a first aspect of the present disclosure provides a vehicle aerodynamic optimization method, including:

Constructing a high-precision analysis model and a medium-precision analysis model of a vehicle, wherein the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, and the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle;

Generating a low-precision analysis model according to the medium-precision analysis model;

Determining a sample point set, wherein the sample point set comprises: high-precision sample points;

constructing a multi-credibility approximation model according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model and the low-precision analysis model; and

And carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model.

An embodiment of a second aspect of the present disclosure provides a vehicle aerodynamic optimization device, including:

The system comprises a first model construction module, a second model construction module and a third model construction module, wherein the first model construction module is used for constructing a high-precision analysis model and a medium-precision analysis model of a vehicle, the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, and the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle;

the second model building module is used for generating a low-precision analysis model according to the medium-precision analysis model;

a determining module, configured to determine a sample point set, where the sample point set includes: high-precision sample points;

the third model construction module is used for constructing a multi-credibility approximation model according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model and the low-precision analysis model; and

And the optimization module is used for carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model.

An embodiment of a third aspect of the present disclosure provides a vehicle, including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to: a vehicle aerodynamic optimization method as proposed by an embodiment of the first aspect of the present disclosure is implemented.

An embodiment of a fourth aspect of the present disclosure proposes a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a vehicle aerodynamic optimization method as proposed by an embodiment of the first aspect of the present disclosure.

Embodiments of a fifth aspect of the present disclosure propose a computer program product, which when executed by a processor, performs a vehicle aerodynamic optimization method as proposed by embodiments of the first aspect of the present disclosure.

The vehicle aerodynamic optimization method provided by the embodiment of the disclosure can comprise the following beneficial effects: the method comprises the steps of constructing a high-precision analysis model and a medium-precision analysis model of a vehicle, wherein the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle, generating a low-precision analysis model according to the medium-precision analysis model, and determining a sample point set, wherein the sample point set comprises: the high-precision sample point is used for constructing a multi-credibility approximation model according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model and the low-precision analysis model and carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model, so that the obtained multi-credibility approximation model can give consideration to the prediction performance and the calculation cost in the aerodynamic analysis process of the vehicle, and the aerodynamic optimization effect on the vehicle can be effectively improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The foregoing and/or additional aspects and advantages of the present disclosure will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

FIG. 1 is a flow chart illustrating a vehicle aerodynamic optimization method according to some embodiments of the present disclosure;

FIG. 2 is a flow chart illustrating a vehicle aerodynamic optimization method according to some embodiments of the present disclosure;

FIG. 3 is a schematic diagram of sample point distribution proposed in accordance with the present disclosure;

FIG. 4 is a flow chart illustrating a vehicle aerodynamic optimization method according to some embodiments of the present disclosure;

FIG. 5 is a flow chart illustrating a vehicle aerodynamic optimization method according to some embodiments of the present disclosure;

FIG. 6 is a flow chart of an aerodynamic design optimization method based on a multi-credibility approximation model and multi-precision data fusion, according to the present disclosure;

FIG. 7 is a block diagram of a vehicle aerodynamic optimization device shown in accordance with some embodiments of the present disclosure;

fig. 8 is a block diagram of a vehicle, according to an exemplary embodiment.

Detailed Description

Reference will now be made in detail to some embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. Various changes, modifications, and equivalents of the methods, devices, and/or systems described herein will become apparent after an understanding of the present disclosure. For example, the order of operations described herein is merely an example and is not limited to those set forth herein, but may be altered as will become apparent after an understanding of the disclosure, except where necessary to perform the operations in a particular order. In addition, descriptions of features known in the art may be omitted for the sake of clarity and conciseness.

The implementations described below in some examples of the disclosure are not representative of all implementations consistent with the disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.

FIG. 1 is a flow chart illustrating a vehicle aerodynamic optimization method, as shown in FIG. 1, according to some embodiments of the present disclosure, including the steps of:

s101: and constructing a high-precision analysis model and a medium-precision analysis model of the vehicle, wherein the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, and the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle.

The high-precision analysis model is a model which is constructed based on wind tunnel experimental data of a vehicle and is suitable for high-precision analysis.

In the embodiment of the disclosure, when a high-precision analysis model of the vehicle is constructed, wind tunnel experiments can be performed on the vehicle, and the model vehicle is used for collecting wind tunnel experiment data in a controlled wind field. These wind tunnel experimental data may include aerodynamic forces (e.g., lift and drag) of the vehicle at different speeds, pressure distribution, flow field information, and the like. Preprocessing the acquired data, including data cleaning, denoising, interpolation and the like. Ensuring the accuracy and usability of the data. Useful features are extracted from the raw data for describing the aerodynamic behaviour of the vehicle in wind tunnel experiments. This may include lift coefficient, drag coefficient, side force coefficient, pressure coefficient distribution, etc. By analyzing wind tunnel experimental data, a mathematical model can be established to describe the aerodynamic behavior of the vehicle. This typically involves expressing the motion, mechanics and flow behavior of the vehicle using laws of physics and mathematical equations. The mathematical model used may be, without limitation, a aerodynamic balance equation, a Navier-Stokes equation, a boundary layer equation, and the like.

The medium-precision analysis model is a simulation model which is configured based on the geometric data of the vehicle and is suitable for medium-precision aerodynamic analysis. The geometric data may be, for example, information such as the vehicle body outline, the dimensions, the wheel positions, and the like. The geometric data may be obtained from documents provided by the vehicle manufacturer, CAD models, or actual measurements.

In embodiments of the present disclosure, in constructing a medium-precision analytical model of a vehicle, a three-dimensional vehicle model may be created from geometric data of the vehicle using Computer Aided Design (CAD) software or similar tools. The model is ensured to accurately reflect the appearance and the size of the vehicle, and is divided into small grid cells to form a discretized geometrical body so as to calculate hydrodynamic parameters in simulation.

It can be appreciated that in the aerodynamic analysis process, there is generally a high requirement on the accuracy of the model, and in the embodiments of the present disclosure, when constructing a high-accuracy analysis model and a medium-accuracy analysis model of a vehicle, data support of the high-accuracy analysis dimension and the medium-accuracy analysis dimension may be provided for the subsequent construction of a multi-credibility approximation model.

S102: and generating a low-precision analysis model according to the medium-precision analysis model.

The low-precision analysis model is a model suitable for low-precision aerodynamic analysis.

It will be appreciated that the low-precision analytical model may introduce errors during aerodynamic analysis that are greater than the medium-precision analytical model described above, but the low-precision analytical model is typically more efficient in analysis, which can effectively reduce the overall time cost of the aerodynamic analysis.

In the embodiment of the disclosure, when the low-precision analysis model is generated according to the medium-precision analysis model, an empirical formula and a simplifying assumption can be used to replace detailed physical modeling, so that the low-precision analysis model is generated, or the precision of data can be reduced by reducing the sampling rate of a medium-precision data set or reducing the number of data points, and then the low-precision analysis model is trained by using the downsampled data, so that the low-precision analysis model is not limited.

S103: determining a sample point set, wherein the sample point set comprises: high precision sample points.

It will be appreciated that in performing aerodynamic analysis, the vehicle model is typically divided into a number of small surface elements or grids in order to accurately describe the stress conditions of the vehicle. The four corner points of each surface element or grid are typically considered sample points. By measuring parameters such as pressure and flow rate at these sample points, air flow information on the surface of the vehicle can be obtained.

The high-precision sample point is a sample point suitable for high-precision analysis. A sample point set refers to a set of a plurality of high precision sample points.

