Model training method, data prediction method, device, equipment and vehicle

Abstract

The invention provides a model training method, a data prediction method and device, equipment and a vehicle, relates to the technical field of artificial intelligence, and is used for solving the problems of long time consumption and low efficiency of data analysis. The method comprises the following steps: acquiring external flow field characteristic data of a vehicle at multiple moments; based on the external flow field characteristic data at the multiple moments, training an external flow field analysis model to obtain a trained external flow field analysis model; the external flow field analysis model comprises an encoder and a decoder; the encoder is used for converting the external flow field characteristic data into characteristic data and obtaining predicted characteristic data based on the characteristic data; the decoder is used for converting the predicted characteristic data into predicted external flow field characteristic data; and external flow field characteristic data can be efficiently predicted through the external flow field analysis model.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a model training method, a data prediction method, an apparatus, equipment, and a vehicle. Background Art

[0002] The aerodynamic characteristics of a vehicle's external flow field affect its fuel economy, driving stability, and noise levels, making it a primary research topic in vehicle aerodynamic design. Accurately analyzing the characteristics of this external flow field can optimize vehicle shape, reduce air resistance, and ultimately improve vehicle performance.

[0003] However, analyzing the flow field around a vehicle involves complex three-dimensional nonlinear fluid dynamics, which places high demands on computing resources and time. Existing techniques can describe the flow field around a vehicle using finite element analysis, but this method is time-consuming and inefficient, making it difficult to meet the efficiency requirements of practical applications. Summary of the Invention

[0004] One of the purposes of the present invention is to provide a model training method to solve the problem of time-consuming and inefficient analysis of vehicle external flow field characteristic data in the prior art; the second purpose is to provide a method for predicting external flow field characteristic data; the third purpose is to provide a model training device; the fourth purpose is to provide a device for predicting external flow field characteristic data; the fifth purpose is to provide an electronic device; and the sixth purpose is to provide a vehicle.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] In a first aspect, the present invention provides a model training method, comprising: obtaining external flow field characteristic data of a vehicle at multiple moments, the external flow field characteristic data being used to characterize the distribution of flow field variables in a spatiotemporal domain;

[0007] Based on the external flow field characteristic data at the multiple moments, the external flow field analysis model is trained to obtain a trained external flow field analysis model; wherein, the external flow field analysis model includes an encoder and a decoder; the encoder is used to convert the external flow field characteristic data into feature data, and obtain predicted feature data based on the feature data; the decoder is used to convert the predicted feature data into predicted external flow field characteristic data; wherein, the dimension of the feature data is lower than the dimension of the external flow field characteristic data.

[0008] According to the above technical means, data dimensionality reduction and feature extraction are achieved through the encoder of the external flow field analysis model, and the extracted feature data is restored through the decoder to obtain the predicted external flow field characteristic data, which reduces the dimension and complexity of the external flow field characteristic data and reduces the calculation time, thereby achieving high-precision prediction of the external flow field characteristic data.

[0009] In one possible embodiment, the external flow field characteristic data of the vehicle at multiple moments are obtained, including: simulating the external flow field of the vehicle and extracting initial external flow field characteristic data at multiple moments from the simulation results; wherein the initial external flow field characteristic data includes data of n dimensions, where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results; and performing dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

[0010] According to the above technical means, the initial external flow field characteristic data at multiple moments can be obtained through software simulation, which expands the data volume and improves the data acquisition efficiency.

[0011] In one possible embodiment, the initial external flow field characteristic data at the multiple moments are subjected to dimensionality reduction processing to obtain the external flow field characteristic data at the multiple moments, including: constructing the initial external flow characteristic data at the multiple moments into a first data matrix; performing singular value decomposition on the first data matrix, and using the left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as first POD modal basis vectors; wherein the set number of singular values ​​includes multiple singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; constructing a second data matrix based on the first POD modal basis vectors; performing dimensionality reduction processing on the second data matrix to obtain a projection matrix; performing dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

[0012] According to the above technical means, the initial external flow characteristic data can be reduced in dimension through singular value decomposition and mode extraction to obtain external flow field characteristic data, thereby reducing the data dimension of the input external flow field analysis model and improving the data analysis efficiency of the external flow field analysis model.

[0013] In one possible implementation, the second data matrix is ​​subjected to dimensionality reduction processing to obtain a projection matrix, including: performing singular value decomposition and modal extraction on the second data matrix to obtain second POD modal basis vectors; and interpolating and indexing the second POD modal basis vectors using a DEIM algorithm to construct the projection matrix, wherein the dimension of the projection matrix is ​​lower than the dimension of the second data matrix. According to the above technical means, the projection matrix can be constructed using the DEIM algorithm to reduce the dimension of the second data matrix.

[0014] In a possible implementation, the projection matrix is ​​subjected to dimensionality reduction processing to obtain the external flow field characteristic data at the multiple moments, including: according to the second POD modal basis vector and the projection matrix, through the equation F≈Ψ(P T Ψ) - 1 PT F, reconstruct the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; the dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

[0015] According to the above technical means, the data dimension of the second data matrix can be further reduced by nonlinear term reconstruction to obtain data with lower dimension.

[0016] In a possible implementation manner, the flow field variable includes at least one of the following: gas velocity and gas pressure.

[0017] In a possible implementation, obtaining the external flow field characteristic data of the vehicle at multiple moments includes: obtaining the external flow field characteristic data at multiple moments under multiple parameter conditions; wherein different parameter conditions have different gas velocities and / or gas pressures.

[0018] In a possible implementation, the external flow field analysis model is constructed based on an LSTM neural network.

