Performance prediction model training and performance prediction method, device, equipment, medium and program

Abstract

Embodiments of the present application disclose a performance prediction model training and performance prediction method, device, equipment, medium and program, comprising: generating a model training sample set according to reference design variables of each rod of a vehicle body framework; determining a model topology structure of a first vehicle body framework performance prediction model and a second vehicle body framework performance prediction model; iteratively updating model parameters of the first vehicle body framework performance prediction model according to the model training sample set by using a local search updating algorithm, and screening a first target vehicle body framework performance prediction model; iteratively updating model parameters of the second vehicle body framework performance prediction model according to the model training sample set by using a global search updating algorithm, and screening a second target vehicle body framework performance prediction model; and generating a target vehicle body framework performance prediction model according to the first target vehicle body framework performance prediction model and the second target vehicle body framework performance prediction model.

Description

Technical Field

[0001] The embodiments of the present invention relate to the fields of vehicle frame structure optimization and swarm intelligence optimization technology, and particularly to a performance prediction model training method, performance prediction method, device, electronic device, storage medium and program. Background Technology

[0002] The vehicle body frame is a crucial component that bears the weight of the entire vehicle and transmits ground loads. Optimizing the design of the vehicle body frame structure often requires comprehensive consideration of performance indicators such as bending stiffness, torsional stiffness, and low-order modal frequencies, making it an expensive, multidisciplinary optimization problem.

[0003] Currently, finite element model (FEM), response surface methodology (RSM), Kriging model, and neural network model are commonly used for optimizing vehicle body frame structures. Solving these optimization models often requires multiple iterative analyses. If the FEM model is directly called in each iteration, it will result in a significant waste of computational resources and time. RSM, Kriging, and neural network models are three commonly used surrogate models. Compared to RSM and Kriging models, neural network models offer better prediction accuracy. However, since the accuracy of single neural network models is still not ideal, using a single neural network model for vehicle body optimization will significantly reduce the convergence speed. Summary of the Invention

[0004] This invention provides a performance prediction model training method, performance prediction method, device, electronic device, storage medium, and program, which can improve the training efficiency and performance prediction accuracy of the vehicle body frame performance prediction model.

[0005] According to one aspect of the present invention, a method for training a performance prediction model is provided, comprising: A model training sample set is generated based on the reference design variables of each member of the vehicle body frame; The model topology of the first body frame performance prediction model and the second body frame performance prediction model is determined; wherein the first body frame performance prediction model and the second body frame performance prediction model are used to predict multidimensional body frame performance indicators. The local search update algorithm is used to iteratively update the model parameters of the first vehicle body frame performance prediction model based on the model training sample set, and the first target vehicle body frame performance prediction model is selected based on the iterative update results. A global search update algorithm is used to iteratively update the model parameters of the second body frame performance prediction model based on the model training sample set, and the second target body frame performance prediction model is selected based on the iterative update results. A target body frame performance prediction model is generated based on the first target body frame performance prediction model and the second target body frame performance prediction model.

[0006] According to another aspect of the present invention, a performance prediction method is provided, comprising: Obtain the target design variables for each member of the target vehicle body frame; The target body frame performance prediction model is used to predict the multidimensional body frame performance index of the target body frame based on the target design variables, and the prediction result of the multidimensional body frame performance index is obtained. The target body frame performance prediction model includes a first body frame performance prediction model and a second body frame performance prediction model; the first body frame performance prediction model and the second body frame performance prediction model are trained by the performance prediction model training method described in any embodiment of the present invention.

[0007] According to another aspect of the present invention, a performance prediction model training apparatus is provided, comprising: The model training sample set generation module is used to generate a model training sample set based on the reference design variables of each member of the vehicle body frame. The model topology determination module is used to determine the model topology of the first body frame performance prediction model and the second body frame performance prediction model; wherein, the first body frame performance prediction model and the second body frame performance prediction model are used to predict multi-dimensional body frame performance indicators. The first target body frame performance prediction model training module is used to iteratively update the model parameters of the first body frame performance prediction model according to the model training sample set using a local search update algorithm, and to select the first target body frame performance prediction model according to the iterative update results. The training module for the second target body frame performance prediction model is used to iteratively update the model parameters of the second body frame performance prediction model based on the model training sample set using a global search update algorithm, and to select the second target body frame performance prediction model based on the iterative update results. The target body frame performance prediction model generation module is used to generate a target body frame performance prediction model based on the first target body frame performance prediction model and the second target body frame performance prediction model.

[0008] According to another aspect of the present invention, a performance prediction apparatus is provided, characterized in that it comprises: The target design variable acquisition module is used to acquire the target design variables of each member of the target vehicle body frame; The multi-dimensional body frame performance index prediction module is used to predict the multi-dimensional body frame performance index of the target body frame according to the target design variables through the target body frame performance prediction model, and obtain the multi-dimensional body frame performance index prediction result. The target body frame performance prediction model includes a first body frame performance prediction model and a second body frame performance prediction model; the first body frame performance prediction model and the second body frame performance prediction model are trained by the performance prediction model training method described in any embodiment of the present invention.

[0009] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and a memory communicatively connected to said at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to execute the performance prediction model training method or performance prediction method according to any embodiment of the present invention.

[0010] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement the performance prediction model training method or performance prediction method according to any embodiment of the present invention.

[0011] According to another aspect of the present invention, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the performance prediction model training method or performance prediction method described in any embodiment of the present invention.

[0012] This invention generates a model training sample set based on the reference design variables of each member of the vehicle body frame, and determines the model topology of a first and a second vehicle body frame performance prediction model used to predict multi-dimensional vehicle body frame performance indicators. Further, a local search update algorithm is used to iteratively update the model parameters of the first vehicle body frame performance prediction model based on the model training sample set, and a first target vehicle body frame performance prediction model is selected based on the iterative update results. Simultaneously, a global search update algorithm is used to iteratively update the model parameters of the second vehicle body frame performance prediction model based on the model training sample set, and a second target vehicle body frame performance prediction model is selected based on the iterative update results. Finally, a target vehicle body frame performance prediction model is generated based on the first and second target vehicle body frame performance prediction models. The trained target vehicle body frame performance prediction model can be used to predict the multi-dimensional performance indicators of the vehicle body frame. This technical solution effectively reduces the computational load during model training by utilizing the reference design variables selected from each member of the vehicle body frame to generate the model training sample set. Meanwhile, the first vehicle body frame performance prediction model employs an update algorithm focused on local search during parameter optimization training, which significantly improves the model's linear prediction capability. The second vehicle body frame performance prediction model, on the other hand, uses an update algorithm focused on global search during parameter optimization training, which improves the model's nonlinear prediction capability while reducing computational cost. By using two different model parameter update algorithms for parameter optimization during model training, the overall computational load of the vehicle body frame performance prediction model can be effectively reduced while maintaining its accuracy, thereby improving the training efficiency and performance prediction accuracy of the vehicle body frame performance prediction model.

[0013] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of a performance prediction model training method provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram illustrating the three-dimensional structure of a vehicle frame and the corresponding positions of design variables provided in an embodiment of the present invention. Figure 3 This is a flowchart of a performance prediction model training method provided in Embodiment 2 of the present invention; Figure 4 This is a flowchart illustrating a performance prediction model training method provided in Embodiment 2 of the present invention; Figure 5 This is a schematic diagram of the sensitivity calculation results of the bending stiffness of a vehicle frame provided in Embodiment 2 of the present invention; Figure 6 This is a schematic diagram of the sensitivity calculation results of the torsional stiffness of a vehicle frame provided in Embodiment 2 of the present invention; Figure 7 This is a schematic diagram of the sensitivity calculation results of the first-order torsional mode frequency of a vehicle frame provided in Embodiment 2 of the present invention; Figure 8 This is a schematic diagram of the sensitivity calculation results of the first-order bending mode frequency of a vehicle frame provided in Embodiment 2 of the present invention; Figure 9 This is a schematic diagram of the sensitivity calculation results of the vehicle body frame mass provided in Embodiment 2 of the present invention; Figure 10 This is a schematic diagram of the model topology of a first vehicle body frame performance prediction model provided in Embodiment 2 of the present invention; Figure 11 This is a schematic diagram of the model topology of a second vehicle body frame performance prediction model provided in Embodiment 2 of the present invention; Figure 12 This is a flowchart of a performance prediction method provided in Embodiment 3 of the present invention; Figure 13 This is a schematic diagram of a performance prediction model training device provided in Embodiment 4 of the present invention; Figure 14 This is a schematic diagram of a performance prediction device provided in Embodiment 5 of the present invention; Figure 15 This is a schematic diagram of the structure of an electronic device provided in Embodiment Six of the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] Example 1 Figure 1 This is a flowchart of a performance prediction model training method provided in Embodiment 1 of the present invention. This embodiment is applicable to situations where different vehicle body frame performance prediction models are trained using different parameter update algorithms based on a model training sample set generated from selected reference design variables. This method can be executed by a performance prediction model training device, which can be implemented in software and / or hardware, and is generally integrated into an electronic device. This electronic device can be a terminal device or a server device, as long as it can execute the performance prediction model training method. The present invention does not limit the specific type of electronic device. Correspondingly, as... Figure 1 As shown, the method includes the following operations: S110. Generate a model training sample set based on the reference design variables of each member of the vehicle body frame.

