Aerodynamic evaluation method, device and electronic equipment for an automobile exterior

Abstract

The present disclosure provides a method, an apparatus, and an electronic device for aerodynamic evaluation of an automobile exterior. By acquiring three-dimensional point cloud data describing the automobile exterior and generating a forward index and a reverse index corresponding to the point cloud data according to a preset space-filling curve; determining a spatial similarity matrix corresponding to the point cloud data, and extracting large-scale features corresponding to the point cloud data according to the spatial similarity matrix; serializing the point cloud data according to the forward index, extracting small-scale features corresponding to the point cloud data, and returning the point cloud data to the original order according to the reverse index; fusing the large-scale features and the small-scale features to generate point cloud fusion features, and determining the surface flow field of the automobile according to the point cloud fusion features. It can effectively extract large-scale and small-scale structural features in the automobile exterior, accurately predict the surface pressure distribution and wall shear stress distribution of the automobile exterior and calculate the drag coefficient under lower computational complexity and video memory occupancy, and improve the evaluation efficiency and prediction accuracy.

Description

Technical Field

[0001] The present disclosure relates to the technical field of automotive design, and in particular, to a method, device, and electronic device for aerodynamic evaluation of an automotive exterior Background Technique

[0002] With the rapid development of the new energy vehicle industry, the energy utilization efficiency of vehicles has become an important factor affecting vehicle performance and market competitiveness. Among them, the drag coefficient, as an important factor when a vehicle is driving at high speed, is closely related to the aerodynamic performance of the vehicle exterior design. Optimizing the vehicle exterior to reduce wind resistance can not only significantly increase the driving range but also effectively reduce energy consumption, which is in line with the industry development trend of energy conservation and environmental protection. Therefore, how to quickly and accurately evaluate the aerodynamic performance of an automotive exterior has become a key technical issue in the research and development of new energy vehicles.

[0003] Traditionally, the aerodynamic performance evaluation of an automotive exterior relies on computational fluid dynamics simulation methods. By means of numerical calculation, the interaction between the fluid and the solid surface can be simulated, and the surface pressure, wall shear stress distribution, and drag coefficient of the automotive exterior can be accurately calculated. However, the demand for computing resources is extremely large, especially when dealing with complex three-dimensional geometric models, and the calculation process often takes several days or even weeks. This high time cost limits the efficiency of automotive design optimization and is not conducive to the actual needs of rapid iterative design. Summary of the Invention

[0004] The embodiments of the present disclosure at least provide a method, device, and electronic device for aerodynamic evaluation of an automotive exterior. By introducing the spatial aggregation technology and serialization technology of point clouds, the large-scale and small-scale structural features in the automotive exterior can be effectively extracted, and the surface pressure, wall shear stress distribution, and drag coefficient of the automotive exterior can be accurately predicted with lower computational complexity and video memory occupancy, thereby significantly improving the evaluation efficiency and prediction accuracy.

[0005] The embodiments of the present disclosure provide a method for aerodynamic evaluation of an automotive exterior, including:

[0006] Obtaining three-dimensional point cloud data describing the automotive exterior, and generating a forward index and a reverse index corresponding to the point cloud data according to a preset space-filling curve;

[0007] Determining a spatial similarity matrix corresponding to the point cloud data, and extracting large-scale features corresponding to the point cloud data according to the spatial similarity matrix;

[0008] Serializing the point cloud data according to the forward index, extracting small-scale features corresponding to the point cloud data, and returning the point cloud data to the original order according to the reverse index;

[0009] Fuse the large-scale features and the small-scale features to generate point cloud fusion features, and determine the vehicle surface flow field according to the point cloud fusion features.

[0010] In an alternative embodiment, generate a forward index and a reverse index corresponding to the point cloud data according to a preset space filling curve, specifically including:

[0011] Divide the three-dimensional space into multiple cubic blocks;

[0012] Sort the cubic blocks according to the preset space filling curve, and the preset space filling curve at least includes a Z-order curve, a Hilbert curve, a Tran Z-order curve, and a Tran Hilbert curve;

[0013] Assign corresponding indexes to each of the cubic blocks, and arrange the point cloud data belonging to the same cubic block in ascending order of the indexes to determine the forward index and the reverse index for serializing the point cloud data.

[0014] In an alternative embodiment, determine a spatial similarity matrix corresponding to the point cloud data, and extract large-scale features corresponding to the point cloud data according to the spatial similarity matrix, specifically including:

[0015] Encode the point cloud data into high-dimensional point cloud basic features through a linear transformation;

[0016] Perform a point-by-point linear transformation on the point cloud basic features, calculate the corresponding spatial similarity weights using a global Softmax function, and integrate all the spatial similarity weights into the spatial similarity matrix;

[0017] Convert the point cloud data into multiple spatial slices, and determine the spatial aggregation features between the spatial slices according to the spatial similarity matrix;

[0018] Process the spatial aggregation features through a self-attention mechanism to determine the large-scale features.