[0019] In a second aspect, the present invention provides a method for predicting external flow field characteristic data, the method comprising:

[0020] Acquiring external flow field characteristic data of the vehicle at multiple times;

[0021] The external flow field characteristic data at the multiple moments are input into the external flow field analysis model to obtain the predicted external flow field characteristic data output by the external flow field analysis model; wherein, the external flow field analysis model is trained by the model training method of the first aspect mentioned above.

[0022] According to the above technical means, the external flow field characteristic data can be predicted through the external flow field analysis model, thereby improving the efficiency of data prediction.

[0023] In one possible embodiment, the external flow field characteristic data of the vehicle at multiple moments are obtained, including: simulating the external flow field of the vehicle and extracting initial external flow field characteristic data at multiple moments from the simulation results; wherein the initial external flow field characteristic data includes data of n dimensions, where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results; and performing dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

[0024] According to the above technical means, the initial external flow field characteristic data at multiple moments can be obtained through software simulation, which expands the data volume and improves the data acquisition efficiency.

[0025] In one possible embodiment, the initial external flow field characteristic data at the multiple moments are subjected to dimensionality reduction processing to obtain the external flow field characteristic data at the multiple moments, including: constructing the initial external flow characteristic data at the multiple moments into a first data matrix; performing singular value decomposition on the first data matrix, and using the left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as first POD modal basis vectors; wherein the set number of singular values ​​includes multiple singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; constructing a second data matrix based on the first POD modal basis vectors; performing dimensionality reduction processing on the second data matrix to obtain a projection matrix; performing dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

[0026] According to the above technical means, the initial external flow characteristic data can be reduced in dimension through singular value decomposition and mode extraction to obtain external flow field characteristic data, thereby reducing the data dimension of the input external flow field analysis model and improving the data analysis efficiency of the external flow field analysis model.

[0027] In one possible implementation, the second data matrix is ​​subjected to dimensionality reduction processing to obtain a projection matrix, including: performing singular value decomposition and modal extraction on the second data matrix to obtain a second POD modal basis vector; interpolating and indexing the second POD modal basis vector through a DEIM algorithm to construct the projection matrix, wherein the dimension of the projection matrix is ​​lower than the dimension of the second data matrix.

[0028] According to the above technical means, the projection matrix can be constructed through the DEIM algorithm to reduce the dimension of the second data matrix.

[0029] In a possible implementation, the projection matrix is ​​subjected to dimensionality reduction processing to obtain the external flow field characteristic data at the multiple moments, including: according to the second POD modal basis vector and the projection matrix, through the equation F≈Ψ(P T Ψ) - 1 P T F, reconstruct the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; the dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

[0030] According to the above technical means, the data dimension of the second data matrix can be further reduced by nonlinear term reconstruction to obtain data with lower dimension.

[0031] In a possible implementation manner, the flow field variable includes at least one of the following: gas velocity and gas pressure.

[0032] In a possible implementation, obtaining the external flow field characteristic data of the vehicle at multiple moments includes: obtaining the external flow field characteristic data at multiple moments under multiple parameter conditions; wherein different parameter conditions have different gas velocities and / or gas pressures.

[0033] In a possible implementation, the external flow field analysis model is constructed based on an LSTM neural network.

[0034] In a third aspect, the present invention provides a device for model training, comprising:

[0035] A first acquisition module is used to acquire external flow field characteristic data of the vehicle at multiple moments, where the external flow field characteristic data is used to characterize the distribution of flow field variables in the time and space domain;

[0036] A training module is used to train an external flow field analysis model based on the external flow field characteristic data at multiple moments to obtain a trained external flow field analysis model; wherein, the external flow field analysis model includes an encoder and a decoder; the encoder is used to convert the external flow field characteristic data into feature data, and obtain predicted feature data based on the feature data; the decoder is used to convert the predicted feature data into predicted external flow field characteristic data; wherein, the dimension of the feature data is lower than the dimension of the external flow field characteristic data.

[0037] According to the above technical means, data dimensionality reduction and feature extraction are achieved through the encoder of the external flow field analysis model, and the extracted feature data is restored through the decoder to obtain the predicted external flow field characteristic data, which reduces the dimension and complexity of the external flow field characteristic data and reduces the calculation time, thereby achieving high-precision prediction of the external flow field characteristic data.

[0038] The first acquisition module is used to simulate the external flow field of the vehicle and extract initial external flow field characteristic data at multiple moments from the simulation results; wherein the initial external flow field characteristic data includes data of n dimensions, where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results; and perform dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

[0039] The first acquisition module is used to construct the initial external flow characteristic data of the multiple moments into a first data matrix; perform singular value decomposition on the first data matrix, and use the left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as the first POD modal basis vectors; wherein the set number of singular values ​​includes multiple singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; construct a second data matrix based on the first POD modal basis vectors; perform dimensionality reduction processing on the second data matrix to obtain a projection matrix; perform dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

[0040] The first acquisition module is used to perform singular value decomposition and modal extraction on the second data matrix to obtain a second POD modal basis vector; the second POD modal basis vector is interpolated and indexed by the DEIM algorithm to construct the projection matrix, and the dimension of the projection matrix is ​​lower than the dimension of the second data matrix.

[0041] The first acquisition module is used to obtain the second POD modal basis vector and the projection matrix through the equation F≈Ψ(P T Ψ) -1 P T F, reconstruct the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; the dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

[0042] The flow field variable includes at least one of the following: gas velocity and gas pressure.

[0043] The first acquisition module is used to acquire external flow field characteristic data at multiple moments under multiple parameter conditions; wherein different parameter conditions have different gas velocities and / or gas pressures.

[0044] The external flow field analysis model is built based on the LSTM neural network.