[0019] The vehicle body frame can be the frame structure of any type of vehicle, including but not limited to unmanned inspection vehicles, cars, buses, and trucks, as well as the frame structure of subway or high-speed rail carriages. This embodiment of the invention does not limit the vehicle type to which the vehicle body frame belongs. Optionally, the design variables for each member of the vehicle body frame can be the member thickness parameters at corresponding positions. The reference design variables can be a subset of design variables selected from all design variables at all positions of the members of the vehicle body frame. The model training sample set can be a sample set used to train the first vehicle body frame performance prediction model and the second vehicle body frame performance prediction model.

[0020] Figure 2 This is a schematic diagram illustrating the three-dimensional structure of a vehicle body frame and the corresponding positions of design variables according to an embodiment of the present invention. In a specific example, such as... Figure 2As shown, various locations within the vehicle body frame can include multiple types of design variables. Let X represent the design variables for each member of the vehicle body frame, where the value can be the thickness of the corresponding member, such as 3mm. Find X represents the set of design variables for each member of the vehicle body frame. For example, as... Figure 2 As shown, the design variables for each member of the vehicle body frame can include a total of 16 positional parameters, namely... It is understandable that only the positions of certain members in a vehicle body frame have a significant impact on the overall performance of the frame. Therefore, to improve model training efficiency, design variables that significantly contribute to the performance of the vehicle body frame can be selected from the design variables of each member as reference design variables, thereby reducing the input dimensions of the model. For example, one could select from the design variables of each member... Ten design variables that significantly contribute to the performance of the vehicle body frame were selected as reference design variables.

[0021] Accordingly, after selecting reference design variables based on the design variables of each member of the vehicle body frame, a model training sample set for model training can be generated based on the selected reference design variables. Specifically, multiple sets of selectable values ​​can be configured for the selected reference design variables, and then simulations can be performed on various vehicle body frame performance indicators based on each set of reference design variables to obtain the multi-dimensional vehicle body frame performance indicator simulation values, i.e., performance simulation data, corresponding to each set of reference design variables. Each set of reference design variables and its corresponding multi-dimensional vehicle body frame performance indicator simulation values ​​(performance simulation data) can together constitute the model training sample set. Among them, the multi-dimensional vehicle body frame performance indicator simulation values ​​corresponding to each set of reference design variables can be used to evaluate the accuracy of model training.

[0022] S120. Determine the model topology of the first body frame performance prediction model and the second body frame performance prediction model; wherein, the first body frame performance prediction model and the second body frame performance prediction model are used to predict multidimensional body frame performance indicators.

[0023] The first vehicle body frame performance prediction model can be a surrogate model composed of a neural network model. The second vehicle body frame performance prediction model can also be a surrogate model composed of a neural network model. The multi-dimensional vehicle body frame performance indicators can include various different types of vehicle body frame performance indicators, such as, but not limited to, frame bending stiffness, torsional stiffness, first-order torsional mode frequency, first-order bending mode frequency, and vehicle mass. This embodiment of the invention does not limit the types or number of indicators included in the multi-dimensional vehicle body frame performance indicators.

[0024] In this embodiment of the invention, two different prediction models, namely a first vehicle body frame performance prediction model and a second vehicle body frame performance prediction model, can be used to jointly constitute a surrogate model for predicting multi-dimensional vehicle body frame performance indicators. Optionally, the first and second vehicle body frame performance prediction models can be of the same type, for example, both can be neural network models, but with different model topologies. That is, the multi-dimensional performance of the vehicle body frame can be predicted by combining the advantages of the two models through surrogate models with different model topologies. Optionally, the first vehicle body frame performance prediction model can focus on the prediction accuracy of multi-dimensional vehicle body frame performance indicators, while the second vehicle body frame performance prediction model can reduce computational consumption while focusing on the prediction accuracy of multi-dimensional vehicle body frame performance indicators.

[0025] S130. The local search update algorithm is used to iteratively update the model parameters of the first vehicle body frame performance prediction model according to the model training sample set, and the first target vehicle body frame performance prediction model is selected according to the iterative update results.

[0026] Among them, the first target body frame performance prediction model can be the first body frame performance prediction model with the highest prediction accuracy selected during the training process of the first body frame performance prediction model.

[0027] In this embodiment of the invention, during the optimization of model parameters in the first vehicle body frame performance prediction model, a model parameter update algorithm focusing on local search can be used to update the model parameters of the first vehicle body frame performance prediction model, and the model training sample set can be used to train the first vehicle body frame performance prediction model. The core principle of the local search update algorithm is a greedy iteration based on neighborhood: starting from the initial solution of the model parameters, a better solution is found in the predefined "neighborhood" of the model parameters and the current solution is replaced until no improvement can be made. Its essence is to quickly converge to the local optimum through local development. The core advantage of the local search update algorithm lies in its extremely high computational efficiency and powerful local fine-tuning ability. Using the local search update algorithm to optimize the model parameters can improve the linear prediction ability of the first vehicle body frame performance prediction model.

[0028] Understandably, the accuracy of the first vehicle body frame performance prediction model varies after multiple iterations of training. Therefore, after the first vehicle body frame performance prediction model has been iteratively trained, the model with the highest prediction accuracy from all iterations can be selected as the final target vehicle body frame performance prediction model.

[0029] S140. The global search update algorithm is used to iteratively update the model parameters of the second vehicle body frame performance prediction model based on the model training sample set, and the second target vehicle body frame performance prediction model is selected based on the iterative update results.

[0030] Among them, the second target body frame performance prediction model can be the second body frame performance prediction model with the highest prediction accuracy selected during the training process of the second body frame performance prediction model.

[0031] In this embodiment of the invention, during the optimization of model parameters for the second vehicle body frame performance prediction model, a model parameter update algorithm focusing on global search can be used to update the model parameters of the second vehicle body frame performance prediction model, and the model training sample set can be used to train the second vehicle body frame performance prediction model. The core principle of the global search update algorithm is to determine the direction of model parameter updates based on the individual historical optimal experience and the group's global optimal experience, balancing the "exploration" of new regions with the "development" of known high-quality regions. The core advantage of the global search update algorithm is that it actively escapes local extreme points, prevents the model from locking into suboptimal parameter configurations too early, and can perform extensive sampling throughout the entire parameter space, thereby improving the nonlinear prediction capability of the second target vehicle body frame performance prediction model while reducing computational consumption.

[0032] Understandably, the accuracy of the second vehicle body frame performance prediction model varies after multiple iterations of training. Therefore, after the second vehicle body frame performance prediction model has been iteratively trained, the model with the highest prediction accuracy from all iterations can be selected as the final target vehicle body frame performance prediction model.

[0033] S150. Generate a target body frame performance prediction model based on the first target body frame performance prediction model and the second target body frame performance prediction model.

[0034] Among them, the target body frame performance prediction model can be the final model used to predict multi-dimensional body frame performance index data.

[0035] After training and obtaining the first and second target vehicle body frame performance prediction models, a further target vehicle body frame performance prediction model can be constructed based on these models. Since the multi-dimensional vehicle body frame performance index includes various types of vehicle body frame performance indices, each type of vehicle body frame performance index can correspond to a specific target vehicle body frame performance prediction model. That is, a single target vehicle body frame performance prediction model is used to predict only one type of vehicle body frame performance index data.

[0036] Optionally, the weights corresponding to the first target vehicle body frame performance prediction model and the second target vehicle body frame performance prediction model can be determined respectively, and then the first target vehicle body frame performance prediction model and the second target vehicle body frame performance prediction model can be weighted and fused to obtain the target vehicle body frame performance prediction model. The weights configured for the first target vehicle body frame performance prediction model and the second target vehicle body frame performance prediction model may also be different in different target vehicle body frame performance prediction models. Among them, the minimum value of the weights configured for the weighted fusion of the first target vehicle body frame performance prediction model and the second target vehicle body frame performance prediction model can be 0, and the maximum value can be 1. The target vehicle body frame performance prediction model trained by the above model training method has the advantages of high training efficiency and low computational cost. The trained target vehicle body frame performance prediction model has good generalization ability and can accurately predict various performance indicators of the vehicle body frame structure, which has great practical engineering application value.

[0037] It should be noted that there is no sequential relationship between steps S110 and S120. That is, step S110 can be executed first and then step S120, or step S120 can be executed first and then step S110, or steps S110 and S120 can be executed simultaneously. This embodiment of the invention does not limit this.

[0038] This invention generates a model training sample set based on the reference design variables of each member of the vehicle body frame, and determines the model topology of a first and a second vehicle body frame performance prediction model used to predict multi-dimensional vehicle body frame performance indicators. Further, a local search update algorithm is used to iteratively update the model parameters of the first vehicle body frame performance prediction model based on the model training sample set, and a first target vehicle body frame performance prediction model is selected based on the iterative update results. Simultaneously, a global search update algorithm is used to iteratively update the model parameters of the second vehicle body frame performance prediction model based on the model training sample set, and a second target vehicle body frame performance prediction model is selected based on the iterative update results. Finally, a target vehicle body frame performance prediction model is generated based on the first and second target vehicle body frame performance prediction models. The trained target vehicle body frame performance prediction model can be used to predict the multi-dimensional performance indicators of the vehicle body frame. This technical solution effectively reduces the computational load during model training by utilizing the reference design variables selected from each member of the vehicle body frame to generate the model training sample set. Meanwhile, the first vehicle body frame performance prediction model employs an update algorithm focused on local search during parameter optimization training, which significantly improves the model's linear prediction capability. The second vehicle body frame performance prediction model, on the other hand, uses an update algorithm focused on global search during parameter optimization training, which improves the model's nonlinear prediction capability while reducing computational cost. By using two different model parameter update algorithms for parameter optimization during model training, the overall computational load of the vehicle body frame performance prediction model can be effectively reduced while maintaining its accuracy, thereby improving the training efficiency and performance prediction accuracy of the vehicle body frame performance prediction model.