[0019] In an alternative embodiment, serialize the point cloud data according to the forward index, and extract small-scale features corresponding to the point cloud data, specifically including:

[0020] Sort the point cloud data according to the forward index to generate the serialized point cloud data;

[0021] Divide the serialized point cloud data into multiple point cloud groups, and extract local features of each point cloud group through a self-attention mechanism to determine the small-scale features corresponding to the point cloud group.

[0022] In an alternative embodiment, returning the point cloud data to the original order according to the reverse index specifically includes:

[0023] Connecting the point cloud groups to form the point cloud data in the original shape;

[0024] Sorting the point cloud data formed by connecting the point cloud groups according to the reverse index, and returning the point cloud data to the original order.

[0025] In an alternative embodiment, fusing the large-scale features and the small-scale features to generate point cloud fusion features, and determining the vehicle surface flow field according to the point cloud fusion features, specifically including:

[0026] Performing residual connection on the large-scale features, the small-scale features and the point cloud basic features, and inputting them into a fully connected network for feature fusion to determine the point cloud fusion features;

[0027] Inputting the point cloud fusion features into a pre-trained pressure distribution prediction model to determine the pressure value at each point on the vehicle surface corresponding to the point cloud data;

[0028] Inputting the point cloud fusion features into a pre-trained wall shear stress distribution prediction model to determine the wall shear stress vector at each point on the vehicle surface corresponding to the point cloud data;

[0029] Determining the pressure distribution on the vehicle surface according to the pressure value, and determining the wall shear stress distribution on the vehicle surface according to the wall shear stress vector;

[0030] Determining the aerodynamic drag coefficient of the vehicle according to the pressure distribution and the wall shear stress distribution.

[0031] In an alternative embodiment, after extracting the large-scale features corresponding to the point cloud data according to the spatial similarity matrix, the method further includes:

[0032] Returning the large-scale features to the point cloud space according to the spatial similarity matrix.

[0033] The embodiments of the present disclosure further provide an aerodynamic evaluation device for a vehicle exterior, including:

[0034] A point cloud acquisition module, configured to acquire three-dimensional point cloud data describing the vehicle exterior, and generate a forward index and a reverse index corresponding to the point cloud data according to a preset space filling curve;

[0035] A large-scale feature extraction module, configured to determine the spatial similarity matrix corresponding to the point cloud data, and extract the large-scale features corresponding to the point cloud data according to the spatial similarity matrix;

[0036] A small-scale feature extraction module, configured to serialize the point cloud data according to the forward index, extract the small-scale features corresponding to the point cloud data, and return the point cloud data to the original order according to the reverse index;

[0037] An automobile surface flow field prediction module, configured to fuse the large-scale features and the small-scale features to generate point cloud fusion features, and determine the automobile surface flow field according to the point cloud fusion features.

[0038] An embodiment of the present disclosure further provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device runs, the processor communicates with the memory through the bus. When the machine-readable instructions are executed by the processor, the above-mentioned aerodynamic evaluation method for an automobile exterior, or the steps in any possible implementation manner of the above-mentioned aerodynamic evaluation method for an automobile exterior are executed.

[0039] An embodiment of the present disclosure further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the above-mentioned aerodynamic evaluation method for an automobile exterior, or the steps in any possible implementation manner of the above-mentioned aerodynamic evaluation method for an automobile exterior are executed.

[0040] An embodiment of the present disclosure further provides a computer program product, including a computer program / instructions. When the computer program and instructions are executed by a processor, the above-mentioned aerodynamic evaluation method for an automobile exterior, or the steps in any possible implementation manner of the above-mentioned aerodynamic evaluation method for an automobile exterior are implemented.

[0041] An aerodynamic evaluation method, device, and electronic device for an automobile exterior provided by an embodiment of the present disclosure obtain three-dimensional point cloud data describing the automobile exterior, and generate a forward index and a reverse index corresponding to the point cloud data according to a preset space filling curve; determine a spatial similarity matrix corresponding to the point cloud data, and extract large-scale features corresponding to the point cloud data according to the spatial similarity matrix; serialize the point cloud data according to the forward index, extract small-scale features corresponding to the point cloud data, and return the point cloud data to the original order according to the reverse index; fuse the large-scale features and the small-scale features to generate point cloud fusion features, and determine the automobile surface flow field according to the point cloud fusion features. By introducing the spatial aggregation technology and serialization technology of point clouds, the large-scale and small-scale structural features in the automobile exterior can be effectively extracted, and the surface pressure, wall shear stress distribution, and drag coefficient of the automobile exterior can be accurately predicted with lower computational complexity and video memory occupancy, thereby significantly improving the evaluation efficiency and prediction accuracy.