[0045] In a fourth aspect, the present invention provides a device for predicting external flow field characteristic data, the device comprising:

[0046] A second acquisition module is used to acquire external flow field characteristic data of the vehicle at multiple moments;

[0047] The prediction module is used to input the external flow field characteristic data at multiple moments into the external flow field analysis model to obtain the predicted external flow field characteristic data output by the external flow field analysis model; wherein, the external flow field analysis model is trained by the model training method of the first aspect mentioned above.

[0048] According to the above technical means, the external flow field characteristic data can be predicted through the external flow field analysis model, thereby improving the efficiency of data prediction.

[0049] The second acquisition module is used to simulate the external flow field of the vehicle and extract initial external flow field characteristic data at multiple moments from the simulation results; wherein the initial external flow field characteristic data includes data of n dimensions, where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results; and perform dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

[0050] The second acquisition module is used to construct the initial external flow characteristic data of the multiple moments into a first data matrix; perform singular value decomposition on the first data matrix, and use the left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as the first POD modal basis vectors; wherein the set number of singular values ​​includes multiple singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; construct a second data matrix based on the first POD modal basis vectors; perform dimensionality reduction processing on the second data matrix to obtain a projection matrix; perform dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

[0051] The second acquisition module is used to perform singular value decomposition and modal extraction on the second data matrix to obtain a second POD modal basis vector; the second POD modal basis vector is interpolated and indexed by the DEIM algorithm to construct the projection matrix, and the dimension of the projection matrix is ​​lower than the dimension of the second data matrix.

[0052] The second acquisition module is used to obtain the second POD modal basis vector and the projection matrix through the equation F≈Ψ(P T Ψ) -1 P T F, reconstruct the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; the dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

[0053] The flow field variable includes at least one of the following: gas velocity and gas pressure.

[0054] The second acquisition module is used to acquire external flow field characteristic data at multiple moments under multiple parameter conditions; wherein different parameter conditions have different gas velocities and / or gas pressures.

[0055] The external flow field analysis model is built based on the LSTM neural network.

[0056] In a fifth aspect, the present invention provides an electronic device comprising a memory and a processor, wherein the memory and the processor are connected, and the processor is used to execute the model training method of the first aspect or the external flow field characteristic data prediction method of the second aspect stored in the memory.

[0057] In a sixth aspect, the present invention provides a vehicle comprising the device for predicting external flow field characteristic data according to the fourth aspect.

[0058] Beneficial effects of the present invention:

[0059] (1) According to the above technical means, feature data extraction and feature data prediction are realized through the encoder of the external flow field analysis model, and the predicted feature data is restored through the decoder to obtain predicted external flow field characteristic data, thereby reducing the dimension and complexity of the external flow field characteristic data, reducing the calculation time, and realizing efficient external flow field characteristic data prediction;

[0060] (2) According to the above technical means, the external flow field characteristic data can be predicted through the external flow field analysis model, thereby improving the efficiency of data prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A flowchart of a model training method provided by the present invention.

[0062] Figure 2 A flowchart of another model training method provided by the present invention.

[0063] Figure 3 The present invention provides a flowchart of a method for constructing a projection matrix using a DEIM algorithm.

[0064] Figure 4 The present invention provides a flow chart of a method for predicting external flow field characteristic data.

[0065] Figure 5 A block diagram of a model training device provided by the present invention.

[0066] Figure 6 A block diagram of a device for predicting external flow field characteristic data provided by the present invention.

[0067] Figure 7This is a block diagram of an electronic device provided by the present invention.

[0068] Figure 8 A block diagram of a vehicle provided by the present invention. DETAILED DESCRIPTION

[0069] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and are not intended to limit the scope of protection of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention.

[0070] First, let's introduce the application scenario of this invention, which involves analyzing and predicting vehicle external flow field characteristic data. The aerodynamic characteristics of a vehicle's external flow field affect its fuel economy, driving stability, and noise level, and are a primary research topic in vehicle aerodynamic design. Therefore, by accurately analyzing the characteristics of this external flow field, the vehicle body shape can be optimized, thereby reducing air resistance and improving vehicle performance.

[0071] Finite element analysis (FEM) is currently used to analyze the characteristic data of the vehicle's external flow field. However, this data analysis involves complex three-dimensional nonlinear fluid dynamics, typically requiring the solution of high-dimensional partial differential equations and multiple calculations. Consequently, this method is time-consuming and inefficient, making it difficult to meet the efficiency requirements of practical applications.

[0072] Therefore, the present invention proposes a model training method, data prediction method, device, equipment and vehicle, which obtain external flow field characteristic data of the vehicle at multiple moments, and the external flow field characteristic data is used to characterize the distribution of flow field variables in the time and space domain; based on the external flow field characteristic data at multiple moments, the external flow field analysis model is trained to obtain a trained external flow field analysis model; wherein, the external flow field analysis model includes an encoder and a decoder; the encoder is used to convert the external flow field characteristic data into feature data, and obtain predicted feature data based on the feature data; the decoder is used to convert the predicted feature data into predicted external flow field characteristic data; wherein the dimension of the feature data is lower than the dimension of the external flow field characteristic data; feature data extraction and feature data prediction are realized by the encoder of the external flow field analysis model in the above technical solution, and the predicted feature data is restored by the decoder to obtain predicted external flow field characteristic data, thereby reducing the dimension and complexity of the external flow field characteristic data, reducing the calculation time, and realizing efficient external flow field characteristic data prediction; and, the external flow field characteristic data can be predicted by the external flow field analysis model to improve the efficiency of data prediction.