[0039] Example 2 Figure 3 This is a flowchart of a performance prediction model training method provided in Embodiment 2 of the present invention. Figure 4 This is a flowchart illustrating a performance prediction model training method according to Embodiment 2 of the present invention. This embodiment is a specific implementation based on the above embodiment. In this embodiment, various specific optional implementation methods are given for generating a model training sample set, determining the model topology, iteratively updating the model parameters, and generating a target vehicle body frame performance prediction model. Correspondingly, as... Figure 3 and Figure 4 As shown, the method in this embodiment may include: S310. Determine the reference design variables of the vehicle body frame based on the multi-dimensional sensitivity analysis results of each member of the vehicle body frame.

[0040] Among them, the results of multidimensional sensitivity analysis can include analysis results of multiple different types of sensitivity.

[0041] In this embodiment of the invention, the sensitivity of various types of links at the positions of the vehicle frame can be analyzed using the finite element analysis model of the vehicle frame. Based on the results of the sensitivity analysis of various types of links at the positions of the vehicle frame, the design variables of the links that contribute significantly to the performance of the vehicle frame can be selected from all design variables as reference design variables.

[0042] In an optional embodiment of the present invention, determining the reference design variables of the vehicle frame based on the multi-dimensional sensitivity analysis results of each member of the vehicle frame may include: calculating the single-dimensional sensitivity analysis results at the positions of each target member of the vehicle frame, and determining the number of reference design variables to be selected for each single-dimensional sensitivity analysis result; and sequentially selecting each reference design variable from the positions of each target member of the vehicle frame according to the number of reference design variables to be selected for each single-dimensional sensitivity analysis result based on the single-dimensional sensitivity analysis results at the positions of each target member of the vehicle frame.

[0043] Specifically, the first step is to calculate the single-dimensional sensitivity analysis results at the locations of each target member in the vehicle body frame. For example, the sensitivity of a member to the bending stiffness of the vehicle body frame can be obtained using the following formula: in, For the bending stiffness of the car body frame. For the combined force loaded onto the vehicle body frame, This represents the maximum displacement along the Z-axis. Let i be the i-th design variable. Figure 2 As shown, assuming a total of 16 design variables, the calculated results of the sensitivity of the members to the bending stiffness of the vehicle body frame are as follows. Figure 5 As shown.

[0044] For example, the sensitivity of a member to the bending stiffness of the vehicle body frame can be derived using the following formula: In the formula, For the torsional stiffness of the car body frame The wheelbase is the distance between the wheels. A pair of force couples applied at the front wheels. This represents the maximum displacement along the Z-axis. Let i be the i-th design variable. Figure 2 As shown, assuming a total of 16 design variables, the sensitivity calculation results of the members to the torsional stiffness of the vehicle body frame are as follows: Figure 6 As shown.

[0045] For example, the sensitivity of a link to the low-order modal frequencies of the vehicle body frame can be obtained by the following formula: In the formula, K is the overall stiffness matrix of the vehicle body frame, M is the overall mass matrix of the vehicle body frame, and γ is the square of the characteristic frequency of the vehicle body frame. The characteristic mode shape matrix, Let i be the i-th design variable. Figure 2 As shown, assuming a total of 16 design variables, the calculated sensitivity of the members to the first-order torsional modal frequency of the vehicle body frame is as follows: Figure 7 As shown, the calculated sensitivity of the member to the first-order bending mode frequency of the vehicle body frame is as follows: Figure 8 As shown, the sensitivity calculation results of the rods to the mass of the vehicle body frame are as follows: Figure 9 As shown.

[0046] Sensitivity analysis results for each dimension can select one or more reference design variables. The number of reference design variables selected for sensitivity analysis results in different dimensions may be the same or different. For example, the number of reference design variables selected for the sensitivity configuration of the body frame bending stiffness may be 2, and the number of reference design variables selected for the sensitivity configuration of the body frame first-order bending modal frequency may be 4, etc. Optionally, the weight of each single-dimensional sensitivity analysis result can be determined based on its importance. Then, the number of reference design variables selected can be determined based on the weight matched to the single-dimensional sensitivity analysis results according to the total number of reference design variables. Then, based on the single-dimensional sensitivity analysis results at each target member position of the body frame, the design variables with the highest single-dimensional sensitivity analysis results are selected as reference design variables according to the number of reference design variables matched to each single-dimensional sensitivity analysis result. For example, the sensitivity analysis results of each member on the body frame bending stiffness, torsional stiffness, low-order modal frequency, and vehicle mass can be sorted, and the members with the highest number of selections for each sensitivity analysis result can be selected as reference design variables.

[0047] For example, such as Figure 2 As shown, the components are ranked according to their contribution to the bending stiffness, torsional stiffness, low-order modal frequencies, and overall vehicle mass, and the top ten components in terms of contribution are selected as reference design variables. The final selected component number is... The member is used as a reference design variable, denoted as: .

[0048] S320. Generate the model training sample set based on the reference design variables and multi-dimensional vehicle body frame performance indicators.

[0049] Specifically, the input sample values ​​of the model training sample set can be determined based on the values ​​of the reference design variables, and the output sample values ​​of the model training sample set can be determined based on the simulated values ​​of each vehicle body frame performance index corresponding to the reference design variables. One input sample value and its corresponding output sample value constitute a complete sample. For example, the values ​​of a set of 10 reference design variables and the simulated values ​​of the corresponding 5 dimensions of vehicle body frame performance indices constitute a complete sample. The output sample values ​​of the model training sample set are used to evaluate the accuracy of the prediction results of the input sample values. Furthermore, the number of samples in the model training sample set can be determined based on the number of performance index types included in the multi-dimensional vehicle body frame performance indices. For example, the number of samples in the model training sample set can be N = 100n, where N represents the number of samples in the model training sample set, and n represents the number of vehicle body frame performance indices. For example, n can be 5.

[0050] S330. Determine the model topology of the first body frame performance prediction model and the second body frame performance prediction model.

[0051] In an optional embodiment of the present invention, determining the model topology of the first body frame performance prediction model and the second body frame performance prediction model may include: configuring the input layers of the first body frame performance prediction model and the second body frame performance prediction model according to the number of reference design variables; configuring the output layers of the first body frame performance prediction model and the second body frame performance prediction model according to the number of body frame performance indicators; configuring the number of hidden layers and hidden factors of the first body frame performance prediction model and the second body frame performance prediction model; wherein the number of hidden layers of the first body frame performance prediction model is a first number, the number of hidden layers of the second body frame performance prediction model is a second number, and the first number is less than the second number.

[0052] Figure 10 This is a schematic diagram of the model topology of a first vehicle body frame performance prediction model provided in Embodiment 2 of the present invention. Figure 11 This is a schematic diagram of the model topology of a second vehicle body frame performance prediction model provided in Embodiment 2 of the present invention. In a specific example, such as Figure 10 and Figure 11As shown, assuming the number of reference design variables is 10, the input layer of the first and second vehicle body frame performance prediction models can include 10 neurons, each receiving the value of one reference design variable. Simultaneously, the output layers of the first and second vehicle body frame performance prediction models can be configured according to the number of vehicle body frame performance indicators. Assuming the number of vehicle body frame performance indicators is 5, namely bending stiffness, torsional stiffness, first-order torsional modal frequency, first-order bending modal frequency, and vehicle mass, the output layer of the first and second vehicle body frame performance prediction models can include 5 neurons, each corresponding to the prediction result of one vehicle body frame performance indicator.

[0053] The significant difference between the first and second vehicle body frame performance prediction models lies in the hidden layer. For example, such as... Figure 10 As shown, the first vehicle body frame performance prediction model can have only one hidden layer. One hidden layer provides high prediction accuracy for performance indicators with low nonlinearity. Figure 11 As shown, the first vehicle body frame performance prediction model can have two hidden layers. Two hidden layers provide higher prediction accuracy for performance indicators with high nonlinearity. Assuming the number of hidden factors in the first vehicle body frame performance prediction model is *a*, and the number of hidden factors in the second vehicle body frame performance prediction model is *b*, then *a* = 2*b*. Optionally, *a* can be 16, and *b* can be 8.

[0054] The first and second vehicle body frame performance prediction models, constructed with different numbers of hidden layers, exhibit different predictive characteristics. Correspondingly, the target vehicle body frame performance prediction model improves the overall robustness and accuracy by integrating the advantages of neural network proxy models with different hidden layers.

[0055] S340. The local search update algorithm is used to iteratively update the model parameters of the first vehicle body frame performance prediction model according to the model training sample set, and the first target vehicle body frame performance prediction model is selected according to the iterative update results.