It will be appreciated that too few sample points may fail to capture detail and local variations, resulting in poor model accuracy; too many sample points increase the amount of computation and computation time. Thus, in embodiments of the present disclosure, a set of sample points may be selected with an appropriate number and distribution of sample points according to particular needs and model complexity.

In some embodiments, in determining the sample point set, a three-dimensional scanner or measurement device may be used to scan and measure the actual vehicle to obtain its geometry. By this method, high-precision sample points, i.e. accurate surface data of the model, can be obtained.

In other embodiments, computer Aided Design (CAD) software may also be used to manually model the geometry of the vehicle when determining the sample point set. CAD software provides a series of tools and functions that can draw three-dimensional models of automobiles. In the modeling process, the density and distribution of the sample points can be customized to meet the requirements, and the generated CAD model is the high-precision sample points.

Of course, any other possible method may be used to determine the high-precision sample points of the vehicle to determine the sample point set, without limitation.

S104: and constructing a multi-credibility approximation model according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model and the low-precision analysis model.

The multi-credibility approximation model is a method for establishing and utilizing models with different precision levels to approximate an objective function. In the multi-credibility approximation model, a plurality of models with different precision levels can be used for replacing the evaluation of the original objective function so as to improve the efficiency of model construction and optimization.

It will be appreciated that, in general, when the model accuracy is higher, more computing resources and time are typically required for training and evaluation, while a lower accuracy model may produce results faster. The multi-credibility approximation model realizes more effective model establishment and optimization under limited resources by combining the advantages of different precision models.

In the embodiment of the disclosure, when the multi-credibility approximation model is constructed according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model and the low-precision analysis model, the multi-credibility approximation model can be based on an interpolation method, a machine learning algorithm, gaussian process regression or the like, and the multi-credibility approximation model is not limited to the above.

In the embodiment of the disclosure, when the multi-credibility approximation model is constructed according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model and the low-precision analysis model, the suitability of the obtained multi-credibility approximation model for the vehicle can be effectively improved.

S105: and carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model.

That is, in the embodiment of the disclosure, when the aerodynamic optimization is performed on the vehicle, the contradictory relation between the analysis efficiency and the analysis accuracy can be balanced based on the multi-credibility approximation model, so that the time cost consumed in the aerodynamic analysis process of the vehicle is reduced, and the personalized requirement of the aerodynamic optimization effect on the vehicle in the application scene is met.

In the embodiment of the disclosure, when the wind resistance performance of the vehicle is required to be high, the wind resistance suite can be increased or decreased, such as chassis packaging is reinforced, a low-rolling-resistance tire is adopted, and the wind resistance reducing effect is improved.

In the embodiment of the disclosure, the establishment of the automobile windage whole automobile model and the calculation domain can be realized based on the following steps:

(1) The domain settings are calculated. In order to more accurately align with the wind tunnel test, a five-band system considering the ground effect is established, and the calculation domain is set as follows: the distance from the head to the entrance of the calculation domain is 4 times of the length of the vehicle, the distance from the tail to the exit of the calculation domain is 6 times of the length of the vehicle, and the total length of the calculation domain is 11 times of the length of the vehicle; the distance from the left side and the right side of the vehicle to the left side and the right side of the calculation domain is 4 times of the vehicle width, and the total width of the calculation domain is 9 times of the vehicle width; the roof is 5 times the vehicle height to the top of the calculation domain, and the total height of the calculation domain is 6 times the vehicle height. The section of the incoming flow direction right in front of the whole automobile is an inlet boundary and is set as a speed inlet boundary. The right rear section is an outlet boundary, set as a pressure outlet boundary. The left side, the right side and the section right above the whole vehicle are set as sliding boundary conditions, the five-belt system is set as non-sliding ground, and the running speed is equal to the vehicle speed. While taking into account wheel rotation, a constant tangential velocity of Local Rotaion Rate is employed.

(2) The grid is discrete. The size of the surface mesh is 2-8 mm. In order to more accurately consider the influence of the vehicle surface on the fluid flow, boundary layer meshing is performed on the surface, and the 1 st layer has a thickness of 0.02mm and a 10-layer boundary layer with a total thickness of 8mm. While the encryption process of the parts 4, 8 and 16mm is performed in the fluid separation area.

(3) And solving the setting. SST k-omega turbulence model in RANS method is adopted, standard wall function, second order windward format and SIMPLE calculation algorithm are adopted. According to the existing vehicle model, accuracy verification is carried out, starCCM + simulation results are compared with tests, the maximum difference is within 3counts, the whole vehicle simulation model and analysis are effective, the method has high accuracy, and the method can be expanded to other vehicle models and is regarded as a high-reliability analysis model and standard analysis working condition.

FIG. 2 is a flow chart illustrating a vehicle aerodynamic optimization method according to some embodiments of the present disclosure.

As shown in fig. 2, the vehicle aerodynamic optimization method includes the steps of:

S201: and constructing a high-precision analysis model and a medium-precision analysis model of the vehicle, wherein the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, and the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle.

The description of S201 may be specifically referred to the above embodiments, and will not be repeated here.

S202: adjusting the size of the contour and/or the modeling related to the vehicle architecture in the middle precision analysis model to obtain a parameterized model, wherein the parameterized model comprises design variables which are not related to the local features of the vehicle, and the parameterized model comprises: the model portion corresponding to the local feature, and/or the first face mesh size, and/or the total number of first mesh units.

The parameterized model is a model obtained by adjusting the size of a contour and/or a model related to a vehicle architecture in the centering accuracy analysis model.

Wherein, the design variable refers to a variable which can be adjusted in the parameterized model.

The local feature may be a detail feature such as a rearview mirror or radar of the vehicle.

The first surface mesh size refers to the mesh size of the parameterized model in an initial state.

The first grid cell total number refers to the grid cell total number of the parameterized model in an initial state.

For example, in the embodiment of the disclosure, when the contour and/or the dimension of the model related to the vehicle architecture in the centering accuracy analysis model are adjusted to obtain the parameterized model, the overall modeling contour dimension adjustment may be performed in the conceptual modeling design stage. Such as wheelbase, wheel track, vehicle height, minimum ground clearance, approach angle, departure angle, front windshield inclination angle, nominal back angle and the like, adopting ANSA software VolumeMorph parameterization function, establishing MORPHBOX containing each design characteristic, adjusting MORPHBOX the position of a control point or a reference edge to enable MORPHERBOC to be attached to the modeling contour as much as possible, setting design variables through MORPHBOX to define functions of movement, rotation, scaling and the like, and realizing parameterization automatic adjustment of the whole modeling contour.

It will be appreciated that medium-precision analytical models may be relatively complex, with internal expressions or structures not directly dependent on parameters. For more convenient optimization, analysis, or application, the model may be adapted in embodiments of the present disclosure to be represented in a parameterized manner. That is, a set of parameters is introduced, and by adjusting the values of these parameters, the behavior, shape, or output of the model can be changed. These parameters may control the weights, biases, coefficients, etc. of the model. By converting the model into a parameterized form, the model optimization, parameter tuning and inference process can be performed more efficiently. Parameterized models may also be better integrated with other models or algorithms and are generally easier to understand and interpret.

S203: and generating a low-precision analysis model according to the parameterized model.

That is, in the embodiments of the present disclosure, after the dimensions of the contours and/or shapes related to the vehicle architecture in the middle-precision analysis model are adjusted to obtain the parameterized model, the low-precision analysis model may be quickly and accurately generated based on the parameterized model.