[0073] Since the present invention relates to the training of the external flow field analysis model, that is, the external flow field analysis model is trained through the external flow field characteristic data at multiple moments, the model training process can be executed through the model training end, and the model training end can be a software and hardware system for completing the artificial intelligence model parameter learning, and by computing and processing the labeled data, a set of model parameters with prediction or decision-making capabilities is generated. In addition, the present invention also relates to the application of the external flow field analysis model, that is, the prediction of the vehicle's external flow field characteristic data is achieved through the external flow field analysis model, and the model application process can be executed through the model application end, and the model application end can be a software and hardware system that uses the trained model to perform inference and prediction on real-time input data, and converts the model parameters into decisions or output results for specific business scenarios. The above-mentioned model training end and model application end can be the same or different.

[0074] like Figure 1 As shown, the present invention provides a method for model training, which may include the following steps.

[0075] S101. Acquire external flow field characteristic data of a vehicle at multiple moments.

[0076] The external flow field characteristic data is used to characterize the distribution of flow field variables in the time and space domain.

[0077] For example, the flow field variables may include at least one of the following: gas velocity and gas pressure. The main dynamic characteristics of the external flow field can be characterized by the distribution of the flow field variables.

[0078] S102 : Based on the external flow field characteristic data at the multiple moments, an external flow field analysis model is trained to obtain a trained external flow field analysis model.

[0079] In which, the external flow field analysis model may include an encoder and a decoder; the encoder is used to convert the external flow field characteristic data into feature data, and obtain predicted feature data based on the feature data; the decoder is used to convert the predicted feature data into predicted external flow field characteristic data; wherein, the dimension of the feature data is lower than the dimension of the external flow field characteristic data.

[0080] For example, the external flow field analysis model can be constructed based on an LSTM (Long Short-Term Memory) neural network. The LSTM neural network can include multiple layers, each layer can include multiple neural units. For example, the LSTM neural network can include 2 layers, each layer can include 128 neural units. The LSTM neural network is a special recurrent neural network (RNN) that can include memory units, gating mechanisms, and state update processes.

[0081] In one possible embodiment, the external flow field analysis model may include an encoder and a decoder, and the encoder may include a convolution layer and an activation function; the convolution layer can be used to convert the external flow field data at multiple moments into feature data, and obtain predicted feature data based on the feature data, and the activation function is used to perform linear rectification on the predicted feature data.

[0082] The decoder may include a transposed convolution layer and an optimization function; the transposed convolution layer may be used to convert the feature data into predicted external flow field data, and the optimization function may be used to process abnormal data during the data conversion process.

[0083] For example, the encoder input can be external flow field data at multiple moments, and the output can be feature data predicted at multiple future moments. For example, the multiple moments can be 80 time steps, and the multiple future moments can be 20 time steps. That is, the input data can be z_t (t=1 to 80), and the output data can be z_{t+1} to z_{t+20}.

[0084] For example, the encoder's convolutional layer can be a convolutional neural network (CNN), which can include five layers, each with a convolution kernel size of 3×3×3, and the number of channels per layer being 64, 32, 16, 8, or 4, respectively. The encoder's activation function can be ReLU. The decoder's optimization function can be an Adam optimizer, with an initial learning rate of 1e-4, a batch size of 32, and a training cycle of 500 cycles.

[0085] In one possible implementation, the external flow field analysis model can be optimized using a loss function based on the loss value between the predicted value and the true value output by the external flow field analysis model, so that the optimized external flow field analysis model can achieve data prediction more accurately and efficiently.

[0086] For example, the loss function can be a mean square error loss function, each data sample (prediction value) can correspond to a loss value, and the mean square error can be calculated using the loss values ​​of multiple data samples. The loss value of the loss function can range from 0.05 to 0.15, for example, 0.05, 0.1 or 0.15, etc., which is not limited here. The learning rate of the mean square error loss function can be a decay of 0.5 every 50 cycles. According to the above technical means, the external flow field analysis model can be optimized by the loss function to improve the accuracy of the external flow field analysis model training.

[0087] Through the above technical solution, feature data extraction and feature data prediction can be realized through the encoder of the external flow field analysis model, and the predicted feature data can be restored through the decoder to obtain the predicted external flow field characteristic data, thereby reducing the dimension and complexity of the external flow field characteristic data, reducing the calculation time, and realizing efficient external flow field characteristic data prediction.

[0088] In one possible implementation, step S101 may include acquiring external flow field characteristic data at multiple moments under multiple parameter conditions, wherein the different parameter conditions may differ in gas velocity and / or gas pressure. For example, the parameters may include gas velocity and / or gas pressure.

[0089] For example, the parameters may include a first gas velocity, a first gas pressure, a second gas velocity, and a second gas pressure, and the different parameter conditions may include a first parameter condition, a second parameter condition, a third parameter condition, a fourth parameter condition, and a fifth parameter condition, etc. The first parameter condition may refer to a situation where the first gas velocity and the first gas pressure exist, the second parameter condition may refer to a situation where the first gas velocity and the second gas pressure exist, the third parameter condition may refer to a situation where the second gas velocity and the second gas pressure exist, the fourth parameter condition may refer to a situation where the first gas pressure exists, the fifth parameter condition may refer to a situation where the second gas velocity exists, etc. Similarly, when the parameters include multiple gas pressures and multiple gas velocities, the different parameter conditions may refer to situations including different combinations of multiple gas pressures and multiple gas velocities, which will not be described in detail here.

[0090] According to the above technical means, the external flow field characteristic data under different circumstances can be obtained based on different parameter conditions, the amount and type of characteristic data can be increased, and the accuracy of model training can be improved.

[0091] In one possible implementation, Figure 2 As shown, the above step S101 may include the following steps.

[0092] S1011. Simulate the external flow field of the vehicle and extract initial external flow field characteristic data at multiple moments from the simulation results.

[0093] The initial external flow field characteristic data includes data of n dimensions, where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results.