[0056] In an optional embodiment of the present invention, the step of iteratively updating the model parameters of the first vehicle body frame performance prediction model using the local search update algorithm based on the model training sample set may include: determining the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration round using the local search update algorithm; determining the current vehicle body frame performance prediction result of the first vehicle body frame performance prediction model in the current iteration round based on the model training sample set; evaluating the prediction accuracy of the current vehicle body frame performance prediction result of the first vehicle body frame performance prediction model based on the parameter optimization objective function of the first vehicle body frame performance prediction model; and returning to execute the operation of determining the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration round using the local search update algorithm, until it is determined that the current iteration round has reached a preset number of iterations.

[0057] The current first model optimization parameters can be the applicable model parameters of the first vehicle body frame performance prediction model in the current iteration round. The current vehicle body frame performance prediction result can be the model's prediction result for each vehicle body frame performance index in the current iteration round. The parameter optimization objective function can be used to evaluate the accuracy of the model's prediction results for the vehicle body frame performance index. The preset number of iterations can be the maximum number of iterations.

[0058] Specifically, a local search update algorithm can be used to determine the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration. These current first model optimization parameters are then used to assign values ​​to the model parameters of the first vehicle body frame performance prediction model, resulting in the first vehicle body frame performance prediction model for the current iteration. Further, the input sample values ​​from the model training sample set are input into the first vehicle body frame performance prediction model. For example, the values ​​of 10 reference design variables are input into the corresponding neurons of the first vehicle body frame performance prediction model, obtaining the prediction results of multiple vehicle body frame performance indicators output by the output layer of the first vehicle body frame performance prediction model. These prediction results serve as the current vehicle body frame performance prediction result for the first vehicle body frame performance prediction model in the current iteration. Further, based on the parameter optimization objective function of the first vehicle body frame performance prediction model, the prediction accuracy of the current vehicle body frame performance prediction result is evaluated using the output sample values ​​from the model training sample set, thus obtaining the training effect of the first vehicle body frame performance prediction model in the current iteration. The local search update algorithm is configured with a maximum number of iterations. If the current iteration round has not reached the maximum number of iterations, it can return to execute the operation of using the local search update algorithm to determine the current first model optimization parameters of the first body frame performance prediction model in the current iteration round, so as to iteratively update and train the first body frame performance prediction model until it is determined that the current iteration round has reached the maximum number of iterations, and the training of the first body frame performance prediction model ends.

[0059] In an optional embodiment of the present invention, before iteratively updating the model parameters of the first vehicle body frame performance prediction model using the local search update algorithm based on the model training sample set, the method may further include: merging the model parameters of the first vehicle body frame performance prediction model into a one-dimensional vector to obtain first model optimization parameters; generating a design variable matrix based on the reference design variables of each training sample in the model training sample set, and generating a sample data performance simulation value matrix based on the performance simulation data of each training sample in the model training sample set; generating a first model prediction value matrix based on the first model optimization parameters and the design variable matrix; and generating a parameter optimization objective function for the first vehicle body frame performance prediction model based on the sample data performance simulation value matrix and the first model prediction value matrix.

[0060] Here, the first model optimization parameters are the vector set of model parameters for the first vehicle body frame performance prediction model. The design variable matrix can be a matrix generated from each reference design variable. The performance simulation data of each training sample in the model training sample set are the output sample values ​​of the model training sample set. The sample data performance simulation value matrix can be a matrix composed of the simulation values ​​of each vehicle body frame performance index of a sample. The first model prediction value matrix can be a matrix composed of the predicted values ​​of each vehicle body frame performance index by the first neural network surrogate model.

[0061] In a specific example, suppose the first neural network surrogate model includes one hidden layer. Since a hidden layer typically includes two thresholds and two weights, the model parameters of the first neural network surrogate model can be a first threshold W, a second threshold U, a first weight C, and a second weight D. Combining these model parameters of the first neural network surrogate model into a one-dimensional vector yields the first model optimization parameters X, i.e. .in: The design variable matrix p can be generated based on the reference design variables of each training sample in the model training sample set, i.e. ; Generate a sample data performance simulation value matrix q based on the performance simulation data of each training sample in the model training sample set, i.e. , Let be the simulated value of the i-th vehicle body frame performance index among all training samples. , This represents the simulated value of the nth vehicle body frame performance index of the Nth training sample. Further, based on the first model optimization parameters and the design variable matrix, the first model prediction matrix f is generated, i.e. .in, At this time, x can be . Let be the predicted value of the i-th vehicle body frame performance index of the first neural network surrogate model for all training samples. , This represents the predicted value of the nth vehicle body frame performance index for the Nth training sample by the first neural network surrogate model.

[0062] Finally, the objective function Y for optimizing the parameters of the first vehicle body frame performance prediction model can be generated based on the sample data performance simulation value matrix q and the first model prediction value matrix f. The optimization objective of the objective function Y can be to minimize the error between the predicted performance value and the sample performance value, i.e.: Y = min .

[0063] in, The number of reference design variables to be input; The total number of hidden factors in the first neural network surrogate model; denoted by , where is the number of performance indicators for the vehicle body frame. N represents the number of training samples in the model training sample set.

[0064] In an optional embodiment of the present invention, determining the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration using the local search update algorithm may include: determining the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration based on the following formula: ; ; in, Optimize the parameters of the first model in the (t+1)th iteration. The target first model optimization parameters are the first model optimization parameters for the first vehicle body frame performance prediction model in the current iteration round. These target first model optimization parameters can be the first model optimization parameters that achieve the highest prediction accuracy, i.e., the optimal first model optimization parameters. That is, This can be the model parameters of the first body frame performance prediction model with the highest prediction accuracy among all first body frame performance prediction models trained before (and including) the current iteration. λ is the influence factor. Let be the first random variable, which can take values ​​between [0,1], and t be the current iteration round. This represents the preset number of iterations for the first vehicle body frame performance prediction model. Let [the variable] be the second random variable, which can take values ​​between [0,1]. The first model optimization parameter for the current iteration is given. e is a natural constant, which can take a value of 2.71828.

[0065] By iteratively updating the training, the optimal parameters for the first model can be obtained, denoted as... The optimal value of the first threshold in the first vehicle body frame performance prediction model; The optimal value of the second threshold in the first vehicle body frame performance prediction model; This represents the optimal first weight value in the first vehicle body frame performance prediction model. This represents the optimal second weight value in the first vehicle body frame performance prediction model. The predicted value of the i-th vehicle body frame performance index corresponding to the optimal first model optimization parameters.

[0066] S350. The global search update algorithm is used to iteratively update the model parameters of the second body frame performance prediction model according to the model training sample set, and the second target body frame performance prediction model is selected according to the iterative update results.

[0067] In an optional embodiment of the present invention, the step of iteratively updating the model parameters of the second vehicle body frame performance prediction model using the global search update algorithm based on the model training sample set may include: determining the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration round using the global search update algorithm; determining the current vehicle body frame performance prediction result of the second vehicle body frame performance prediction model in the current iteration round based on the model training sample set; evaluating the prediction accuracy of the current vehicle body frame performance prediction result of the second vehicle body frame performance prediction model based on the parameter optimization objective function of the second vehicle body frame performance prediction model; and returning to execute the operation of determining the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration round using the global search update algorithm, until it is determined that the current iteration round has reached a preset number of iterations.

[0068] The current optimization parameters of the second model can be the applicable model parameters of the second vehicle body frame performance prediction model in the current iteration round. The current vehicle body frame performance prediction result can be the prediction result of the model for each vehicle body frame performance index in the current iteration round. The parameter optimization objective function can be used to evaluate the accuracy of the model's prediction results for the vehicle body frame performance index. The preset number of iterations can be the maximum number of iterations.

[0069] Specifically, a global search update algorithm can be used to determine the current optimization parameters of the second body frame performance prediction model in the current iteration. These current optimization parameters are then used to assign values ​​to the model parameters of the second body frame performance prediction model, resulting in the second body frame performance prediction model for the current iteration. Further, the input sample values ​​from the model training sample set are input into the second body frame performance prediction model. For example, the values ​​of 10 reference design variables are input into the corresponding neurons of the second body frame performance prediction model, resulting in the prediction results of multiple body frame performance indicators output by the output layer of the second body frame performance prediction model. These predictions serve as the current body frame performance prediction result for the second body frame performance prediction model in the current iteration. Finally, based on the parameter optimization objective function of the second body frame performance prediction model, the prediction accuracy of the current body frame performance prediction result is evaluated using the output sample values ​​from the model training sample set, thus obtaining the training effect of the second body frame performance prediction model in the current iteration. The global search update algorithm is configured with a maximum number of iterations. If the current iteration round has not reached the maximum number of iterations, it can return to execute the operation of using the local search update algorithm to determine the current second model optimization parameters of the second body frame performance prediction model in the current iteration round, so as to iteratively update and train the second body frame performance prediction model until it is determined that the current iteration round has reached the maximum number of iterations, and the training of the second body frame performance prediction model ends.

[0070] Optionally, the maximum number of iterations configured for the local search update algorithm and the global search update algorithm can be the same or different, and this embodiment of the invention does not limit this.