Optionally, in some embodiments, when the low-precision analysis model is generated according to the parameterized model, the first surface mesh size of the parameterized model may be increased to the second surface mesh size, and the total number of the first mesh units is reduced to the total number of the second mesh units, so as to obtain the low-precision analysis model, where the low-precision analysis model does not support refinement processing on a model portion corresponding to the local feature, and thus, model simplification of the parameterized model may be achieved by increasing the surface mesh size and reducing to the total number of mesh units, so that reliability of the obtained low-precision analysis model is effectively improved.

The second surface mesh size may be any surface mesh size larger than the first surface mesh size.

The second total number of grid cells may be any total number of grid cells less than the first total number of grid cells.

For example, in the embodiment of the disclosure, when the low-precision analysis model is generated according to the parameterized model, the large outline and the modeling size related to the whole vehicle structure may be adjusted, the design variable does not relate to the local feature, at this time, the refinement processing of the local feature such as the rearview mirror, the laser radar and the like is canceled, the surface mesh size is coarsened to 8-14mm, and the total number of the model mesh units is reduced as the corresponding low-precision model.

That is, after the medium-precision analysis model of the vehicle is built in the embodiment of the present disclosure, the dimensions of the contour and/or the model related to the vehicle architecture in the medium-precision analysis model may be adjusted to obtain a parameterized model, where the parameterized model includes design variables that are not related to the local features of the vehicle, and the parameterized model includes: and generating a low-precision analysis model according to the parameterized model by the model part corresponding to the local features and/or the first surface grid size and/or the total number of the first grid units, so that the flexibility of the low-precision analysis model generating process can be effectively improved by parameterizing the medium-precision analysis model.

In the embodiment of the disclosure, after the medium-precision analysis model and the low-precision analysis model are obtained, the multi-precision simulation analysis working condition fusion can be realized based on the following steps:

(1) And establishing a high/low precision analysis model and an associated flow of analysis working conditions. Integrating a high/low precision parameterized analysis model and an analysis working condition, taking a predefined automatic parameterized design feature as an input design variable, associating the same design variable with the high/low precision analysis model at the same time, automatically calling a StarCCM + solver to solve the high/low precision parameterized analysis model, and automatically editing the solved result to obtain a calculated wind resistance coefficient C _d, wherein the wind resistance coefficient C _d is a performance index concerned by optimization analysis.

(2) A multi-sample self-driven parallel iterative method. Aerodynamic computation is time-consuming, requiring the invocation of high-performance computing resources; the invention develops StarCCM + and high-performance cluster interfaces, adopts OpenPBS parallel resource management system, realizes that the working conditions can automatically submit parallel calculation in batches, and automatically returns the calculation result to the corresponding sample point. Performing simulation debugging based on the built association flow, performing analysis and calculation of an initial design state sample, and comparing an automatic simulation result of the integrated flow with a manually submitted calculation result to ensure that the results of the two are consistent; and further performing analysis and calculation on a plurality of samples to ensure that the association flow can drive the automatic iterative updating of the design variables.

S204: and generating a plurality of low-precision sample points through a Latin hypercube experiment design algorithm.

That is, in the embodiment of the present disclosure, the value range of each parameter may be divided into equal subintervals based on the latin hypercube design algorithm, and then one sample point is selected from each subinterval to ensure better coverage in each parameter dimension. In this process, a larger distance between sample points is ensured, thereby obtaining better sample representation capability.

Thus, uniformly distributed low precision sample points can be selected in the multidimensional parameter space to provide a comprehensive, uniform, representative parameter sample.

S205: the medium-precision sample points are selected from a plurality of low-precision sample points by a D-optimization criterion, wherein the total number of low-precision sample points is greater than the total number of medium-precision sample points.

It will be appreciated that the D-best criterion is a criterion for sample point selection. In statistical experimental design, the D-best criterion aims at selecting the set of sample points with the best design performance to provide the most accurate information about parameter estimation or model prediction.

In the disclosed embodiments, when a medium-precision sample point is selected from a plurality of low-precision sample points by a D-optimization criterion, the sample point may be selected by minimizing a variance of parameter estimation or a variance of model prediction. The variance is an index for measuring uncertainty of a parameter estimation or model prediction result, so that the uncertainty of the parameter estimation or model prediction can be reduced as much as possible by adopting a D-optimal criterion through designing a combination of sample points, and the quality of experiments or modeling is improved.

For example, as shown in fig. 3, fig. 3 is a schematic diagram of sample point distribution proposed according to the present disclosure. Wherein, the small hollow circles are low-precision sample points. The small circle containing the cross is the middle-precision sample point.

S206: high-precision sample points are selected from a plurality of medium-precision sample points by a D-optimal criterion, wherein the total number of medium-precision sample points is greater than the total number of high-precision sample points.

In the embodiment of the disclosure, the high-precision sample points are selected from the plurality of middle-precision sample points through the D-optimal criterion, so that the re-selection of the plurality of middle-precision sample points can be realized, and the practicability of the obtained high-precision sample points is ensured.

That is, in the embodiment of the present disclosure, after the low-precision analysis model is generated, a plurality of low-precision sample points may be generated by a latin hypercube design algorithm, a medium-precision sample point may be selected from the plurality of low-precision sample points by a D-optimal criterion, wherein the total number of low-precision sample points is greater than the total number of medium-precision sample points, and a high-precision sample point may be selected from the plurality of medium-precision sample points by a D-optimal criterion, wherein the total number of medium-precision sample points is greater than the total number of high-precision sample points, thereby effectively improving the practicality of the obtained high-precision sample points based on the latin hypercube design algorithm and the D-optimal criterion.

S207: and establishing a first scale function between the high-precision analysis model and the medium-precision analysis model and a second scale function between the high-precision analysis model and the low-precision analysis model according to the high-precision sample points.

Where the scaling function, which may be a simple linear function or a more complex nonlinear function, may normalize the responses of the different models to the same scale so that they may be directly compared. Different models may have different response ranges and magnitudes, which can be mapped to the same probability distribution by a scaling function, ensuring that their comparisons are comparable. The first scaling function may map the high precision data to a range corresponding to the medium precision data. While the second scaling function may map the high precision data to a range corresponding to the low precision data.

In the disclosed embodiments, when a first scale function between a high-precision analysis model and a medium-precision analysis model and a second scale function between the high-precision analysis model and a low-precision analysis model are established from high-precision sample points, high-precision data can be converted into a representation of medium-precision data or low-precision data by the first scale function and the second scale function for use in analysis or modeling.

S208: and constructing an initial multi-credibility approximation model according to the first scale function, the second scale function, the high-precision sample points, the medium-precision analysis model and the low-precision analysis model.

The initial multi-credibility approximation model is a multi-credibility approximation model in an initial state constructed based on a first scale function, a second scale function, a high-precision sample point, a medium-precision analysis model and a low-precision analysis model.

That is, in the embodiments of the present disclosure, after a first scale function between a high-precision analysis model and a medium-precision analysis model, and a second scale function between the high-precision analysis model and a low-precision analysis model are established from high-precision sample points, an initial multi-credibility approximation model may be constructed from the first scale function, the second scale function, the high-precision sample points, the medium-precision analysis model, and the low-precision analysis model.

S209: and performing target processing on the initial multi-credibility approximation model to obtain a multi-credibility approximation model.

The target processing may refer to a model adjustment method configured in advance for the initial multi-credibility approximation model.

That is, in the embodiment of the disclosure, after the high-precision sample point is determined, a first scale function between the high-precision analysis model and the middle-precision analysis model and a second scale function between the high-precision analysis model and the low-precision analysis model may be established according to the high-precision sample point, and an initial multi-reliability approximation model may be constructed according to the first scale function, the second scale function, the high-precision sample point, the middle-precision analysis model and the low-precision analysis model, and the initial multi-reliability approximation model may be subjected to target processing to obtain the multi-reliability approximation model, so that a data mapping relationship between different precision analysis models may be accurately indicated based on the scale function, thereby effectively improving the prediction performance of the obtained multi-reliability approximation model.