[0094] For example, the external flow field of the vehicle can be simulated using simulation software to obtain a simulation result of the external flow field of the vehicle. The simulation result can include multiple grids, and each grid can include external flow field characteristic data at different times.

[0095] S1012: Perform dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

[0096] According to the above technical means, the initial external flow field characteristic data at multiple moments can be obtained through software simulation, which expands the amount of characteristic data and improves the data acquisition efficiency; and by performing dimensionality reduction processing on the initial external flow field characteristic data, lower-dimensional data can be obtained, and the external flow field analysis model can be trained using the lower-dimensional data, which can further improve the efficiency of model training.

[0097] In a possible embodiment, the above-mentioned step S1012 may include the following steps: constructing the initial external flow characteristic data of the multiple moments into a first data matrix; performing singular value decomposition on the first data matrix, and using the left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as the first POD modal basis vectors; wherein the set number of singular values ​​includes multiple singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; constructing a second data matrix based on the first POD modal basis vectors; performing dimensionality reduction processing on the second data matrix to obtain a projection matrix; performing dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

[0098] For example, the range of the set number of singular values ​​can be set by the user, for example, 10-30, i.e., 10, 15, 20, or 30, etc., which is not limited here. For example, the set number of singular values ​​can be the top 20 singular values ​​in the decomposed first data matrix, that is, the left singular vectors corresponding to the top 20 singular values ​​in the decomposed first data matrix can be used as the first POD modal basis vectors.

[0099] For example, the multiple moments may refer to the time steps of the flow field grid points in the simulation results; the range of the multiple moments may be set by the user, for example, 80-220, that is, the time steps may be 80, 90, 100, 150, 200 or 220, etc., which is not limited here; the first data matrix may be a linear data matrix.

[0100] For example, the first data matrix can be S∈R^(n×m), where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results, and m is the number of time steps. The range of the set number can be set by the user, for example, 10-30, that is, the set number can be 10, 15, 20, or 30, etc., and is not limited here.

[0101] According to the above technical means, by performing singular value decomposition and modal extraction on the first data matrix, the dimensionality reduction processing of the initial external flow characteristic data is achieved, the external flow field characteristic data is obtained, and the data dimension of the input external flow field analysis model is reduced, thereby improving the data analysis efficiency of the external flow field analysis model.

[0102] In a possible implementation, the above-mentioned dimensionality reduction processing of the second data matrix to obtain the projection matrix may include: performing singular value decomposition and modal extraction on the second data matrix to obtain second POD modal basis vectors; interpolating indexes on the second POD modal basis vectors through the DEIM algorithm to construct the projection matrix, wherein the dimension of the projection matrix is ​​lower than the dimension of the second data matrix.

[0103] For example, the second data matrix may be a nonlinear data matrix. After performing singular value decomposition on the second data matrix, left singular vectors corresponding to a set number of singular values ​​in the decomposed second data matrix may be used as second POD modal basis vectors to achieve modal extraction of the second data matrix.

[0104] For example, the DEIM (DETR with Improved Matching for Fast Convergence) algorithm can be used to interpolate and index the second POD modal basis vector according to a set value to construct the projection matrix. The set value refers to the value corresponding to the number of iterative cycles in the DEIM algorithm when it stops. The set value can be set by the user. For example, the set value can be in the range of 40-60, such as 40, 45, 50 or 60, etc., which is not limited here. In this way, a projection matrix can be constructed based on the second data matrix through the DEIM algorithm, and the data dimension of the second data matrix can be reduced to the data dimension corresponding to the set value, thereby achieving dimensionality reduction processing of the second data matrix.

[0105] like Figure 3 FIG. 1 is a flow chart of a method for constructing a projection matrix using a DEIM algorithm. The method may include the following steps.

[0106] S301, performing singular value decomposition on the second data matrix N,

[0107] S302: Determine the second POD modal basis vector Ξ of the second data matrix p =[ξ1 ξ2…ξ p ].

[0108] S303 . Initialize the first index [ρ, λ1]=max|ξ1|.

[0109] S304, construct the first column P1 of the projection matrix P = [e λ1 ].

[0110] S305 , perform interpolation indexing and set the number of iteration cycles P (j=2, 3, . . . , p).

[0111] The number of iteration cycles refers to the number of iteration cycles in the DEIM algorithm.

[0112] S306, calculate c j and residual R j+1 , R j+1 =ξ j+1 -Ξ j c j .

[0113] S307, determine the maximum residual index [ρ, γ j ]=max|R j+1 |.

[0114] S308. Add a new column to the projection matrix P j+1 =[P j eλj ].

[0115] S309: Determine whether the number of iterative cycles reaches the set value.

[0116] The set value refers to the value corresponding to the stop of the number of loop iterations in the DEIM algorithm, which can be set by the user and is not limited here.

[0117] If it is determined that the number of iterative cycles reaches the set value, step S310 is executed; if it is determined that the number of iterative cycles does not reach the set value, the process returns to step S306.

[0118] S310: Obtain the constructed projection matrix.

[0119] According to the above technical means, the projection matrix can be constructed through the DEIM algorithm to reduce the dimension of the second data matrix.

[0120] In a possible implementation, the above-mentioned dimensionality reduction processing of the projection matrix to obtain the external flow field characteristic data at the multiple moments may include: according to the second POD modal basis vector and the projection matrix, by the equation F≈Ψ(P T Ψ) -1 P T F, reconstruct the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; the dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

[0121] For example, the equation F≈Ψ(P T Ψ) -1 P T F can be a pre-constructed equation that can be directly called upon to perform relevant operations when performing nonlinear term reconstruction on the second data matrix. The Navier-Stokes equations (NS equations) are already disclosed in relevant technical literature and will not be described in detail here. According to the above technical means, the data dimension of the second data matrix can be further reduced by nonlinear term reconstruction to obtain data of lower dimension.