[0071] In an optional embodiment of the present invention, before iteratively updating the model parameters of the second vehicle body frame performance prediction model using a global search update algorithm based on the model training sample set, the method may further include: merging the model parameters of the second vehicle body frame performance prediction model into a one-dimensional vector to obtain second model optimization parameters; generating a design variable matrix based on the reference design variables of each training sample in the model training sample set, and generating a sample data performance simulation value matrix based on the performance simulation data of each training sample in the model training sample set; generating a second model prediction value matrix based on the second model optimization parameters and the design variable matrix; and generating a parameter optimization objective function for the second vehicle body frame performance prediction model based on the sample data performance simulation value matrix and the second model prediction value matrix.

[0072] The second model optimization parameters are the vector set of model parameters for the second vehicle body frame performance prediction model. The second model prediction matrix can be a matrix composed of the predicted values ​​of each vehicle body frame performance index by the second neural network surrogate model.

[0073] In a specific example, suppose the second neural network surrogate model includes two hidden layers. Since two hidden layers typically include three thresholds and three weights, the model parameters of the second neural network surrogate model can be the first threshold. Second threshold Third threshold First weight Second weight and third weight The model parameters of the second neural network surrogate model are combined into a one-dimensional vector to obtain the optimized parameters of the second model. ,Right now .in: The design variable matrix p can be generated based on the reference design variables of each training sample in the model training sample set, i.e. ; Generate a sample data performance simulation value matrix q based on the performance simulation data of each training sample in the model training sample set, i.e. , Let be the simulated value of the i-th vehicle body frame performance index among all training samples. , This represents the simulated value of the nth vehicle body frame performance index for the Nth training sample. Further, a second model prediction matrix is ​​generated based on the second model optimization parameters and the design variable matrix. ,Right now .in, At this time, x can be . Let be the predicted value of the i-th vehicle body frame performance index of the second neural network surrogate model for all training samples. , This represents the predicted value of the nth vehicle body frame performance index for the Nth training sample by the second neural network surrogate model.

[0074] Finally, the simulated performance matrix q and the predicted value matrix of the second model can be used as a basis for determining the performance of the sample data. Generate the parameter optimization objective function Y' for the second vehicle body frame performance prediction model. The optimization objective of the parameter optimization objective function Y' can be to minimize the error between the predicted performance value and the sample performance value, i.e.: Y' = min .

[0075] in, The number of reference design variables to be input; The number of performance indicators for the vehicle body frame; is the number of factors in each hidden layer within the second neural network surrogate model; N is the number of training samples in the model training sample set.

[0076] In an optional embodiment of the present invention, determining the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration using the global search update algorithm may include: determining the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration based on the following formula: ; ; ; in, For the (t+1)th iteration, optimize the parameters of the second model. As the first learning factor, The target second model optimization parameters are the second model optimization parameters for the second vehicle body frame performance prediction model in the current iteration round. These target second model optimization parameters can be the second model optimization parameters that achieve the highest prediction accuracy, i.e., the optimal second model optimization parameters. That is, The model parameters can be selected from all second body frame performance prediction models trained in previous iterations (including the current iteration) to determine the model with the highest prediction accuracy. Here, t represents the second model optimization parameter for the current iteration round. It is a third random variable that can take values ​​between [0,1]. This represents the preset number of iterations for the performance prediction model of the second vehicle body frame. As the second learning factor, Let [the variable] be the fourth random variable, and it can take values ​​between [0, 1]. and The two second model optimization parameters are randomly selected from the second model optimization parameters before the current iteration round. Let [the variable] be the fifth random variable, which can take values ​​between [0,1]. It is a natural constant, and its value can be 2.71828.

[0077] By iteratively updating the training, the optimal parameters for the second model can be obtained, denoted as... The optimal value of the first threshold in the second vehicle body frame performance prediction model; This is the optimal value of the second threshold in the second vehicle body frame performance prediction model; The optimal value of the third threshold in the second vehicle body frame performance prediction model; This is the optimal first weight value in the performance prediction model of the second vehicle body frame. This represents the optimal second weight value in the second body frame performance prediction model. This is the optimal value of the third weight in the second vehicle body frame performance prediction model. The predicted value of the i-th vehicle body frame performance index corresponding to the optimal second model optimization parameters.

[0078] S360. Calculate the first mean square error of the prediction results of the first target body frame performance prediction model for each body frame performance index.

[0079] The first mean square error can be the mean square error between the prediction results of the first target body frame performance prediction model for each body frame performance index and the performance simulation value of the sample data.

[0080] In an optional embodiment of the present invention, calculating the first mean square error of the prediction results of the first target body frame performance prediction model for each body frame performance index may include: calculating the first mean square error based on the following formula: in, This represents the first mean square error of the first vehicle frame performance prediction model for the predicted value of the i-th vehicle frame performance index. The predicted value of the i-th body frame performance index for all training samples corresponding to the optimal first model optimization parameters; Let be the simulated value of the i-th vehicle body frame performance index among all training samples. This represents the predicted value of the i-th vehicle body frame performance index for the N-th training sample corresponding to the optimal first model optimization parameters. This represents the simulated value of the i-th vehicle body frame performance index of the N-th training sample.

[0081] S370. Calculate the second mean square error of the prediction results of the second target body frame performance prediction model for each body frame performance index.

[0082] The second mean square error can be the mean square error between the prediction results of the second target body frame performance prediction model for each body frame performance index and the performance simulation value of the sample data.

[0083] In an optional embodiment of the present invention, calculating the second mean square error of the prediction results of the second target body frame performance prediction model for each body frame performance index may include: calculating the second mean square error based on the following formula: in, This represents the second mean square error of the predicted value of the i-th body frame performance index by the second body frame performance prediction model. The predicted value of the i-th vehicle body frame performance index corresponding to the optimal second model optimization parameters. The simulated value of the i-th vehicle body frame performance index for all training samples. This represents the predicted value of the i-th vehicle body frame performance index for the N-th training sample corresponding to the optimal second model optimization parameters. This represents the simulated performance value of the i-th vehicle body frame performance index corresponding to the N-th sample.

[0084] S380. Calculate the first fusion weight of the first target vehicle body frame performance prediction model and the second fusion weight of the second target vehicle body frame performance prediction model based on the first mean square error and the second mean square error.

[0085] The first fusion weight is the fusion weight of the first target vehicle body frame performance prediction model. The second fusion weight is the fusion weight of the second target vehicle body frame performance prediction model.

[0086] In an optional embodiment of the present invention, calculating the first fusion weight of the first target vehicle body frame performance prediction model based on the first mean square error and the second mean square error may include: calculating the first fusion weight of the first target vehicle body frame performance prediction model based on the following formula: Calculating the second fusion weights of the first target vehicle body frame performance prediction model based on the first mean square error and the second mean square error may include: calculating the second fusion weights of the first target vehicle body frame performance prediction model based on the following formula: in, This represents the first fusion weight of the first vehicle body frame performance prediction model for the i-th vehicle body frame performance index. This represents the second fusion weight of the second vehicle body frame performance prediction model for the i-th vehicle body frame performance index. This represents the first mean square error of the predicted value of the i-th body frame performance index by the first body frame performance prediction model. This represents the second mean square error of the predicted value of the i-th body frame performance index by the second body frame performance prediction model.

[0087] As can be seen from the above formula for calculating the fusion weight, the larger the mean squared error, the more unstable the prediction accuracy of the model, and the smaller the corresponding fusion weight; conversely, the smaller the mean squared error, the more stable the prediction accuracy of the model, and the larger the corresponding fusion weight.

[0088] S390. The first target vehicle body frame performance prediction model and the second target vehicle body frame performance prediction model are weighted and fused according to the first fusion weight and the second fusion weight to obtain the target vehicle body frame performance prediction model.

[0089] For example, the target vehicle body frame performance prediction model can be represented as: ,in: in, The predicted value of the i-th body frame performance index by the target body frame performance prediction model; This represents the first fusion weight of the first vehicle body frame performance prediction model for the i-th vehicle body frame performance index; This represents the second fusion weight of the second body frame performance prediction model for the i-th body frame performance index; This is the matrix of predicted values ​​from the first model; The predicted value of the i-th body frame performance index corresponding to the optimal first model optimization parameters; This is the matrix of predicted values ​​from the second model; The predicted value of the i-th body frame performance index corresponding to the optimal second model optimization parameters; These are the reference design variables for input.

[0090] In a specific example, Table 1 is a list of model training sample sets provided in Embodiment 2 of the present invention.

[0091] Table 1 List of Model Training Sample Sets Based on the five samples in the model training sample set provided in Table 1, the maximum number of iterations for optimizing the model parameters of the first vehicle body frame performance prediction model is set. After iterative training, the optimal parameters for the first model are as follows: , Based on the five samples in the model training sample set provided in Table 1, the maximum number of iterations for optimizing the model parameters of the second body frame performance prediction model is set. After iterative training, the optimal parameters for the second model are as follows: , , Based on the five samples from the model training sample set provided in Table 1, the first mean square error of each prediction result of the first target body frame performance prediction model for each body frame performance index is calculated as follows: , , , as well as The second mean square error of each prediction result of the second target body frame performance prediction model for each body frame performance index is as follows: , , , as well as .