S210: and carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model.

The description of S210 may be specifically referred to the above embodiments, and will not be repeated here.

In an embodiment of the present disclosure, a parameterized model is obtained by adjusting a contour and/or a dimension of a model related to a vehicle architecture in a middle-precision analysis model, where a design variable included in the parameterized model is not related to a local feature of the vehicle, and the parameterized model includes: and generating a low-precision analysis model according to the parameterized model by the model part corresponding to the local features and/or the first surface grid size and/or the total number of the first grid units, so that the flexibility of the low-precision analysis model generating process can be effectively improved by parameterizing the medium-precision analysis model. The first grid size of the parameterized model is increased to the second grid size, and the total number of the first grid cells is reduced to the total number of the second grid cells, so that a low-precision analysis model is obtained, wherein the low-precision analysis model does not support refinement processing of a model part corresponding to local features, and therefore model simplification of the parameterized model can be achieved by increasing the surface grid size and reducing the surface grid size to the total number of the grid cells, and reliability of the obtained low-precision analysis model is effectively improved. Generating a plurality of low-precision sample points through a Latin hypercube test design algorithm, and selecting middle-precision sample points from the plurality of low-precision sample points through a D-optimal criterion, wherein the total number of the low-precision sample points is larger than the total number of the middle-precision sample points, and selecting high-precision sample points from the plurality of middle-precision sample points through the D-optimal criterion, wherein the total number of the middle-precision sample points is larger than the total number of the high-precision sample points, so that the practicability of the obtained high-precision sample points can be effectively improved based on the Latin hypercube test design algorithm and the D-optimal criterion. By establishing a first scale function between the high-precision analysis model and the medium-precision analysis model according to the high-precision sample points and a second scale function between the high-precision analysis model and the low-precision analysis model, an initial multi-credibility approximation model is established according to the first scale function, the second scale function, the high-precision sample points, the medium-precision analysis model and the low-precision analysis model, and target processing is carried out on the initial multi-credibility approximation model to obtain a multi-credibility approximation model, therefore, the data mapping relation between different precision analysis models can be accurately indicated based on the scale functions, and the prediction performance of the obtained multi-credibility approximation model is effectively improved.

Fig. 4 is a flow chart illustrating a vehicle aerodynamic optimization method according to some embodiments of the present disclosure.

As shown in fig. 4, the vehicle aerodynamic optimization method includes the steps of:

s401: and constructing a high-precision analysis model and a medium-precision analysis model of the vehicle, wherein the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, and the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle.

The description of S401 may be specifically referred to the above embodiments, and will not be repeated here.

S402: adjusting a first modeling characteristic of the vehicle in the centering accuracy analysis model to obtain a parameterized model, wherein the parameterized model comprises: the model portion corresponding to the local feature of the vehicle, and/or the first mesh size corresponding to the model portion of the second modeling feature, and/or the total number of third mesh cells, the first modeling feature and the second modeling feature being different.

The first molding feature may be referred to as a vehicle molding feature that has a greater impact on aerodynamic analysis. The second modeling feature may then refer to a vehicle modeling feature that has less impact on aerodynamic analysis.

The third grid cell total number refers to the grid cell total number of the parameterized model obtained after the first modeling characteristic of the vehicle in the centering accuracy analysis model is adjusted.

For example, in the embodiment of the disclosure, when the first modeling feature of the vehicle in the centering accuracy analysis model is adjusted, the adjustment of the key modeling feature in the modeling scheme design and development stage may be performed when the parameterized model is obtained. Key structural areas with complex curved surface characteristics such as a big tail wing, a duck wing, a side wing, a rearview mirror, a front lip, a front bumper and the like are parameterized by ANSA software SurfaceMorph, MORPHBOX containing various design characteristics is established, MORPHBOX is adjusted to enable modeling contours to be attached, and the functions of movement, rotation, scaling and the like are defined by setting design variables through MORPHBOX, so that parameterization automatic adjustment of the key modeling contours is realized.

That is, in the embodiment of the present disclosure, after the medium-precision analysis model of the vehicle is constructed, the first molding feature of the vehicle in the medium-precision analysis model may be parameterized, so that each parameter in the model may be adjusted to obtain the low-precision analysis model.

S403: and generating a low-precision analysis model according to the parameterized model.

Optionally, in some embodiments, when the low-precision analysis model is generated according to the parameterized model, the first mesh size of the parameterized model may be increased to the second mesh size, and the total number of the third mesh units may be reduced to the total number of the fourth mesh units, so as to obtain the low-precision analysis model, where a ratio value between the second mesh size and the first mesh size is a preset ratio value, and the low-precision analysis model does not support refinement processing on a model portion corresponding to the local feature, so that relevance between the obtained low-precision analysis model and the medium-precision analysis model may be effectively improved.

The fourth total number of grid cells may refer to any total number of grid cells less than the third total number of grid cells.

The preset proportion value can be flexibly configured according to the application scene, for example, can be 2 times, and is not limited to the preset proportion value.

For example, in the embodiment of the disclosure, when the low-precision analysis model is generated according to the parameterized model, the key modeling feature (the first modeling feature) may be adjusted, and then the mesh size of the model part other than the key feature is increased by 2 times, the refinement processing of the local feature which does not participate in the optimization is cancelled, and the total number of the mesh units of the model is reduced as the corresponding low-precision analysis model.

That is, in the embodiment of the present disclosure, after the medium-precision analysis model of the vehicle is constructed, the first molding feature of the vehicle in the medium-precision analysis model may be adjusted to obtain a parameterized model, where the parameterized model includes: the model part corresponding to the local feature of the vehicle, and/or the first grid size corresponding to the model part of the second modeling feature, and/or the total number of the third grid units, the first modeling feature and the second modeling feature are different, and a low-precision analysis model is generated according to the parameterized model, so that the generation efficiency of the low-precision analysis model can be effectively improved while the prediction performance of the obtained low-precision analysis model is ensured.

S404: and generating a plurality of low-precision sample points through a Latin hypercube experiment design algorithm.

S405: the medium-precision sample points are selected from a plurality of low-precision sample points by a D-optimization criterion, wherein the total number of low-precision sample points is greater than the total number of medium-precision sample points.

S406: high-precision sample points are selected from a plurality of medium-precision sample points by a D-optimal criterion, wherein the total number of medium-precision sample points is greater than the total number of high-precision sample points.

S407: and establishing a first scale function between the high-precision analysis model and the medium-precision analysis model and a second scale function between the high-precision analysis model and the low-precision analysis model according to the high-precision sample points.

S408: and constructing an initial multi-credibility approximation model according to the first scale function, the second scale function, the high-precision sample points, the medium-precision analysis model and the low-precision analysis model.

S409: and performing target processing on the initial multi-credibility approximation model to obtain a multi-credibility approximation model.

S410: and carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model.

The descriptions of S404-S410 may be specifically referred to the above embodiments, and are not repeated here.