[0122] In one possible embodiment, solving the dimensionality reduction equation based on the third data matrix to obtain the external flow field characteristic data at the multiple moments may include normalizing the solved external flow field characteristic data to obtain the external flow field characteristic data at the multiple moments. For example, each physical quantity in the data, namely, the components of the gas velocity in the x, y, and z directions and the gas pressure, may be normalized to obtain the external flow field characteristic data at the multiple moments. In this way, by normalizing the data, data processing efficiency can be improved, facilitating subsequent model training.

[0123] In one possible implementation, the normalized external flow field characteristic data can be divided to obtain training data and verification data; the external flow field analysis model is trained by the training data, and the external flow field analysis model is verified by the verification data. If the verification passes, the trained external flow field prediction model is obtained.

[0124] For example, the normalized external flow field characteristic data can be randomly divided to obtain the training data and the verification data. For example, 60%-90% of the external flow field characteristic data can be divided into training data, and 10%-40% of the external flow field characteristic data can be divided into verification data.

[0125] According to the above technical features, the external flow field analysis model can be trained and verified through the processed external flow field characteristic data, thereby improving the accuracy of model training.

[0126] like Figure 4 FIG. 1 is a flow chart of a method for predicting external flow field characteristic data, and the method may include the following steps.

[0127] S401. Obtain external flow field characteristic data of a vehicle at multiple moments.

[0128] S402: Input the external flow field characteristic data at the multiple moments into an external flow field analysis model to obtain predicted external flow field characteristic data output by the external flow field analysis model.

[0129] The external flow field analysis model is obtained by training using the above-mentioned model training method.

[0130] According to the above technical means, the external flow field characteristic data can be predicted through the external flow field analysis model, thereby improving the efficiency of data prediction.

[0131] For example, software simulation can be used to obtain simulation results of the vehicle's external flow field. The simulation results can include multiple flow field grid points. The characteristic data of each flow field grid point at multiple time points can be used as input to the external flow field analysis model, and the characteristic data of each flow field grid point output by the external flow field analysis model at future time points can be obtained, thereby realizing data prediction for all flow field grid points in the external flow field. According to the above technical means, the external flow field analysis model can be used to predict the external flow field characteristic data, thereby improving the efficiency of data prediction.

[0132] In one possible embodiment, the external flow field characteristic data of the vehicle at multiple moments are obtained, including: simulating the external flow field of the vehicle and extracting initial external flow field characteristic data at multiple moments from the simulation results; wherein the initial external flow field characteristic data includes data of n dimensions, where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results; and performing dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

[0133] According to the above technical means, the initial external flow field characteristic data at multiple moments can be obtained through software simulation, which expands the data volume and improves the data acquisition efficiency.

[0134] In one possible embodiment, the initial external flow field characteristic data at the multiple moments are subjected to dimensionality reduction processing to obtain the external flow field characteristic data at the multiple moments, including: constructing the initial external flow characteristic data at the multiple moments into a first data matrix; performing singular value decomposition on the first data matrix, and using the left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as first POD modal basis vectors; wherein the set number of singular values ​​includes multiple singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; constructing a second data matrix based on the first POD modal basis vectors; performing dimensionality reduction processing on the second data matrix to obtain a projection matrix; performing dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

[0135] According to the above technical means, the initial external flow characteristic data can be reduced in dimension through singular value decomposition and mode extraction to obtain external flow field characteristic data, thereby reducing the data dimension of the input external flow field analysis model and improving the data analysis efficiency of the external flow field analysis model.

[0136] In one possible implementation, the second data matrix is ​​subjected to dimensionality reduction processing to obtain a projection matrix, including: performing singular value decomposition and modal extraction on the second data matrix to obtain a second POD modal basis vector; interpolating and indexing the second POD modal basis vector through a DEIM algorithm to construct the projection matrix, wherein the dimension of the projection matrix is ​​lower than the dimension of the second data matrix.

[0137] According to the above technical means, the projection matrix can be constructed through the DEIM algorithm to reduce the dimension of the second data matrix.

[0138] In a possible implementation, the projection matrix is ​​subjected to dimensionality reduction processing to obtain the external flow field characteristic data at the multiple moments, including: according to the second POD modal basis vector and the projection matrix, through the equation F≈Ψ(P T Ψ) - 1 P T F, reconstruct the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; the dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

[0139] According to the above technical means, the data dimension of the second data matrix can be further reduced by nonlinear term reconstruction to obtain data with lower dimension.

[0140] In a possible implementation manner, the flow field variable includes at least one of the following: gas velocity and gas pressure.

[0141] In a possible implementation, obtaining the external flow field characteristic data of the vehicle at multiple moments includes: obtaining the external flow field characteristic data at multiple moments under multiple parameter conditions; wherein different parameter conditions have different gas velocities and / or gas pressures.

[0142] In a possible implementation, the external flow field analysis model is constructed based on an LSTM neural network.

[0143] The above mainly introduces the solution provided by the embodiment of the present invention from the perspective of method. In order to realize the above functions, the device for model training, the device for predicting external flow field characteristic data or the electronic device contains hardware structures and / or software modules corresponding to the execution of each function. It should be easy for those skilled in the art to realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present invention can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0144] According to the above method, the embodiment of the present invention can exemplarily divide the functional modules of the device for model training, the device for predicting external flow field characteristic data or the electronic device. For example, the device for model training, the device for predicting external flow field characteristic data or the electronic device can include various functional modules corresponding to the various functional divisions, or two or more functions can be integrated into one processing module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present invention is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0145] like Figure 5 FIG. 5 is a block diagram of a model training apparatus 500. The model training apparatus 500 may include: a first acquisition module 510 and a training module 520;

[0146] The first acquisition module 510 is used to acquire the external flow field characteristic data of the vehicle at multiple moments, and the external flow field characteristic data is used to characterize the distribution of flow field variables in the time and space domain;

[0147] The training module 520 is used to train the external flow field analysis model based on the external flow field characteristic data at the multiple moments to obtain a trained external flow field analysis model; wherein, the external flow field analysis model includes an encoder and a decoder; the encoder is used to convert the external flow field characteristic data into feature data, and obtain predicted feature data based on the feature data; the decoder is used to convert the predicted feature data into predicted external flow field characteristic data; wherein, the dimension of the feature data is lower than the dimension of the external flow field characteristic data.