[0042] To make the above objects, features, and advantages of the present disclosure more apparent and understandable, the following provides preferred embodiments in conjunction with the accompanying drawings and describes them in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of this specification. These drawings show embodiments that conform to the present disclosure and, together with the specification, are used to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only show certain embodiments of the present disclosure and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can be obtained based on these drawings.

[0044] Figure 1 The flowchart showing a method for aerodynamic evaluation of a vehicle exterior provided by an embodiment of the present disclosure;

[0045] Figure 2 The flowchart showing another method for aerodynamic evaluation of a vehicle exterior provided by an embodiment of the present disclosure;

[0046] Figure 3 The schematic diagram showing a device for aerodynamic evaluation of a vehicle exterior provided by an embodiment of the present disclosure;

[0047] Figure 4 The schematic diagram showing an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present disclosure clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present disclosure in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of them. Usually, the components of the embodiments of the present disclosure described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure to be protected, but only represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present disclosure.

[0049] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0050] As used herein, the term "and / or" merely describes an associated relationship, indicating that there can be three relationships. For example, A and / or B can represent three cases: A exists alone, A and B exist simultaneously, and B exists alone. Additionally, the term "at least one" as used herein means any one of multiple items or any combination of at least two of multiple items. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set composed of A, B, and C.

[0051] It has been found through research that traditionally, the aerodynamic performance evaluation of vehicle exterior shapes relies on computational fluid dynamics simulation methods. By means of numerical calculations, the interaction between fluids and solid surfaces can be simulated, and the surface pressure, wall shear stress distribution, and drag coefficient of the vehicle exterior shape can be accurately calculated. However, the demand for computing resources is extremely large. Especially when dealing with complex three-dimensional geometric models, the calculation process often takes several days or even weeks. This high time cost limits the efficiency of vehicle design optimization and is not conducive to the actual needs of rapid iterative design.

[0052] Based on the above research, the present disclosure provides a method, apparatus, and electronic device for aerodynamic evaluation of vehicle exterior shapes. By obtaining three-dimensional point cloud data describing the vehicle exterior shape and generating a forward index and a reverse index corresponding to the point cloud data according to a preset space-filling curve; determining a spatial similarity matrix corresponding to the point cloud data, and extracting large-scale features corresponding to the point cloud data according to the spatial similarity matrix; serializing the point cloud data according to the forward index, extracting small-scale features corresponding to the point cloud data, and returning the point cloud data to the original order according to the reverse index; fusing the large-scale features and the small-scale features to generate point cloud fusion features, and determining the vehicle surface flow field according to the point cloud fusion features. By introducing the spatial aggregation technology and serialization technology of point clouds, the large-scale and small-scale structural features in the vehicle exterior shape can be effectively extracted, and the surface pressure, wall shear stress distribution, and drag coefficient of the vehicle exterior shape can be accurately predicted with lower computational complexity and video memory occupancy, thereby significantly improving the evaluation efficiency and prediction accuracy.

[0053] For the convenience of understanding this embodiment, first, a detailed introduction is given to an aerodynamic evaluation method for an automobile exterior shape disclosed in the embodiments of the present disclosure. The execution subject of the aerodynamic evaluation method for the automobile exterior shape provided in the embodiments of the present disclosure is generally a computer device with certain computing capabilities. Such a computer device includes, for example: a terminal device, a server, or other processing devices. The terminal device may be a user equipment (UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, an in-vehicle device, a wearable device, etc. In some possible implementation manners, the aerodynamic evaluation method for the automobile exterior shape may be implemented by a processor invoking computer-readable instructions stored in a memory.

[0054] See Figure 1 As shown, it is a flowchart of an aerodynamic evaluation method for an automobile exterior shape provided in the embodiments of the present disclosure. The method includes steps S101 to S104, where:

[0055] S101. Obtain three-dimensional point cloud data describing the automobile exterior shape, and generate a forward index and a reverse index corresponding to the point cloud data according to a preset space-filling curve.

[0056] In specific implementation, the point cloud data consists of a set of unordered points in three-dimensional space. Each point is represented by three-dimensional coordinates and is used to accurately describe the geometric shape of the automobile exterior. The acquisition method of the point cloud data can generate three-dimensional point cloud data by laser scanning the automobile exterior, or generate point clouds by projecting a grating pattern and recording the reflected light, or obtain point cloud data by taking multi-view cameras and performing three-dimensional reconstruction.

[0057] Here, the point cloud data has disorderliness, there is no fixed order between points, and the number of points may be very large. The automobile surface usually contains hundreds of thousands to millions of points.