In an embodiment of the present disclosure, a parameterized model is obtained by adjusting a first modeling feature of a vehicle in a centering accuracy analysis model, where the parameterized model includes: the model part corresponding to the local feature of the vehicle, and/or the first grid size corresponding to the model part of the second modeling feature, and/or the total number of the third grid units, the first modeling feature and the second modeling feature are different, and a low-precision analysis model is generated according to the parameterized model, so that the generation efficiency of the low-precision analysis model can be effectively improved while the prediction performance of the obtained low-precision analysis model is ensured. The first grid size of the parameterized model is increased to the second grid size, and the total number of the third grid cells is reduced to the total number of the fourth grid cells, so that a low-precision analysis model is obtained, wherein a ratio value between the second grid size and the first grid size is a preset ratio value, and the low-precision analysis model does not support refinement processing of a model part corresponding to local features, so that relevance of the obtained low-precision analysis model and the medium-precision analysis model can be effectively improved.

Fig. 5 is a flow chart illustrating a vehicle aerodynamic optimization method according to some embodiments of the present disclosure.

As shown in fig. 5, the vehicle aerodynamic optimization method includes the steps of:

S501: and constructing a high-precision analysis model and a medium-precision analysis model of the vehicle, wherein the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, and the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle.

S502: and generating a low-precision analysis model according to the medium-precision analysis model.

S503: determining a sample point set, wherein the sample point set comprises: high precision sample points.

The descriptions of S501-S503 may be specifically referred to the above embodiments, and are not repeated here.

S504: a first output response of the high-precision analytical model at the high-precision sample point is determined.

The first output response refers to a prediction result or output of the high-precision analysis model at a high-precision sample point.

In embodiments of the present disclosure, when determining a first output response of a high-precision analytical model at a high-precision sample point, reference information of an output dimension of the high-precision analytical model may be provided for a subsequent determination of a scale function.

S505: a second output response of the medium precision analytical model at the high precision sample point is determined.

The second output response refers to the prediction result or output of the middle-precision analysis model at the high-precision sample point.

In embodiments of the present disclosure, when determining a first output response of a medium precision analytical model at a high precision sample point, reference information of the output dimension of the medium precision analytical model may be provided for a subsequent determination of a scale function.

S506: a third output response of the low-precision analytical model at the high-precision sample point is determined.

The third output response refers to the prediction result or output of the low-precision analysis model at the high-precision sample point.

In embodiments of the present disclosure, when determining a third output response of the low-precision analytical model at the high-precision sample point, reference information of the output dimension of the low-precision analytical model may be provided for a subsequent determination of the scaling function.

S507: a first scaling function is determined based on the first output response, the second output response, and the high precision sample points.

That is, in embodiments of the present disclosure, after determining the first output response/second output response of the high/medium precision analytical model at the high precision sample point, a first scale function may be constructed from the first output response, the second output response, and the high precision sample point.

For example, in the multiple credibility approximation modeling method (ADDITIVE SCALING function based multi-fidelity surrogate modeling, AS-MFS) based on additive scale, high-precision sample pointsOutput responses of the high/medium precision analysis model are/>, respectivelyAnd/>The difference between the two is expressed as: Wherein/> High-precision sample point/>, which is a high-precision analytical modelScale factors at. The set of scale factors δ _m＝{δ₁,δ₂,……,δ_Nh is obtained based on the high-precision sample points, and the added scale function δ _m (x), namely the first scale function, is constructed as input and output, and a Kriging model method can be selected.

S508: a second scaling function is determined based on the first output response, the third output response, and the high precision sample points.

For example, in the multiple credibility approximation modeling method (ADDITIVE SCALING function based multi-fidelity surrogate modeling, AS-MFS) based on additive scale, high-precision sample pointsThe output response of the high/low precision analytical model is/>, respectivelyAnd/>The difference between the two is expressed as: Wherein/> High-precision sample point/>, which is a high-precision analytical modelScale factors at. The set of scale factors delta _l＝{δ₁,δ₂,……,δ_Nh is obtained based on the high-precision sample points, and the added scale function delta _l (x), namely the second scale function, is constructed as input and output, and a Kriging model method can be selected.

That is, in the embodiment of the present disclosure, after the sample point set is determined, a first output response of the high-precision analysis model at the high-precision sample point may be determined, a second output response of the medium-precision analysis model at the high-precision sample point may be determined, a third output response of the low-precision analysis model at the high-precision sample point may be determined, a first scale function may be determined according to the first output response, the second output response and the high-precision sample point, and a second scale function may be determined according to the first output response, the third output response and the high-precision sample point, thereby effectively improving the difference of the prediction results of the obtained scale function for different precision analysis models, and thus effectively improving the indication effect of the scale function in the process of constructing the initial multi-reliability approximation model.

S509: an initial multi-reliability approximation model is constructed from the first scale function, the second output response, the third output response, the first reference coefficient and the second reference coefficient, wherein the first reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the medium-precision analysis model, and the second reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the low-precision analysis model.

Optionally, in some embodiments, when the initial multi-reliability approximation model is constructed according to the first scale function, the second output response, the third output response, the first reference coefficient and the second reference coefficient, the second output response and the first reference coefficient may be multiplied, and the multiplied result and the first scale function are summed to obtain a first summed result, the third output response and the second reference coefficient are multiplied, and the multiplied result and the second scale function are summed to obtain a second summed result, and the first summed result and the second summed result are summed, and the summed result is used as the initial multi-reliability approximation model, thereby, a fusion process of the scale function and the low-precision analysis model may be implemented, and thus, the output flexibility of the obtained initial multi-reliability approximation model is effectively improved.

For example, the multi-credibility approximation model in the embodiments of the present disclosure may be:

f_mf(x)＝ρ_lf^l(x)+δ_l(x)+ρ_mf^m(x)+δ_m(x)

Wherein ρ _l is a linear constant coefficient (i.e. the second reference coefficient) between the high/low precision analysis model responses, and f ^l (x) is the high precision sample point of the low precision analysis model An output response (i.e., the third output response described above). ρ _m is the linear constant coefficient between the high/medium precision analytical model responses (i.e., the first reference coefficient described above), and f ^m (x) is the high precision sample point/>, of the medium precision analytical modelOutput response (i.e. the second output response)

That is, in the embodiment of the present disclosure, after the first scale function and the second scale function are determined, the initial multi-reliability approximation model may be constructed according to the first scale function, the second output response, the third output response, the first reference coefficient and the second reference coefficient, wherein the first reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the medium-precision analysis model, and the second reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the low-precision analysis model, whereby the reliability of the initial multi-reliability approximation model construction process may be effectively improved.

S510: and performing global optimization on the initial multi-credibility approximation model based on a preset optimization algorithm to obtain an initial solution.

The preset optimization algorithm is an optimization algorithm configured for an initial multi-credibility approximation model in advance. For example, a genetic algorithm, a particle swarm algorithm, or the like in the global optimization algorithm may be used, and this is not a limitation.

The initial solution may be an initial global optimal solution obtained by performing global optimization on the initial multi-credibility approximation model based on a preset optimization algorithm.

In the embodiment of the disclosure, when the initial multi-credibility approximation model is globally optimized based on the preset optimization algorithm to obtain an initial solution, the obtained initial solution can provide a reference point or a reference solution for the subsequent multi-objective optimization.

S511: a plurality of target sample points are determined in a neighborhood of the initial solution, wherein each target sample point has a corresponding objective function.

The target sample point is a sample point selected in the neighborhood of the initial solution.

Wherein, the objective function can be used to indicate the performance index of the corresponding target sample point in the global optimization process.

In the embodiment of the present disclosure, the plurality of target sample points may be all composed of sample points suitable for high-precision analysis, or may be composed of a plurality of sample point combinations suitable for different-precision analysis, without limitation.

S512: and determining a scale function value corresponding to the target sample point according to a preset scale function.

The preset scale function refers to a scale function configured for a model fusion process in advance. For example, it may be an addition scale function, a multiplication scale function, an exponential scale function, or the like, or may be a mixed scale function composed of a plurality of types of scale functions, without limitation.