[0148] The encoder of the external flow field analysis model in the above technical solution is used to realize feature data extraction and feature data prediction, and the predicted feature data is restored by the decoder to obtain predicted external flow field characteristic data, thereby reducing the dimension and complexity of the external flow field characteristic data, reducing the calculation time, and realizing efficient external flow field characteristic data prediction.

[0149] The first acquisition module 510 is used to simulate the external flow field of the vehicle and extract the initial external flow field characteristic data at multiple moments from the simulation results; wherein the initial external flow field characteristic data includes data of n dimensions, where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results; and perform dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

[0150] The first acquisition module 510 is used to construct the initial external flow characteristic data of the multiple moments into a first data matrix; perform singular value decomposition on the first data matrix, and use the left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as the first POD modal basis vectors; wherein the set number of singular values ​​includes multiple singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; construct a second data matrix based on the first POD modal basis vectors; perform dimensionality reduction processing on the second data matrix to obtain a projection matrix; perform dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

[0151] The first acquisition module 510 is used to perform singular value decomposition and modal extraction on the second data matrix to obtain a second POD modal basis vector; interpolation indexing is performed on the second POD modal basis vector through the DEIM algorithm to construct the projection matrix, and the dimension of the projection matrix is ​​lower than the dimension of the second data matrix.

[0152] The first acquisition module 510 is used to obtain the second POD modal basis vector and the projection matrix through the equation F≈Ψ(P T Ψ) -1 P T F, reconstruct the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; the dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

[0153] The flow field variable includes at least one of the following: gas velocity and gas pressure.

[0154] The first acquisition module 510 is used to acquire external flow field characteristic data at multiple moments under multiple parameter conditions; wherein different parameter conditions have different gas velocities and / or gas pressures.

[0155] The external flow field analysis model is built based on the LSTM neural network.

[0156] like Figure 6 , which is a block diagram of a device 600 for predicting external flow field characteristic data. The device 600 may include a second acquisition module 610 and a prediction module 620;

[0157] The second acquisition module 610 is used to acquire the external flow field characteristic data of the vehicle at multiple moments;

[0158] The prediction module 620 is used to input the external flow field characteristic data at multiple moments into the external flow field analysis model to obtain the predicted external flow field characteristic data output by the external flow field analysis model; wherein, the external flow field analysis model is trained by the model training method of the first aspect mentioned above.

[0159] According to the above technical means, the external flow field characteristic data can be predicted through the external flow field analysis model, thereby improving the efficiency of data prediction.

[0160] The second acquisition module 610 is used to simulate the external flow field of the vehicle and extract the initial external flow field characteristic data at multiple moments from the simulation results; wherein the initial external flow field characteristic data includes data of n dimensions, and n is the product of the number of flow field grid points and the number of flow field variables in the simulation results; and perform dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

[0161] The second acquisition module 610 is used to construct the initial external flow characteristic data of the multiple moments into a first data matrix; perform singular value decomposition on the first data matrix, and use the left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as the first POD modal basis vectors; wherein the set number of singular values ​​includes multiple singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; construct a second data matrix based on the first POD modal basis vectors; perform dimensionality reduction processing on the second data matrix to obtain a projection matrix; perform dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

[0162] The second acquisition module 610 is used to perform singular value decomposition and modal extraction on the second data matrix to obtain the second POD modal basis vector; interpolation indexing of the second POD modal basis vector is performed through the DEIM algorithm to construct the projection matrix, and the dimension of the projection matrix is ​​lower than the dimension of the second data matrix.

[0163] The second acquisition module 610 is used to obtain the second POD modal basis vector and the projection matrix through the equation F≈Ψ(P T Ψ) -1 P T F, reconstruct the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; the dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

[0164] The flow field variable includes at least one of the following: gas velocity and gas pressure.

[0165] The second acquisition module 610 is used to acquire external flow field characteristic data at multiple moments under multiple parameter conditions; wherein different parameter conditions have different gas velocities and / or gas pressures.

[0166] The external flow field analysis model is built based on the LSTM neural network.

[0167] According to an embodiment of the present invention, the present invention further provides an electronic device, a readable storage medium, and a computer program product.

[0168] In an exemplary embodiment, the electronic device may include a memory and a processor, the memory and the processor are connected, and the processor is used to execute the above-mentioned model training method or external flow field characteristic data prediction method stored in the memory.

[0169] In an exemplary embodiment, the readable storage medium may be a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute the model training method or the external flow field characteristic data prediction method according to the above embodiments.

[0170] In an exemplary embodiment, a computer program product includes a computer program, which, when executed by a processor, implements the model training method or the external flow field characteristic data prediction method according to the above embodiments.

[0171] like Figure 7 , a block diagram of an example electronic device 700 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0172] like Figure 7As shown, the electronic device 700 may include a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. Various programs and data required for the operation of the electronic device 700 may also be stored in the RAM 703. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0173] Multiple components in the electronic device 700 are connected to the I / O interface 705, including an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the electronic device 700 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0174] The computing unit 701 can be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as the flow control method. For example, in some embodiments, the flow control method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the flow control method described above can be performed. Alternatively, in other embodiments, the computing unit 701 can be configured to perform the flow control method in any other appropriate manner (e.g., by means of firmware).