[0092] Accordingly, the first fusion weights for each performance index of the first target vehicle body frame performance prediction model calculated based on the first mean square error and the second mean square error are as follows: The second fusion weights for each performance index of the second target vehicle frame performance prediction model, calculated based on the first mean square error and the second mean square error, are as follows: For example, the target body frame performance prediction model for each body frame performance index can be as follows: Wherein, F1 represents the target body frame performance prediction model for the first body frame performance index, F2 represents the target body frame performance prediction model for the second body frame performance index, F3 represents the target body frame performance prediction model for the third body frame performance index, F4 represents the target body frame performance prediction model for the fourth body frame performance index, and F5 represents the target body frame performance prediction model for the fifth body frame performance index.

[0093] Table 2 is a comparison table of first-order torsional modal frequency prediction results provided in Embodiment 2 of the present invention; Table 3 is a comparison table of first-order bending modal frequency prediction results provided in Embodiment 2 of the present invention; Table 4 is a comparison table of bending stiffness prediction results provided in Embodiment 2 of the present invention; Table 5 is a comparison table of torsional stiffness prediction results provided in Embodiment 2 of the present invention; and Table 6 is a comparison table of vehicle mass prediction results provided in Embodiment 2 of the present invention. Wherein, the first model prediction value can refer to the prediction value of the first vehicle body frame performance prediction model; the first model error can refer to the error of the first vehicle body frame performance prediction model; the second model prediction value can refer to the prediction value of the second vehicle body frame performance prediction model; the second model error can refer to the error of the second vehicle body frame performance prediction model; the target model prediction value can refer to the prediction value of the target vehicle body frame performance prediction model; and the target model error can refer to the error of the target vehicle body frame performance prediction model.

[0094] Table 2 Comparison of First-Order Torsional Mode Frequency Prediction Results ‌ Table 3 Comparison of First-Order Bending Mode Frequency Prediction Results ‌ Table 4 Comparison of Bending Stiffness Prediction Results ‌ Table 5 Comparison of Torsional Stiffness Prediction Results Table 6 Comparison of Vehicle Quality Prediction Results As shown in Table 2-6, the target vehicle frame performance prediction model has errors of less than 1% in predicting the first-order torsional mode frequency, less than 10% in predicting the first-order bending mode frequency, less than 15% in predicting bending stiffness, less than 5% in predicting torsional stiffness, and less than 1% in predicting the overall vehicle mass.

[0095] Specifically, as shown in Table 2, for the performance index of first-order torsional modal frequency, the target body frame performance prediction model outperforms the first and second target body frame performance prediction models in all five test samples, demonstrating higher prediction accuracy. For the performance index of first-order bending modal frequency, the target body frame performance prediction model's prediction accuracy for sample S1 is slightly lower than that of the first target body frame performance prediction model, and its prediction accuracy for samples S2 and S4 is slightly lower than that of the second target body frame performance prediction model. However, the cumulative prediction error of the target body frame performance prediction model for the five sample data is significantly smaller than that of the first and second target body frame performance prediction models, proving that the target body frame performance prediction model can integrate the advantages of body frame performance prediction models with different hidden layers, thereby improving prediction accuracy and stability. For the performance indices of bending stiffness and torsional stiffness, the target body frame performance prediction model's prediction accuracy for sample S4 is slightly lower than that of the first target body frame performance prediction model, but the cumulative prediction error for the five sample data is significantly smaller than that of the first and second target body frame performance prediction models. Regarding the performance indicators of overall vehicle quality, the prediction accuracy of the target body frame performance prediction model for samples S3 and S4 is slightly lower than that of the first target body frame performance prediction model, and the prediction accuracy for sample S5 is slightly lower than that of the second target body frame performance prediction model. The cumulative prediction error for the five sample data is less than that of the first and second target body frame performance prediction models, and the error is within 1%, indicating extremely high prediction accuracy and stability.

[0096] Before building the target vehicle frame performance prediction model, the above technical solution performs sensitivity analysis on the vehicle frame and selects the members that contribute significantly to the vehicle frame performance as reference design variables, effectively reducing the computational load during network model training and improving the training efficiency of the network model. Specifically, the first vehicle frame performance prediction model employs an update algorithm focused on local search during model parameter optimization, which significantly improves the model's linear prediction ability; the second vehicle frame performance prediction model employs an update algorithm focused on global search during model parameter optimization, which can reduce computational consumption while leveraging the model's nonlinear prediction ability. The target vehicle frame performance prediction model construction method proposed in this embodiment integrates the advantages of the first and second vehicle frame performance prediction models, and uses two different update algorithms for parameter optimization based on the characteristics of each model's parameter optimization process, effectively reducing computational load and complexity while ensuring model accuracy.

[0097] Example 3 Figure 12This is a flowchart of a performance prediction method provided in Embodiment 3 of the present invention. This embodiment is applicable to situations where a multi-dimensional vehicle body frame performance index is predicted based on a target vehicle body frame performance prediction model that includes two different models obtained through training. This method can be executed by a performance prediction device, which can be implemented in software and / or hardware, and is generally integrated into an electronic device. This electronic device can be a terminal device or a server device, as long as it can execute the performance prediction method. The present invention does not limit the specific type of electronic device. Correspondingly, as... Figure 12 As shown, the method includes the following operations: S1210. Obtain the target design variables for each member of the target vehicle body frame.

[0098] The target vehicle body frame can be the vehicle body frame of a type for which multidimensional performance indicators of the vehicle body frame need to be predicted. The target design variables can be some of the design variables selected from the target vehicle body frame.

[0099] Optionally, the target design variable can be of the same type as the reference design variable. The values ​​of the target design variable can be configured according to actual needs.

[0100] S1220. The multi-dimensional body frame performance index of the target body frame is predicted based on the target design variables using the target body frame performance prediction model, and the prediction result of the multi-dimensional body frame performance index is obtained.

[0101] The target vehicle body frame performance prediction model includes a first vehicle body frame performance prediction model and a second vehicle body frame performance prediction model; the first vehicle body frame performance prediction model and the second vehicle body frame performance prediction model are trained using the performance prediction model training method described in any embodiment of the present invention. The multi-dimensional vehicle body frame performance index prediction result can be the prediction result corresponding to each vehicle body frame performance index.

[0102] Correspondingly, each target design variable can be used as input and fed into the first and second body frame performance prediction models of the target body frame performance prediction model corresponding to each body frame performance index. The predicted values ​​of the corresponding body frame performance indexes are then predicted by the first and second body frame performance prediction models. Based on the fusion weights corresponding to each body frame performance index, the predicted values ​​of the corresponding body frame performance indexes predicted by the first and second body frame performance prediction models are weighted and fused to obtain the prediction results for each body frame performance index.

[0103] This invention generates a model training sample set based on the reference design variables of each member of the vehicle body frame, and determines the model topology of a first and a second vehicle body frame performance prediction model used to predict multi-dimensional vehicle body frame performance indicators. Further, a local search update algorithm is used to iteratively update the model parameters of the first vehicle body frame performance prediction model based on the model training sample set, and a first target vehicle body frame performance prediction model is selected based on the iterative update results. Simultaneously, a global search update algorithm is used to iteratively update the model parameters of the second vehicle body frame performance prediction model based on the model training sample set, and a second target vehicle body frame performance prediction model is selected based on the iterative update results. Finally, a target vehicle body frame performance prediction model is generated based on the first and second target vehicle body frame performance prediction models. After the target vehicle body frame performance prediction model is constructed, the target design variables of each member of the target vehicle body frame can be obtained, and the multi-dimensional vehicle body frame performance indicators of the target vehicle body frame can be predicted based on the target design variables using the target vehicle body frame performance prediction model, thus obtaining the multi-dimensional vehicle body frame performance indicator prediction results. The above technical solution effectively reduces the computational load during model training by using the reference design variables selected from each member of the vehicle body frame to generate the model training sample set. Meanwhile, the first vehicle body frame performance prediction model employs an update algorithm focused on local search during model parameter optimization training, which significantly improves the model's predictive ability. The second vehicle body frame performance prediction model, on the other hand, uses an update algorithm focused on global search during model parameter optimization training, which reduces computational cost while leveraging the model's nonlinear predictive capabilities. By using two different model parameter update algorithms for model parameter optimization during training, the overall computational load of the vehicle body frame performance prediction model can be effectively reduced while maintaining its accuracy, thereby improving the training efficiency and performance prediction accuracy of the vehicle body frame performance prediction model.

[0104] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data comply with the relevant laws, regulations and standards of the relevant regions, and do not violate public order and good morals.

[0105] It should be noted that any arrangement or combination of the technical features in the above embodiments also falls within the protection scope of this invention.