The scale function value is an index for measuring the quality or the quality degree of the solution. The scaling function is typically mapped onto a scaling value by calculating the distance, coverage, envelope, etc. of the solution based on the relative relationship between the solutions in the solution set. A smaller scale function value indicates a better quality solution, closer to the real Pareto front.

In the embodiment of the disclosure, when the scale function value corresponding to the target sample point is determined according to the preset scale function, the obtained scale function value can provide reliable reference information for the subsequent model fitting process.

S513: and performing model fitting based on the target function and the scale function value corresponding to the target sample point to obtain a multi-credibility approximation model.

In the embodiment of the disclosure, when model fitting is performed based on the objective function and the scale function value corresponding to the objective sample point to obtain the multi-credibility approximation model, gaussian process regression (Gaussian Process Regression) may be used. Gaussian process regression is a non-parametric Bayesian regression method that can be used to build a mapping between objective function values and scale function values and provide an uncertainty estimate of the predicted outcome.

The following is a reference step for model fitting using gaussian process regression:

(1) The scale function value and the objective function value of the sampling point are collected as a training data set.

(2) And carrying out normalization processing on the scale function value and the objective function value, and mapping the scale function value and the objective function value into the same range so as to improve the fitting effect of the model.

(3) Training data was fitted using a gaussian process regression algorithm. The algorithm is based on bayesian theory, treats the objective function as a random process, and assumes that it obeys a gaussian distribution.

(4) And according to the fitting result, a prediction result of the objective function value and a corresponding confidence interval can be obtained. The confidence interval represents the degree of uncertainty for the predicted outcome.

(5) The performance of the model can be evaluated by cross-validation or the like, and necessary adjustments and optimizations can be made.

(6) When the model meets the requirements, the model can be used for predicting the new scale function value to obtain a corresponding objective function value, so as to obtain a multi-credibility approximation model.

That is, in the embodiment of the disclosure, after an initial multi-credibility approximation model is constructed, global optimization may be performed on the initial multi-credibility approximation model based on a preset optimization algorithm to obtain an initial solution, and a plurality of target sample points are determined in a neighborhood of the initial solution, where the target sample points have corresponding target functions, a scale function value corresponding to the target sample points is determined according to a preset scale function, and model fitting is performed based on the target functions and the scale function values corresponding to the target sample points to obtain the multi-credibility approximation model, so that the prediction performance of the obtained multi-credibility approximation model may be improved to a greater extent.

For example, in embodiments of the present disclosure, the multi-confidence approximation model may be derived based on the following steps:

(1) Determining an initial Pareto front (i.e., a neighborhood of the initial solution described above): some sort of multi-objective optimization algorithm (e.g., genetic algorithm, particle swarm optimization, etc.) is used to search for and determine the initial Pareto front of the problem.

(2) Selecting high-precision sample points: a sample point with high accuracy is selected from the initial Pareto front as a new sample point (i.e., the target sample point described above). The selection of high-precision sample points may be determined based on the magnitude of the objective function value or other evaluation index.

(3) Calculating the weight value of the objective function: for each high-precision sample point, calculating a corresponding objective function weight value. The weight value may be determined based on the specific requirements of the problem and domain knowledge. For example, the weight values may be calculated using normalized objective function values, or based on a comparison between objective functions.

(4) Addition scale correction: and multiplying the objective function value of the high-precision sample point by the corresponding weight value by using an addition scale function to obtain a corrected objective function value. Specifically, for a correction value of a certain objective function, the following formula is used:

correction value=objective function value×weight value

Thus, the corrected objective function value can be obtained, and the contribution of the objective function value corresponding to the sample point in the reconstructed multi-credibility approximation model can be represented.

(5) Reconstructing a multi-credibility approximation model: and taking the corrected objective function value and the input variable corresponding to the high-precision sample point as a new sample point for reconstructing the multi-credibility approximation model. The sample points may be fitted using appropriate regression or interpolation methods to obtain a more accurate multi-confidence approximation model.

In the embodiment of the disclosure, after the multi-credibility approximate model is obtained, the multi-credibility approximate model can be optimized to obtain a Pareto optimal solution, an implementable mode is determined according to the overall requirement of the modeling body, the implementable mode is engineered by a combined modeling and exterior design department, calculation and verification are carried out based on high-credibility simulation physics, performance comparison of a scheme before optimization is carried out, and if wind resistance performance of the engineering scheme is improved equally, optimization is effective.

S514: and carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model.

The description of S514 may be specifically referred to the above embodiments, and will not be repeated here.

In the embodiment of the disclosure, the first output response of the high-precision analysis model at the high-precision sample point is determined, the second output response of the medium-precision analysis model at the high-precision sample point is determined, the third output response of the low-precision analysis model at the high-precision sample point is determined, the first scale function is determined according to the first output response, the second output response and the high-precision sample point, and the second scale function is determined according to the first output response, the third output response and the high-precision sample point, so that the difference of the prediction results of the obtained scale function on different precision analysis models can be effectively improved, and the indication effect of the scale function in the process of constructing the initial multi-credibility approximate model is effectively improved. By constructing the initial multi-reliability approximation model from the first scale function, the second output response, the third output response, the first reference coefficient, and the second reference coefficient, wherein the first reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the medium-precision analysis model, and the second reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the low-precision analysis model, the reliability of the initial multi-reliability approximation model construction process can be effectively improved. The second output response is multiplied by the first reference coefficient, the multiplied result is summed with the first scale function to obtain a first summed result, the third output response is multiplied by the second reference coefficient, the multiplied result is summed with the second scale function to obtain a second summed result, the first summed result is summed with the second summed result, and the summed result is used as an initial multi-credibility approximation model, so that fusion processing of the scale function and the low-precision analysis model can be realized, and the output flexibility of the obtained initial multi-credibility approximation model is effectively improved. The initial multi-credibility approximation model is globally optimized based on a preset optimization algorithm to obtain an initial solution, a plurality of target sample points are determined in the neighborhood of the initial solution, wherein the target sample points have corresponding target functions, the scale function values corresponding to the target sample points are determined according to the preset scale functions, model fitting is performed based on the target functions and the scale function values corresponding to the target sample points to obtain the multi-credibility approximation model, and therefore the prediction performance of the obtained multi-credibility approximation model can be improved to a greater extent.

For example, as shown in fig. 6, fig. 6 is a schematic flow chart of an aerodynamic design optimization method based on a multi-credibility approximation model and multi-precision data fusion according to the present disclosure, wherein the flow steps can be seen from the description of the above embodiments.

According to the method, a low-precision model is established through the precision model in the whole vehicle, unified parameterized design variables are shared, a high/medium/low-precision sample point nested experimental design method is utilized, a multi-credibility approximate modeling method based on an addition scale function is adopted, the scale function and the low-precision analysis model are fused to construct a variable initial credibility approximate model, the middle-precision sample point is selected as a new sample point at the Pareto optimal front, the multi-credibility approximate model is reconstructed, and iterative optimization is carried out, so that an effective low-wind resistance engineering scheme is obtained. The optimization method disclosed by the invention is developed in a layered manner, and can be used for considering the optimization efficiency and the prediction precision.