[0175] like Figure 8 As shown, it is a block diagram of a vehicle 800 , and the vehicle 800 may include the above-mentioned prediction device 600 for external flow field characteristic data.

[0176] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0177] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0178] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in the present invention can be achieved. This is not limited herein.

[0179] The above embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Any equivalent substitution or modification made by those skilled in the art based on the present invention is within the protection scope of the present invention.

Claims

1. A model training method, characterized in that: The method comprises: Acquiring external flow field characteristic data of the vehicle at multiple moments, wherein the external flow field characteristic data is used to characterize the distribution of flow field variables in the time and space domain; Based on the external flow field characteristic data at the multiple moments, the external flow field analysis model is trained to obtain a trained external flow field analysis model; wherein, the external flow field analysis model includes an encoder and a decoder; the encoder is used to convert the external flow field characteristic data into feature data, and obtain predicted feature data based on the feature data; the decoder is used to convert the predicted feature data into predicted external flow field characteristic data; wherein, the dimension of the feature data is lower than the dimension of the external flow field characteristic data.

2. The method according to claim 1, characterized in that The obtaining of the external flow field characteristic data of the vehicle at multiple moments includes: By simulating the external flow field of the vehicle, initial external flow field characteristic data at multiple moments are extracted from the simulation results; wherein the initial external flow field characteristic data includes data of n dimensions, where n is the product of the number of flow field grid points and the number of flow field variables in the simulation results; Dimensionality reduction processing is performed on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments.

3. The method according to claim 2, characterized in that The performing dimensionality reduction processing on the initial external flow field characteristic data at the multiple moments to obtain the external flow field characteristic data at the multiple moments includes: constructing the initial outflow characteristic data at the plurality of moments into a first data matrix; Performing singular value decomposition on the first data matrix, and using left singular vectors corresponding to a set number of singular values ​​in the decomposed first data matrix as first POD modal basis vectors; wherein the set number of singular values ​​includes a plurality of singular values ​​ranked according to the numerical values ​​of the singular values ​​in the decomposed first data matrix; Constructing a second data matrix based on the first POD modal basis vector; Performing dimensionality reduction processing on the second data matrix to obtain a projection matrix; Performing dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the multiple moments.

4. The method according to claim 3, characterized in that The performing dimensionality reduction processing on the second data matrix to obtain a projection matrix includes: Performing singular value decomposition and mode extraction on the second data matrix to obtain a second POD modal basis vector; The second POD modal basis vector is interpolated and indexed by the DEIM algorithm to construct the projection matrix, where the dimension of the projection matrix is ​​lower than the dimension of the second data matrix.

5. The method according to claim 3, characterized in that The performing dimensionality reduction processing on the projection matrix to obtain the external flow field characteristic data at the plurality of moments includes: According to the second POD modal basis vector and the projection matrix, the equation F≈Ψ(P T Ψ) -1 P T F, reconstructs the second data matrix by nonlinear terms to obtain a third data matrix; the dimension of the third data matrix is ​​lower than that of the projection matrix; wherein F is the second data matrix, P is the projection matrix, and Ψ is the second POD modal basis vector; The dimensionality reduction equation is solved according to the third data matrix to obtain the external flow field characteristic data at the multiple moments; wherein the dimensionality reduction equation is obtained by projecting the NS equation on the POD modal space based on the set number of singular values.

6. The method according to claim 1, characterized in that The flow field variables include at least one of the following: gas velocity and gas pressure.

7. The method according to claim 1, characterized in that The obtaining of the external flow field characteristic data of the vehicle at multiple moments includes: External flow field characteristic data at multiple moments under multiple parameter conditions are obtained; wherein different parameter conditions have different gas velocities and / or gas pressures.

8. A method for predicting external flow field characteristic data, characterized in that: The method comprises: Acquiring external flow field characteristic data of the vehicle at multiple times; The external flow field characteristic data at the multiple moments are input into the external flow field analysis model to obtain the predicted external flow field characteristic data output by the external flow field analysis model; wherein, the external flow field analysis model is trained by the model training method described in any one of claims 1 to 7.

9. A model training device, characterized in that: The device comprises: A first acquisition module is used to acquire external flow field characteristic data of the vehicle at multiple moments, wherein the external flow field characteristic data is used to characterize the distribution of flow field variables in the time and space domain; A training module is used to train an external flow field analysis model based on the external flow field characteristic data at the multiple moments to obtain a trained external flow field analysis model; wherein, the external flow field analysis model includes an encoder and a decoder; the encoder is used to convert the external flow field characteristic data into feature data, and obtain predicted feature data based on the feature data; the decoder is used to convert the predicted feature data into predicted external flow field characteristic data; wherein, the dimension of the feature data is lower than the dimension of the external flow field characteristic data.

10. A device for predicting external flow field characteristic data, characterized in that: The device comprises: A second acquisition module is used to acquire external flow field characteristic data of the vehicle at multiple moments; A prediction module is used to input the external flow field characteristic data at the multiple moments into an external flow field analysis model to obtain the predicted external flow field characteristic data output by the external flow field analysis model; wherein, the external flow field analysis model is trained by the model training method described in any one of claims 1 to 7.

11. An electronic device, characterized in that: It includes a memory and a processor, the memory and the processor are connected, and the processor is used to execute the model training method described in any one of claims 1 to 7 or the external flow field characteristic data prediction method described in claim 8 stored in the memory.

12. A vehicle, characterized in that: The vehicle includes the device for predicting external flow field characteristic data according to claim 10 .

Images