[0058] Specifically, the point cloud is sorted through a preset space-filling curve to construct a serialized index of the point cloud. The forward index is used to convert the unordered point cloud into ordered data, and the reverse index is used to restore the original order of the point cloud after serialization processing to ensure data consistency.

[0059] Here, a space-filling curve is a mathematical method for mapping a high-dimensional space to a low-dimensional space (usually one-dimensional), traversing all points in the high-dimensional space through a continuous curve while maintaining local adjacency. The space-filling curves applied in the embodiments of the present application at least include Z-order curve, Hilbert curve, Tran Z-order curve, and Tran Hilbert curve. The Z-order curve is based on binary partitioning and constructs an index according to the bit-interleaved order of each-dimensional coordinate; the Hilbert curve is a recursively defined curve that can better maintain spatial locality.

[0060] Here, the Tran Z-order curve and the Tran Hilbert curve are obtained by changing the traversal order of different dimensions of the original Z-order curve and Hilbert curve. The traversal order of the Z-order curve and the Hilbert curve is x-y-z; the traversal order of the Tran Z-order curve and the Tran Hilbert curve is z-y-x.

[0061] As a possible implementation manner, the process of generating the forward index and the reverse index corresponding to the point cloud data can be referred to Figure 2 As shown in the flowchart of another method for aerodynamic evaluation of an automobile exterior provided by an embodiment of the present disclosure, the method includes steps S1011 to S1013, where:

[0062] S1011: Divide a three-dimensional space into multiple cube blocks.

[0063] S1012: Sort the cube blocks according to the preset space-filling curve, where the preset space-filling curve at least includes Z-order curve, Hilbert curve, Tran Z-order curve, and Tran Hilbert curve.

[0064] S1013: Assign corresponding indexes to each cube block, and arrange the point cloud data belonging to the same cube block in ascending order of the indexes to determine the forward index and the reverse index for serializing the point cloud data.

[0065] In a specific implementation, for a normalized n-dimensional point cloud set , divide the unit cube space into small cube blocks equally in each dimension, and the side length of each cube block is . Use an n-dimensional vector to represent the index of each small cube block.

[0066] For example, for n = 3 and = 4, the index of a cube block may be , indicating the 2nd division of the cube in the x direction, the 3rd division in the y direction, and the 1st division in the z direction.

[0067] Here, use a space-filling curve to sort the cube blocks, map the high-dimensional index to a one-dimensional sequential index, which can be represented by the following formula:

[0068]

[0069] Among them, is a mapping function from the n-dimensional space to the one-dimensional space to define the access order of the cube blocks; represents the number of cube blocks; represents the index of the cube block.

[0070] Furthermore, each cube block index is assigned a unique sorting value , that is, the sequential position in the space-filling curve. Group the point clouds belonging to the same cube, and then sort the point clouds of different groups according to the formula . After performing integer multiplication on the coordinates of the points with the side length of the cube block, map them to the index of the cube block where they are located.

[0071] Here, for the point cloud groups of different cubes, sort them according to the space-filling curve order of the cube blocks; for the point clouds within the same cube, sort them in ascending order to ensure the consistency of the point cloud order within the group.

[0072] Among them, record the position of the sorted point cloud data in the sequence to generate a forward index, which is used to convert the unordered point cloud into a serialized point cloud; construct an inverse mapping relationship according to the forward index, which is used to restore the original order of the point cloud in subsequent processing.

[0073] S102. Determine the spatial similarity matrix corresponding to the point cloud data, and extract the large-scale features corresponding to the point cloud data according to the spatial similarity matrix.

[0074] In a specific implementation, the point cloud data is encoded into high-dimensional point cloud basic features through linear transformation; a point-by-point linear transformation is performed on the point cloud basic features, and the global Softmax function is used to calculate the corresponding spatial similarity weights, and all the spatial similarity weights are integrated into a spatial similarity matrix; the point cloud data is transformed into multiple spatial slices, and the spatial aggregation features between the spatial slices are determined according to the spatial similarity matrix; the spatial aggregation features are processed through a self-attention mechanism to determine the large-scale features.

[0075] Here, it is assumed that the point cloud on the car surface is , where is the number of points. First, the input layer that will be linearly transformed encodes it into a high-dimensional space , where represents the dimension. The high-dimensional basic features of the point cloud are generated through linear transformation (such as a fully connected layer). Through point-by-point linear transformation, the similarity between each pair of points is calculated to capture the relationship of the point cloud in the global range. The combination of the similarity weights of all point pairs obtains the spatial similarity matrix W. The spatial similarity matrix W is a symmetric matrix, and each row or column corresponds to the global relationship of a point.