For example, when the aerodynamic optimization is performed by the method based on the high-precision analysis model, if the Kriging model is constructed by 120 high-precision sample points, the consumed calculation time is 240h, and the corresponding optimized vehicle windage coefficient is 0.215C _d. When the aerodynamic optimization is performed based on the method for fusing the high/medium/low precision analysis models provided by the disclosure, the selected sample combination may be: 5 wind tunnel experiment high-precision samples, 20 medium-precision samples, 100 low-precision samples and 5 dynamic iteration samples, a Kriging model is selected as a scale function to construct a multi-credibility approximation model for aerodynamic analysis and optimization, the consumed calculation time is 80h, and the corresponding optimized vehicle wind resistance coefficient is 0.190C _d. Therefore, the vehicle aerodynamic optimization method based on the disclosure can greatly improve the aerodynamic optimization effect on the vehicle while reducing the calculation time.

Fig. 7 is a block diagram of a vehicle aerodynamic optimization device shown according to some embodiments of the present disclosure. Referring to fig. 7, the vehicle aerodynamic optimization device 70 includes: a first model building module 701, a second model building module 702, a determination module 703, a third model building module 704, and an optimization module 705.

A first model building module 701, configured to build a high-precision analysis model and a medium-precision analysis model of the vehicle, where the high-precision analysis model is a vehicle analysis model built based on wind tunnel experimental data of the vehicle, and the medium-precision analysis model is a vehicle analysis model built based on geometric data of the vehicle;

a second model building module 702, configured to generate a low-precision analysis model according to the medium-precision analysis model;

A determining module 703, configured to determine a sample point set, where the sample point set includes: high-precision sample points;

a third model construction module 704, configured to construct a multi-credibility approximation model according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model, and the low-precision analysis model; and

An optimization module 705 for aerodynamically optimizing the vehicle according to the multi-confidence approximation model.

In some embodiments of the present disclosure, the second model building module 702 is specifically configured to:

adjusting the size of the contour and/or the modeling related to the vehicle architecture in the middle precision analysis model to obtain a parameterized model, wherein the parameterized model comprises design variables which are not related to the local features of the vehicle, and the parameterized model comprises: model portions corresponding to local features, and/or first face mesh sizes, and/or a total number of first mesh units;

and generating a low-precision analysis model according to the parameterized model.

In some embodiments of the present disclosure, the second model building module 702 is further configured to:

And increasing the first grid size of the parameterized model to the second grid size, and reducing the total number of the first grid cells to the total number of the second grid cells to obtain a low-precision analysis model, wherein the low-precision analysis model does not support refinement processing of model parts corresponding to the local features.

Adjusting a first modeling characteristic of the vehicle in the centering accuracy analysis model to obtain a parameterized model, wherein the parameterized model comprises: a model portion corresponding to a local feature of the vehicle, and/or a first grid size corresponding to a model portion of a second modeling feature, and/or a total number of third grid cells, the first modeling feature and the second modeling feature being different;

and increasing the first grid size of the parameterized model to a second grid size, and reducing the total number of the third grid cells to the total number of the fourth grid cells to obtain a low-precision analysis model, wherein the ratio value between the second grid size and the first grid size is a preset ratio value, and the low-precision analysis model does not support refinement processing of model parts corresponding to local features.

In some embodiments of the present disclosure, the determining module 703 is specifically configured to:

Generating a plurality of low-precision sample points through a Latin hypercube experiment design algorithm;

Selecting medium-precision sample points from a plurality of low-precision sample points by a D-optimal criterion, wherein the total number of the low-precision sample points is larger than that of the medium-precision sample points;

high-precision sample points are selected from a plurality of medium-precision sample points by a D-optimal criterion, wherein the total number of medium-precision sample points is greater than the total number of high-precision sample points.

In some embodiments of the present disclosure, the third model building module 704 is specifically configured to:

according to the high-precision sample points, a first scale function between the high-precision analysis model and the medium-precision analysis model and a second scale function between the high-precision analysis model and the low-precision analysis model are established;

Constructing an initial multi-credibility approximation model according to the first scale function, the second scale function, the high-precision sample point, the medium-precision analysis model and the low-precision analysis model;

and performing target processing on the initial multi-credibility approximation model to obtain a multi-credibility approximation model.

In some embodiments of the present disclosure, the third model building module 704 is further configured to:

Determining a first output response of the high-precision analytical model at the high-precision sample point;

determining a second output response of the medium-precision analytical model at the high-precision sample point;

determining a third output response of the low-precision analytical model at the high-precision sample point;

determining a first scale function according to the first output response, the second output response and the high-precision sample point;

A second scaling function is determined based on the first output response, the third output response, and the high precision sample points.

An initial multi-reliability approximation model is constructed from the first scale function, the second output response, the third output response, the first reference coefficient and the second reference coefficient, wherein the first reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the medium-precision analysis model, and the second reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the low-precision analysis model.

multiplying the second output response by the first reference coefficient, and summing the multiplied result and the first scale function to obtain a first summed result;

Multiplying the third output response by the second reference coefficient, and summing the multiplied result and the second scale function to obtain a second summed result;

and adding the first adding result and the second adding result, and taking the added result as an initial multi-credibility approximation model.

Performing global optimization on the initial multi-credibility approximation model based on a preset optimization algorithm to obtain an initial solution;

determining a plurality of target sample points in a neighborhood of the initial solution, wherein the target sample points have corresponding target functions;

Determining a scale function value corresponding to a target sample point according to a preset scale function;

And performing model fitting based on the target function and the scale function value corresponding to the target sample point to obtain a multi-credibility approximation model.

The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.

In this embodiment, a high-precision analysis model and a medium-precision analysis model of a vehicle are constructed, wherein the high-precision analysis model is a vehicle analysis model constructed based on wind tunnel experimental data of the vehicle, the medium-precision analysis model is a vehicle analysis model constructed based on geometric data of the vehicle, a low-precision analysis model is generated according to the medium-precision analysis model, and a sample point set is determined, and the sample point set includes: the high-precision sample point is used for constructing a multi-credibility approximation model according to the high-precision sample point, the high-precision analysis model, the medium-precision analysis model and the low-precision analysis model and carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model, so that the obtained multi-credibility approximation model can give consideration to the prediction performance and the calculation cost in the aerodynamic analysis process of the vehicle, and the aerodynamic optimization effect on the vehicle can be effectively improved.

Fig. 8 is a block diagram of a vehicle, according to an exemplary embodiment. For example, vehicle 800 may be a hybrid vehicle, but may also be a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other type of vehicle. Vehicle 800 may be an autonomous vehicle, a semi-autonomous vehicle, or a non-autonomous vehicle.

Referring to fig. 8, a vehicle 800 may include various subsystems, such as an infotainment system 810, a perception system 820, a decision control system 830, a drive system 840, and a computing platform 850. Vehicle 800 may also include more or fewer subsystems, and each subsystem may include multiple components. In addition, interconnections between each subsystem and between each component of the vehicle 800 may be achieved by wired or wireless means.

In some embodiments, infotainment system 810 may include a communication system, an entertainment system, a navigation system, and so forth.

The sensing system 820 may include several sensors for sensing information of the environment surrounding the vehicle 800. For example, sensing system 820 may include a global positioning system (which may be a GPS system, or may be a beidou system or other positioning system), an inertial measurement unit (inertial measurement unit, IMU), a lidar, millimeter wave radar, an ultrasonic radar, and a camera device.

Decision control system 830 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

The drive system 840 may include components that provide powered motion to the vehicle 800. In one embodiment, the drive system 840 may include an engine, an energy source, a transmission, and wheels. The engine may be one or a combination of an internal combustion engine, an electric motor, an air compression engine. The engine is capable of converting energy provided by the energy source into mechanical energy.

Some or all of the functions of vehicle 800 are controlled by computing platform 850. Computing platform 850 may include at least one processor 851 and memory 852, and processor 851 may execute instructions 853 stored in memory 852.