[0106] Example 4 Figure 13 This is a schematic diagram of a performance prediction model training device provided in Embodiment 4 of the present invention, as shown below. Figure 13As shown, the device includes: a model training sample set generation module 1310, a model topology determination module 1320, a first target vehicle body frame performance prediction model training module 1330, a second target vehicle body frame performance prediction model training module 1340, and a target vehicle body frame performance prediction model generation module 1350, wherein: The model training sample set generation module 1310 is used to generate a model training sample set based on the reference design variables of each member of the vehicle body frame. The model topology determination module 1320 is used to determine the model topology of the first body frame performance prediction model and the second body frame performance prediction model; wherein, the first body frame performance prediction model and the second body frame performance prediction model are used to predict multi-dimensional body frame performance indicators. The first target vehicle body frame performance prediction model training module 1330 is used to iteratively update the model parameters of the first vehicle body frame performance prediction model according to the model training sample set using a local search update algorithm, and to select the first target vehicle body frame performance prediction model according to the iterative update results. The second target body frame performance prediction model training module 1340 is used to iteratively update the model parameters of the second body frame performance prediction model according to the model training sample set using a global search update algorithm, and to select the second target body frame performance prediction model according to the iterative update results. The target body frame performance prediction model generation module 1350 is used to generate a target body frame performance prediction model based on the first target body frame performance prediction model and the second target body frame performance prediction model.

[0107] Optionally, the model training sample set generation module 1310 is further configured to: determine the reference design variables of the vehicle frame based on the multi-dimensional sensitivity analysis results of each member of the vehicle frame; and generate the model training sample set based on the reference design variables and the multi-dimensional vehicle frame performance indicators.

[0108] Optionally, the model training sample set generation module 1310 is further configured to: calculate the single-dimensional sensitivity analysis results at each target member position of the vehicle body frame, and determine the number of reference design variables matched by each single-dimensional sensitivity analysis result; based on the single-dimensional sensitivity analysis results at each target member position of the vehicle body frame, and according to the number of reference design variables matched by each single-dimensional sensitivity analysis result, sequentially select each reference design variable from each target member position of the vehicle body frame.

[0109] Optionally, the model topology determination module 1320 is further configured to: configure the input layers of the first body frame performance prediction model and the second body frame performance prediction model according to the number of reference design variables; configure the output layers of the first body frame performance prediction model and the second body frame performance prediction model according to the number of body frame performance indicators; configure the number of hidden layers and hidden factors of the first body frame performance prediction model and the second body frame performance prediction model; wherein the number of hidden layers of the first body frame performance prediction model is a first number, the number of hidden layers of the second body frame performance prediction model is a second number, and the first number is less than the second number.

[0110] Optionally, the first target vehicle body frame performance prediction model training module 1330 is further configured to: determine the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration round using the local search update algorithm; determine the current vehicle body frame performance prediction result of the first vehicle body frame performance prediction model in the current iteration round based on the model training sample set; evaluate the prediction accuracy of the current vehicle body frame performance prediction result of the first vehicle body frame performance prediction model based on the parameter optimization objective function of the first vehicle body frame performance prediction model; and return to execute the operation of determining the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration round using the local search update algorithm, until it is determined that the current iteration round has reached a preset number of iterations.

[0111] Optionally, the first target vehicle body frame performance prediction model training module 1330 is further configured to: merge the model parameters of the first vehicle body frame performance prediction model into a one-dimensional vector to obtain the first model optimization parameters; generate a design variable matrix based on the reference design variables of each training sample in the model training sample set, and generate a sample data performance simulation value matrix based on the performance simulation data of each training sample in the model training sample set; generate a first model prediction value matrix based on the first model optimization parameters and the design variable matrix; and generate a parameter optimization objective function for the first vehicle body frame performance prediction model based on the sample data performance simulation value matrix and the first model prediction value matrix.

[0112] Optionally, the first target vehicle body frame performance prediction model training module 1330 is further used to: determine the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration based on the following formula: ; ; in, Optimize the parameters of the first model in the (t+1)th iteration. The first model optimization parameters for the first vehicle body frame performance prediction model in the current iteration are defined as follows: λ is the influencing factor. Let be the first random variable, and t be the current iteration round. This represents the preset number of iterations for the first vehicle body frame performance prediction model. Let be the second random variable. The first model optimization parameters for the current iteration round.

[0113] Optionally, the second target vehicle body frame performance prediction model training module 1340 is further configured to: determine the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration round using the global search update algorithm; determine the current vehicle body frame performance prediction result of the second vehicle body frame performance prediction model in the current iteration round based on the model training sample set; evaluate the prediction accuracy of the current vehicle body frame performance prediction result of the second vehicle body frame performance prediction model based on the parameter optimization objective function of the second vehicle body frame performance prediction model; and return to execute the operation of determining the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration round using the global search update algorithm, until the current iteration round reaches a preset number of iterations.

[0114] Optionally, the second target vehicle body frame performance prediction model training module 1340 is further configured to: merge the model parameters of the second vehicle body frame performance prediction model into a one-dimensional vector to obtain the second model optimization parameters; generate a design variable matrix based on the reference design variables of each training sample in the model training sample set, and generate a sample data performance simulation value matrix based on the performance simulation data of each training sample in the model training sample set; generate a second model prediction value matrix based on the second model optimization parameters and the design variable matrix; and generate a parameter optimization objective function for the second vehicle body frame performance prediction model based on the sample data performance simulation value matrix and the second model prediction value matrix.

[0115] Optionally, the second target vehicle body frame performance prediction model training module 1340 is further used to: determine the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration based on the following formula: ; ; ; in, For the (t+1)th iteration, optimize the parameters of the second model. As the first learning factor, The target second model optimization parameters for the second vehicle body frame performance prediction model in the current iteration round are: Here, t represents the second model optimization parameter for the current iteration round. As the third random variable, This represents the preset number of iterations for the performance prediction model of the second vehicle body frame. As the second learning factor, As the fourth random variable, and The two second model optimization parameters are randomly selected from the second model optimization parameters before the current iteration round. It is the fifth random variable.

[0116] Optionally, the target body frame performance prediction model generation module 1350 is further configured to: calculate the first mean square error of the prediction results of the first target body frame performance prediction model for each body frame performance index; calculate the second mean square error of the prediction results of the second target body frame performance prediction model for each body frame performance index; calculate the first fusion weight of the first target body frame performance prediction model and the second fusion weight of the second target body frame performance prediction model based on the first mean square error and the second mean square error; and perform weighted fusion of the first target body frame performance prediction model and the second target body frame performance prediction model based on the first fusion weight and the second fusion weight to obtain the target body frame performance prediction model.

[0117] Optionally, the target vehicle body frame performance prediction model generation module 1350 is further configured to: calculate the first fusion weight of the first target vehicle body frame performance prediction model based on the following formula: The second fusion weight of the first target vehicle body frame performance prediction model is calculated based on the following formula: in, This represents the first fusion weight of the first vehicle body frame performance prediction model for the i-th vehicle body frame performance index. This represents the second fusion weight of the second vehicle body frame performance prediction model for the i-th vehicle body frame performance index. This represents the first mean square error of the predicted value of the i-th body frame performance index by the first body frame performance prediction model. This represents the second mean square error of the predicted value of the i-th body frame performance index by the second body frame performance prediction model.

[0118] The performance prediction model training apparatus described above can execute the performance prediction model training method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in this embodiment can be found in the performance prediction model training method provided in any embodiment of the present invention.

[0119] Since the performance prediction model training device described above is an apparatus capable of executing the performance prediction model training method in the embodiments of the present invention, those skilled in the art can understand the specific implementation and various variations of the performance prediction model training device in this embodiment based on the performance prediction model training method described in the embodiments of the present invention. Therefore, how the performance prediction model training device implements the performance prediction model training method in the embodiments of the present invention will not be described in detail here. Any apparatus used by those skilled in the art to implement the performance prediction model training method in the embodiments of the present invention falls within the scope of protection of this application.

[0120] Example 5 Figure 14 This is a schematic diagram of a performance prediction device provided in Embodiment 5 of the present invention, as shown below. Figure 14 As shown, the device includes: a target design variable acquisition module 1410 and a multi-dimensional vehicle body frame performance index prediction module 1420, wherein: The target design variable acquisition module 1410 is used to acquire the target design variables of each member of the target vehicle body frame. The multi-dimensional body frame performance index prediction module 1420 is used to predict the multi-dimensional body frame performance index of the target body frame according to the target design variables through the target body frame performance prediction model, and obtain the multi-dimensional body frame performance index prediction result. The target body frame performance prediction model includes a first body frame performance prediction model and a second body frame performance prediction model; the first body frame performance prediction model and the second body frame performance prediction model are trained by the performance prediction model training method described in any embodiment of the present invention.

[0121] The performance prediction device described above can execute the performance prediction method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the performance prediction method provided in any embodiment of the present invention.

[0122] Since the performance prediction device described above is an apparatus capable of executing the performance prediction method in the embodiments of the present invention, those skilled in the art can understand the specific implementation methods and various variations of the performance prediction device in this embodiment based on the performance prediction method described in the embodiments of the present invention. Therefore, how the performance prediction device implements the performance prediction method in the embodiments of the present invention will not be described in detail here. Any apparatus used by those skilled in the art to implement the performance prediction method in the embodiments of the present invention falls within the scope of protection of this application.

[0123] Example 6 Figure 15 A schematic diagram of an electronic device 10, which 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 processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0124] like Figure 15 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 and a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded into the RAM 13 from storage unit 18. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0125] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0126] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as performance prediction model training methods or performance prediction methods.