[0076] Furthermore, using the information of the spatial similarity matrix W, the point cloud is divided into several spatial slices. Each slice contains points with a relatively large correlation with a certain region. Using the above spatial similarity weights, the grid points can be encoded into the spatial aggregation features to obtain the spatial aggregation features.

[0077] Exemplarily, for the point cloud basic features , first use the point-by-point linear transformation and the global Softmax function to calculate the spatial similarity weights:

[0078]

[0079] Among them, represents the spatial similarity weight; represents the point cloud basic features, and .

[0080] Furthermore, encode the grid points into the spatial aggregation features according to the spatial similarity weights, that is:

[0081]

[0082] Among them, It is the spatial aggregation feature. Then, the self-attention mechanism processes these spatial aggregation features according to the query, key, and value of the features, generating dynamic adjustment of the attention weights between points to adjust the global relationship between each point. After passing through the multi-head attention mechanism, the large-scale features of the point cloud data are output.

[0083] It should be noted that after extracting the large-scale features corresponding to the point cloud data according to the spatial similarity matrix, the large-scale features are returned to the point cloud space according to the spatial similarity matrix. The large-scale features represent the correlation of the point cloud in the global space, and it captures the large-scale geometric structure on the surface of the point cloud, such as the overall contour of the car, the streamlined surface area, etc.

[0084] In this way, through the spatial similarity matrix, the global relationship information between all points is integrated. The self-attention mechanism dynamically adjusts the feature weights, retains important global characteristics, and avoids directly calculating the adjacency matrix, reducing resource consumption.

[0085] S103. Serialize the point cloud data according to the forward index, extract the small-scale features corresponding to the point cloud data, and return the point cloud data to the original order according to the reverse index.

[0086] In a specific implementation, the point cloud data is sorted according to the forward index to generate serialized point cloud data; the serialized point cloud data is divided into multiple point cloud groups, and the local features of each point cloud group are extracted through the self-attention mechanism to determine the small-scale features corresponding to the point cloud group; the point cloud groups are connected to form the point cloud data in the original shape; the point cloud data formed by connecting the point cloud groups is sorted according to the reverse index, and the point cloud data is returned to the original order.

[0087] Here, the forward index is generated based on the space-filling curve, which is used to organize the point cloud data into ordered serialized data. The order of the index reflects the spatial locality of the point cloud in the three-dimensional space. For example, adjacent points are also closer in the serialized data.

[0088] Specifically, the serialized point cloud data is divided into multiple point cloud groups, each group contains two adjacent point clouds, and each group of point clouds mainly represents the geometric characteristics within the local range of the point cloud, capturing detailed information. Within each point cloud group, the local small-scale features of the point cloud are extracted through the self-attention mechanism. The query (Query), key (Key), and value (Value) used by the self-attention mechanism are respectively generated through linear transformation, and then the Softmax function is used to calculate the attention weights between points to capture the correlation within the local range of the point cloud. Then, the small-scale features of all point cloud groups are combined to form the small-scale feature representation of the complete point cloud.

[0089] Specifically, the forward index generates serialized point clouds, organizes the point clouds according to the space-filling curve to enhance locality; groups and extracts small-scale features, and captures local geometric details through the self-attention mechanism; the reverse index restores the original order to ensure the integrity and consistency of the point cloud data after serialization.

[0090] Here, the reverse index is the inverse mapping of the forward index, which is used to restore the serialized point cloud data to its original order. The small-scale features are sorted according to the reverse index, so that each point cloud feature returns to its corresponding original position.

[0091] Exemplarily, assuming the point cloud data [a b c], the forward index is [2 0 1], and after serialization, it becomes [c a b]. The corresponding reverse index is [1 2 0], and the original sequence [a b c] can be obtained again using the reverse index.

[0092] Among them, the serialization operation converts the originally disordered point cloud into ordered data, uses the space-filling curve to maintain local adjacency, provides a basis for extracting local features, and effectively captures the geometric details of the point cloud in the local range, such as surface undulations and boundary characteristics, through the grouped self-attention mechanism. The reverse index ensures that serialization and local feature extraction do not damage the original geometric structure of the point cloud, and the restored data can be directly used for subsequent fusion and prediction.

[0093] S104. Fuse the large-scale features and the small-scale features to generate point cloud fusion features, and determine the vehicle surface flow field according to the point cloud fusion features.

[0094] In a specific implementation, the large-scale features, small-scale features and point cloud basic features are connected by residuals and input into a fully connected network for feature fusion to determine the point cloud fusion features; the point cloud fusion features are input into a pre-trained pressure distribution prediction model and a wall shear stress distribution prediction model to determine the pressure value and the wall shear stress vector at each point on the vehicle surface corresponding to the point cloud data; the drag coefficient of the vehicle is determined according to the pressure value and the wall shear stress vector.