The processor 851 may be any conventional processor, such as a commercially available CPU. The processor may also include, for example, an image processor (Graphic Process Unit, GPU), a field programmable gate array (Field Programmable GATE ARRAY, FPGA), a System On Chip (SOC), an Application SPECIFIC INTEGRATED Circuit (ASIC), or a combination thereof.

The memory 852 may be implemented by any type or combination of volatile or nonvolatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disk.

In addition to instructions 853, memory 852 may store data such as road maps, route information, vehicle location, direction, speed, etc. The data stored by memory 852 may be used by computing platform 850.

In an embodiment of the present disclosure, the processor 851 may execute instructions 853 to perform all or part of the steps of the vehicle aerodynamic optimization method described above.

The present disclosure also provides a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the steps of the vehicle aerodynamic optimization method provided by the present disclosure.

Furthermore, the word "exemplary" is used herein to mean serving as an example, instance, illustration. Any aspect or design described herein as "exemplary" is not necessarily to be construed as advantageous over other aspects or designs. Rather, the use of the word exemplary is intended to present concepts in a concrete fashion. As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless specified otherwise, or clear from context, "X application a or B" is intended to mean any one of the natural inclusive permutations. I.e. if X applies a; x is applied with B; or both X applications a and B, "X application a or B" is satisfied under any of the foregoing examples. In addition, the articles "a" and "an" as used in this application and the appended claims are generally understood to mean "one or more" unless specified otherwise or clear from context to be directed to a singular form.

Also, although the disclosure has been shown and described with respect to one or more implementations, equivalent alterations and modifications will occur to others skilled in the art upon the reading and understanding of this specification and the annexed drawings. The present disclosure includes all such modifications and alterations and is limited only by the scope of the claims. In particular regard to the various functions performed by the above described components (e.g., elements, resources, etc.), the terms used to describe such components are intended to correspond, unless otherwise indicated, to any component which performs the specified function of the described component (which is functionally equivalent), even though not structurally equivalent to the disclosed structure. In addition, while a particular feature of the disclosure may have been disclosed with respect to only one of several implementations, such feature may be combined with one or more other features of the other implementations as may be desired and advantageous for any given or particular application. Furthermore, to the extent that the terms "includes," including, "" has, "" having, "or variants thereof are used in either the detailed description or the claims, such terms are intended to be inclusive in a manner similar to the term" comprising.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method of vehicle aerodynamic optimization, the method comprising:

Performing aerodynamic optimization on the vehicle according to the multi-credibility approximation model, wherein the multi-credibility approximation model is used for balancing the contradictory relation between analysis efficiency and analysis accuracy;

Wherein said constructing a multi-reliability approximation model from said high-precision sample points, said high-precision analytical model, said medium-precision analytical model, and said low-precision analytical model comprises:

determining a second output response of the medium precision analytical model at the high precision sample point;

Determining a first scaling function based on the first output response, the second output response, and the high precision sample point;

Determining a second scaling function based on the first output response, the third output response, and the high precision sample point;

Constructing an initial multi-credibility approximation model according to the first scale function, the second scale function, the high-precision sample points, the medium-precision analysis model and the low-precision analysis model;

And performing target processing on the initial multi-credibility approximation model to obtain the multi-credibility approximation model.

2. The method of claim 1, wherein generating a low-precision analytical model from the medium-precision analytical model comprises:

And adjusting the size of the contour and/or the modeling related to the vehicle architecture in the medium-precision analysis model to obtain a parameterized model, wherein the parameterized model comprises design variables which are not related to the local features of the vehicle, and the parameterized model comprises: a model portion corresponding to the local feature, and/or a first face mesh size, and/or a first mesh unit total number;

and generating the low-precision analysis model according to the parameterized model.

3. The method of claim 2, wherein the generating the low-precision analytical model from the parameterized model comprises:

Increasing the first grid size of the parameterized model to a second grid size, and reducing the total number of the first grid cells to the total number of the second grid cells to obtain the low-precision analysis model, wherein the low-precision analysis model does not support refinement processing of model parts corresponding to the local features; the first surface grid size refers to the grid size of the parameterized model in an initial state, the second surface grid size is larger than the first surface grid size, the first grid unit total number refers to the grid unit total number of the parameterized model in the initial state, and the second grid unit total number is smaller than the first grid unit total number.

4. The method of claim 1, wherein generating a low-precision analytical model from the medium-precision analytical model comprises:

Adjusting a first modeling characteristic of the vehicle in the medium-precision analysis model to obtain a parameterized model, wherein the parameterized model comprises: a model portion corresponding to a local feature of the vehicle, and/or a first mesh size corresponding to a model portion of a second modeling feature, and/or a total number of third mesh cells, the first modeling feature and the second modeling feature being different; the first modeling feature refers to a vehicle modeling feature that has a large influence on aerodynamic analysis, and the second modeling feature refers to a vehicle modeling feature that has a small influence on aerodynamic analysis; the third grid cell total number refers to the grid cell total number of the parameterized model; and generating the low-precision analysis model according to the parameterized model.

5. The method of claim 4, wherein generating the low-precision analytical model from the parameterized model comprises:

And increasing the first grid size of the parameterized model to a second grid size, and reducing the total number of the third grid cells to the total number of the fourth grid cells to obtain the low-precision analysis model, wherein a ratio value between the second grid size and the first grid size is a preset ratio value, the low-precision analysis model does not support refinement processing of a model part corresponding to the local feature, the second grid size is larger than the first grid size, and the total number of the third grid cells is larger than the total number of the fourth grid cells.

6. The method of claim 1, wherein the determining the set of sample points comprises:

Selecting middle-precision sample points from a plurality of the low-precision sample points by a D-optimal criterion, wherein the total number of the low-precision sample points is larger than the total number of the middle-precision sample points;

And selecting the high-precision sample points from a plurality of medium-precision sample points through the D-optimal criterion, wherein the total number of the medium-precision sample points is larger than the total number of the high-precision sample points.

7. The method of claim 1, wherein said constructing an initial multi-confidence approximation model from said first scale function, said second scale function, said high precision sample points, said medium precision analytical model, and said low precision analytical model comprises:

Constructing the initial multi-reliability approximation model according to the first scale function, the second output response, the third output response, a first reference coefficient and a second reference coefficient, wherein the first reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the medium-precision analysis model, and the second reference coefficient is a linear constant coefficient between the response of the high-precision analysis model and the response of the low-precision analysis model.

8. The method of claim 7, wherein said constructing the initial multi-confidence approximation model from the first scale function, the second output response, the third output response, a first reference coefficient, and a second reference coefficient comprises:

And adding the first adding result and the second adding result, and taking the added result as the initial multi-credibility approximation model.

9. The method according to any one of claims 1-8, wherein said performing a target process on said initial multi-reliability approximation model to obtain said multi-reliability approximation model comprises:

determining a plurality of target sample points in the neighborhood of the initial solution, wherein the target sample points have corresponding target functions;

Determining a scale function value corresponding to the target sample point according to a preset scale function;

And performing model fitting based on the target function and the scale function value corresponding to the target sample point to obtain the multi-credibility approximation model.

10. A vehicle aerodynamic optimization device, the device comprising:

The optimization module is used for carrying out aerodynamic optimization on the vehicle according to the multi-credibility approximation model, and the multi-credibility approximation model is used for balancing the contradictory relation between analysis efficiency and analysis accuracy;

The third model construction module is specifically configured to:

11. A vehicle, characterized by comprising:

A processor;

a memory for storing processor-executable instructions; wherein the processor is configured to: the steps of carrying out the method of any one of claims 1-9.