[0095] Here, layer normalization (LayerNorm) is performed on the point cloud basic features to standardize each feature dimension to ensure that the features have a stable numerical range. Then, the physical attention mechanism (Physics-Attention) is used to extract the global correlation and large-scale structural features of the point cloud. The physical attention mechanism uses the spatial similarity matrix of the point cloud and combines the self-attention mechanism to perform weighted aggregation on the global information to generate large-scale features, representing the global geometric characteristics of the point cloud, such as the overall contour or aerodynamically important regions.

[0096] Here, the difference between the large-scale features and the point cloud basic features is calculated, and this difference is processed by Layer Normalization (LayerNorm) to ensure that the numerical range of the local features is adapted to the subsequent attention mechanism. Then, the Serialized-Attention mechanism is applied to extract the local features of the point cloud. Based on the local grouping relationship after the serialization of the point cloud, the Serialized-Attention mechanism utilizes the adjacent information between points to capture the detailed features and reflect the local geometric features of the point cloud, such as surface details or boundary features.

[0097] Furthermore, the large-scale features, small-scale features, and point cloud basic features are fused through residual connections to generate point cloud fusion features:

[0098]

[0099] Among them, represents the point cloud fusion feature; represents the large-scale feature; represents the small-scale feature; represents the point cloud basic feature. It consists of two layers of linear transformation and a non-linear activation function, which is used to enhance the expression ability of the features.

[0100] Furthermore, using the trained deep neural network model, the fusion features are input into it to predict the pressure value corresponding to each point cloud. According to the pressure distribution of all points, the overall flow field distribution on the car surface is determined, and the drag coefficient is calculated based on the pressure distribution using a preset physical formula.

[0101] In this way, considering both the global geometric structure and local detailed features of the point cloud, the prediction accuracy of the flow field of complex car shapes is improved. The residual connection ensures the integrity of information, and the non-linear transformation enhances the expression ability after feature fusion.

[0102] An aerodynamic evaluation method for a vehicle exterior provided by an embodiment of the present disclosure obtains three-dimensional point cloud data describing the vehicle exterior and generates a forward index and a reverse index corresponding to the point cloud data according to a preset space filling curve; determines a spatial similarity matrix corresponding to the point cloud data, and extracts large-scale features corresponding to the point cloud data according to the spatial similarity matrix; serializes the point cloud data according to the forward index, extracts small-scale features corresponding to the point cloud data, and returns the point cloud data to the original order according to the reverse index; fuses the large-scale features and the small-scale features to generate point cloud fusion features, and determines the vehicle surface flow field according to the point cloud fusion features. By introducing the spatial aggregation technology and serialization technology of point clouds, large-scale and small-scale structural features in the vehicle exterior can be effectively extracted, and the surface pressure, wall shear stress distribution and drag coefficient of the vehicle exterior can be accurately predicted with lower computational complexity and video memory occupancy, thus significantly improving the evaluation efficiency and prediction accuracy.

[0103] Those skilled in the art can understand that in the above method of the specific embodiment, the writing order of each step does not mean a strict execution order and does not constitute any limitation to the implementation process. The specific execution order of each step should be determined according to its function and possible internal logic.

[0104] Based on the same inventive concept, an aerodynamic evaluation device for a vehicle exterior corresponding to the aerodynamic evaluation method for a vehicle exterior is also provided in an embodiment of the present disclosure. Since the principle of solving problems by the device in the embodiment of the present disclosure is similar to the above aerodynamic evaluation method for a vehicle exterior in the embodiment of the present disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0105] Please refer to Figure 3 , Figure 3 which is a schematic diagram of an aerodynamic evaluation device for a vehicle exterior provided by an embodiment of the present disclosure. As Figure 3 shown in

[0106] A point cloud acquisition module 310 is configured to acquire three-dimensional point cloud data describing the vehicle exterior and generate a forward index and a reverse index corresponding to the point cloud data according to a preset space filling curve.

[0107] A large-scale feature extraction module 320 is configured to determine a spatial similarity matrix corresponding to the point cloud data and extract large-scale features corresponding to the point cloud data according to the spatial similarity matrix.

[0108] A small-scale feature extraction module 330 is configured to serialize the point cloud data according to the forward index, extract small-scale features corresponding to the point cloud data, and return the point cloud data to the original order according to the reverse index.

[0109] The vehicle surface flow field prediction module 340 is configured to fuse the large-scale features and the small-scale features to generate point cloud fusion features, and determine the vehicle surface flow field according to the point cloud fusion features.

[0110] For the processing flow of each module in the device and the interaction flow between the modules, reference may be made to the relevant descriptions in the above method embodiments, which will not be elaborated here.

[0111] An aerodynamic evaluation device for a vehicle exterior provided by an embodiment of the present disclosure obtains three-dimensional point cloud data describing the vehicle exterior, and generates a forward index and a reverse index corresponding to the point cloud data according to a preset space filling curve; determines a spatial similarity matrix corresponding to the point cloud data, and extracts large-scale features corresponding to the point cloud data according to the spatial similarity matrix; serializes the point cloud data according to the forward index, extracts small-scale features corresponding to the point cloud data, and returns the point cloud data to the original order according to the reverse index; fuses the large-scale features and the small-scale features to generate point cloud fusion features, and determines the vehicle surface flow field according to the point cloud fusion features. By introducing the spatial aggregation technology and serialization technology of point clouds, the large-scale and small-scale structural features in the vehicle exterior can be effectively extracted, and the surface pressure distribution and drag coefficient of the vehicle exterior can be accurately predicted with relatively low computational complexity and video memory occupancy, thereby significantly improving the evaluation efficiency and prediction accuracy.

[0112] Corresponding to Figure 1 the aerodynamic evaluation method for a vehicle exterior in Figure 4 as shown, which is a schematic structural diagram of an electronic device 400 provided by an embodiment of the present disclosure, including:

[0113] A processor 41, a memory 42, and a bus 43; the memory 42 is used to store execution instructions, including an internal memory 421 and an external memory 422; the internal memory 421 here is also called the main memory, which is used to temporarily store the operation data in the processor 41 and the data exchanged with the external memory 422 such as a hard disk. The processor 41 exchanges data with the external memory 422 through the internal memory 421. When the electronic device 400 runs, the processor 41 communicates with the memory 42 through the bus 43, so that the processor 41 executes Figure 1 the Figure 2 steps of the aerodynamic evaluation method for a vehicle exterior in

[0114] An embodiment of the present disclosure also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it executes the steps of the aerodynamic evaluation method for the automobile exterior shape described in the above method embodiment. Among them, the storage medium may be a volatile or non-volatile computer-readable storage medium.

[0115] An embodiment of the present disclosure also provides a computer program product, which includes computer instructions. When the computer instructions are executed by a processor, they can execute the steps of the aerodynamic evaluation method for the automobile exterior shape described in the above method embodiment. For details, please refer to the above method embodiment and will not be elaborated here.

[0116] Among them, the above computer program product can be specifically implemented in a way of hardware, software or a combination thereof. In an optional embodiment, the computer program product is specifically embodied as a computer storage medium. In another optional embodiment, the computer program product is specifically embodied as a software product, such as a Software Development Kit (SDK), etc.

[0117] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the above-described device can refer to the corresponding process in the foregoing method embodiment and will not be elaborated here. In several embodiments provided by the present disclosure, it should be understood that the disclosed device and method can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces, and the indirect coupling or communication connection of the device or unit may be in an electrical, mechanical or other form.

[0118] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0119] In addition, each functional unit in various embodiments of the present disclosure may be integrated in a processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit.

[0120] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such an understanding, the technical solution of the present disclosure, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs.

[0121] Finally, it should be noted that the above-mentioned embodiments are only specific implementation manners of the present disclosure, used to illustrate the technical solutions of the present disclosure, rather than limiting them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: any person skilled in the technical field can still modify the technical solutions described in the foregoing embodiments, or can easily conceive of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes, or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should all be covered by the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims

1. A method for evaluating the aerodynamics of a vehicle's exterior, characterized in that: include: Acquire three-dimensional point cloud data describing the appearance of the car, and generate forward indexes and reverse indexes corresponding to the point cloud data according to a preset space filling curve; Determine a spatial similarity matrix corresponding to the point cloud data, and extract large-scale features corresponding to the point cloud data according to the spatial similarity matrix; Serializing the point cloud data according to the forward index, extracting small-scale features corresponding to the point cloud data, and returning the point cloud data to the original order according to the reverse index; fusing the large-scale features with the small-scale features to generate a point cloud fusion feature, and determining the flow field on the surface of the vehicle according to the point cloud fusion feature; Determining a spatial similarity matrix corresponding to the point cloud data, and extracting large-scale features corresponding to the point cloud data according to the spatial similarity matrix, specifically includes: Encoding the point cloud data into high-dimensional point cloud basic features through linear transformation; Performing point-by-point linear transformation on the basic features of the point cloud, calculating the corresponding spatial similarity weights using a global Softmax function, and integrating all the spatial similarity weights into the spatial similarity matrix; Converting the point cloud data into a plurality of spatial slices, and determining spatial aggregation features between the spatial slices according to the spatial similarity matrix; Processing the spatial aggregation features through a self-attention mechanism to determine the large-scale features; The large-scale features and the small-scale features are fused to generate a point cloud fusion feature, and the flow field on the surface of the vehicle is determined according to the point cloud fusion feature, specifically including: Residual connection is performed on the large-scale features, the small-scale features and the point cloud basic features, and the features are input into a fully connected network for feature fusion to determine the point cloud fusion features; Inputting the point cloud fusion features into a pre-trained pressure distribution prediction model to determine the pressure value at each point on the car surface corresponding to the point cloud data; Inputting the point cloud fusion features into a pre-trained wall shear stress distribution prediction model to determine the wall shear stress vector at each point on the automobile surface corresponding to the point cloud data; Determining the pressure distribution on the surface of the vehicle according to the pressure value, and determining the wall shear stress distribution on the surface of the vehicle according to the wall shear stress vector; The drag coefficient of the vehicle is determined according to the pressure distribution and the wall shear stress distribution.

2. The method according to claim 1, characterized in that Generating a forward index and a reverse index corresponding to the point cloud data according to a preset space filling curve specifically includes: Divide the three-dimensional space into multiple cubic blocks; Sorting the cubic blocks according to the preset space filling curve, wherein the preset space filling curve includes at least a Z-order curve, a Hilbert curve, a Tran Z-order curve and a Tran Hilbert curve; A corresponding index is assigned to each of the cubic blocks, and the point cloud data belonging to the same cubic block are arranged in ascending order according to the indexes, and the forward index and the reverse index for serializing the point cloud data are determined.

3. The method according to claim 1, characterized in that Serializing the point cloud data according to the forward index and extracting small-scale features corresponding to the point cloud data specifically includes: Sort the point cloud data according to the forward index to generate serialized point cloud data; The serialized point cloud data is divided into a plurality of point cloud groups, local features of each point cloud group are extracted through a self-attention mechanism, and the small-scale features corresponding to the point cloud group are determined.

4. The method according to claim 3, characterized in that Returning the point cloud data to the original order according to the reverse index specifically includes: Connecting the point cloud groups to form the point cloud data of the original shape; The point cloud data formed by connecting the point cloud groups are sorted according to the reverse index, and the point cloud data are returned to the original order.

5. The method according to claim 1, characterized in that After extracting the large-scale features corresponding to the point cloud data according to the spatial similarity matrix, the method further includes: The large-scale features are returned to the point cloud space according to the spatial similarity matrix.

6. An aerodynamic evaluation device for automobile appearance, characterized in that: include: A point cloud acquisition module, used to acquire three-dimensional point cloud data describing the appearance of the car, and generate a forward index and a reverse index corresponding to the point cloud data according to a preset space filling curve; A large-scale feature extraction module, used to determine a spatial similarity matrix corresponding to the point cloud data, and extract large-scale features corresponding to the point cloud data according to the spatial similarity matrix; The large-scale feature extraction module is specifically used to encode the point cloud data into high-dimensional point cloud basic features through linear transformation; Performing point-by-point linear transformation on the basic features of the point cloud, calculating the corresponding spatial similarity weights using a global Softmax function, and integrating all the spatial similarity weights into the spatial similarity matrix; Converting the point cloud data into a plurality of spatial slices, and determining spatial aggregation features between the spatial slices according to the spatial similarity matrix; Processing the spatial aggregation features through a self-attention mechanism to determine the large-scale features; A small-scale feature extraction module, used to serialize the point cloud data according to the forward index, extract the small-scale features corresponding to the point cloud data, and return the point cloud data to the original order according to the reverse index; A vehicle surface flow field prediction module, used for fusing the large-scale features with the small-scale features to generate a point cloud fusion feature, and determining the vehicle surface flow field according to the point cloud fusion feature; The automobile surface flow field prediction module is specifically used to perform residual connection on the large-scale features, the small-scale features and the point cloud basic features, and input them into a fully connected network for feature fusion to determine the point cloud fusion features; Inputting the point cloud fusion features into a pre-trained pressure distribution prediction model to determine the pressure value at each point on the car surface corresponding to the point cloud data; Inputting the point cloud fusion features into a pre-trained wall shear stress distribution prediction model to determine the wall shear stress vector at each point on the automobile surface corresponding to the point cloud data; Determining the pressure distribution on the surface of the vehicle according to the pressure value, and determining the wall shear stress distribution on the surface of the vehicle according to the wall shear stress vector; The drag coefficient of the vehicle is determined according to the pressure distribution and the wall shear stress distribution.

7. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the aerodynamic evaluation method for the automobile appearance as described in any one of claims 1 to 5 are performed.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the aerodynamic evaluation method of the vehicle appearance as claimed in any one of claims 1 to 5 are executed.

Images