[0127] Optionally, the performance prediction model training method may include: generating a model training sample set based on the reference design variables of each member of the vehicle body frame; determining the model topology of the first vehicle body frame performance prediction model and the second vehicle body frame performance prediction model; wherein the first vehicle body frame performance prediction model and the second vehicle body frame performance prediction model are used to predict multi-dimensional vehicle body frame performance indicators; using a local search update algorithm to iteratively update the model parameters of the first vehicle body frame performance prediction model based on the model training sample set, and selecting a first target vehicle body frame performance prediction model based on the iterative update results; using a global search update algorithm to iteratively update the model parameters of the second vehicle body frame performance prediction model based on the model training sample set, and selecting a second target vehicle body frame performance prediction model based on the iterative update results; and generating a target vehicle body frame performance prediction model based on the first target vehicle body frame performance prediction model and the second target vehicle body frame performance prediction model.

[0128] Optionally, performance prediction may include: obtaining target design variables for each member of the target body frame; predicting multi-dimensional body frame performance indicators of the target body frame based on the target design variables using a target body frame performance prediction model, and obtaining multi-dimensional body frame performance indicator prediction results; wherein, the target body frame performance prediction model includes a first body frame performance prediction model and a second body frame performance prediction model; the first body frame performance prediction model and the second body frame performance prediction model are trained using the performance prediction model training method described in any embodiment of the present invention.

[0129] In some embodiments, the performance prediction model training method or performance prediction method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the performance prediction model training method or performance prediction method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to execute the performance prediction model training method or performance prediction method by any other suitable means (e.g., by means of firmware).

[0130] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0131] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0132] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0133] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, 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 sound input, voice input, or tactile input).

[0134] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0135] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0136] This invention also discloses a computer program product, which includes a computer program that, when executed by a processor, implements the performance prediction model training method or performance prediction method provided in any embodiment of this invention. This program product shares the same inventive concept as the performance prediction model training method or performance prediction method disclosed in the various embodiments of this invention, and therefore will not be described in detail here.

[0137] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the embodiments of the present invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in the embodiments of the present invention can be achieved, and this is not limited herein.

[0138] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for training a performance prediction model, characterized in that, include: A model training sample set is generated based on the reference design variables of each member of the vehicle body frame; The model topology of the first body frame performance prediction model and the second body frame performance prediction model is determined; wherein the first body frame performance prediction model and the second body frame performance prediction model are used to predict multidimensional body frame performance indicators. The local search update algorithm is used to iteratively update the model parameters of the first vehicle body frame performance prediction model based on the model training sample set, and the first target vehicle body frame performance prediction model is selected based on the iterative update results. A global search update algorithm is used to iteratively update the model parameters of the second body frame performance prediction model based on the model training sample set, and the second target body frame performance prediction model is selected based on the iterative update results. A target body frame performance prediction model is generated based on the first target body frame performance prediction model and the second target body frame performance prediction model.

2. The method according to claim 1, characterized in that, The process of generating a model training sample set based on the reference design variables of each member of the vehicle body frame includes: The reference design variables of the vehicle body frame are determined based on the multi-dimensional sensitivity analysis results of each member of the vehicle body frame. The model training sample set is generated based on the reference design variables and multi-dimensional vehicle body frame performance indicators.

3. The method according to claim 2, characterized in that, The step of determining the reference design variables of the vehicle body frame based on the multi-dimensional sensitivity analysis results of each member of the vehicle body frame includes: Calculate the single-dimensional sensitivity analysis results at the positions of each target member of the vehicle body frame, and determine the number of reference design variables to be screened for each single-dimensional sensitivity analysis result; Based on the single-dimensional sensitivity analysis results at each target member position of the vehicle body frame, and according to the number of reference design variables matched by each single-dimensional sensitivity analysis result, each reference design variable is sequentially selected from each target member position of the vehicle body frame.

4. The method according to claim 1, characterized in that, The process of determining the model topology of the first body frame performance prediction model and the second body frame performance prediction model includes: Configure the input layers of the first body frame performance prediction model and the second body frame performance prediction model according to the number of reference design variables; Configure the output layers of the first body frame performance prediction model and the second body frame performance prediction model according to the number of body frame performance indicators; Configure the number of hidden layers and hidden factors of the first vehicle body frame performance prediction model and the second vehicle body frame performance prediction model; wherein, the number of hidden layers of the first vehicle body frame performance prediction model is a first number, the number of hidden layers of the second vehicle body frame performance prediction model is a second number, and the first number is less than the second number.

5. The method according to claim 1, characterized in that, The step of iteratively updating the model parameters of the first vehicle body frame performance prediction model using the local search update algorithm based on the model training sample set includes: The local search update algorithm is used to determine the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration round; Based on the model training sample set, determine the current vehicle frame performance prediction result of the first vehicle frame performance prediction model in the current iteration round; The prediction accuracy of the current body frame performance prediction result of the first body frame performance prediction model is evaluated based on the parameter optimization objective function of the first body frame performance prediction model. Return to the operation of using the local search update algorithm to determine the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration round, until it is determined that the current iteration round has reached the preset number of iterations.

6. The method according to claim 5, characterized in that, Before iteratively updating the model parameters of the first vehicle body frame performance prediction model using the local search update algorithm based on the model training sample set, the method further includes: The model parameters of the first vehicle body frame performance prediction model are merged into a one-dimensional vector to obtain the first model optimization parameters; A design variable matrix is ​​generated based on the reference design variables of each training sample in the model training sample set, and a sample data performance simulation value matrix is ​​generated based on the performance simulation data of each training sample in the model training sample set. Generate a first model prediction value matrix based on the first model optimization parameters and the design variable matrix; The objective function for parameter optimization of the first vehicle body frame performance prediction model is generated based on the sample data performance simulation value matrix and the first model prediction value matrix.

7. The method according to claim 5 or 6, characterized in that, The step of determining the current first model optimization parameters of the first vehicle body frame performance prediction model in the current iteration using the local search update algorithm includes: The optimization parameters of the first model in the current iteration of the first vehicle body frame performance prediction model are determined based on the following formula: ； ； in, Optimize the parameters of the first model in the (t+1)th iteration. The first model optimization parameters for the first vehicle body frame performance prediction model in the current iteration are defined as follows: λ is the influencing factor. Let be the first random variable, and t be the current iteration round. This represents the preset number of iterations for the first vehicle body frame performance prediction model. Let be the second random variable. The first model optimization parameters for the current iteration round.

8. The method according to claim 1, characterized in that, The step of iteratively updating the model parameters of the second vehicle body frame performance prediction model using a global search update algorithm based on the model training sample set includes: The global search and update algorithm is used to determine the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration round; The performance prediction result of the second vehicle body frame in the current iteration round is determined based on the model training sample set. The prediction accuracy of the current body frame performance prediction result of the second body frame performance prediction model is evaluated based on the parameter optimization objective function of the second body frame performance prediction model. Return to the operation of using the global search update algorithm to determine the current second model optimization parameters of the second body frame performance prediction model in the current iteration round, until it is determined that the current iteration round has reached the preset number of iterations.

9. The method according to claim 8, characterized in that, Before iteratively updating the model parameters of the second vehicle body frame performance prediction model using the global search update algorithm based on the model training sample set, the method further includes: The model parameters of the second vehicle body frame performance prediction model are merged into a one-dimensional vector to obtain the second model optimization parameters; A design variable matrix is ​​generated based on the reference design variables of each training sample in the model training sample set, and a sample data performance simulation value matrix is ​​generated based on the performance simulation data of each training sample in the model training sample set. Generate a second model prediction value matrix based on the second model optimization parameters and the design variable matrix; The objective function for parameter optimization of the second vehicle body frame performance prediction model is generated based on the sample data performance simulation value matrix and the second model prediction value matrix.

10. The method according to claim 8, characterized in that, The step of using the global search update algorithm to determine the current second model optimization parameters of the second vehicle body frame performance prediction model in the current iteration includes: The optimization parameters of the second body frame performance prediction model in the current iteration are determined based on the following formula: ； ； ； in, For the (t+1)th iteration, optimize the parameters of the second model. As the first learning factor, The target second model optimization parameters for the second vehicle body frame performance prediction model in the current iteration round are: Here, t represents the second model optimization parameter for the current iteration round. As the third random variable, This represents the preset number of iterations for the performance prediction model of the second vehicle body frame. As the second learning factor, As the fourth random variable, and The two second model optimization parameters are randomly selected from the second model optimization parameters before the current iteration round. It is the fifth random variable.

11. The method according to claim 1, characterized in that, The step of generating a target body frame performance prediction model based on the first target body frame performance prediction model and the second target body frame performance prediction model includes: Calculate the first mean square error of the prediction results of the first target body frame performance prediction model for each body frame performance index; Calculate the second mean square error of the prediction results of the second target body frame performance prediction model for each body frame performance index; The first fusion weight of the first target vehicle body frame performance prediction model and the second fusion weight of the second target vehicle body frame performance prediction model are calculated based on the first mean square error and the second mean square error. The first target vehicle body frame performance prediction model and the second target vehicle body frame performance prediction model are weighted and fused according to the first fusion weight and the second fusion weight to obtain the target vehicle body frame performance prediction model.

12. The method according to claim 11, characterized in that, The first fusion weights of the first target vehicle body frame performance prediction model are calculated based on the first mean square error and the second mean square error, including: The first fusion weight of the first target vehicle body frame performance prediction model is calculated based on the following formula: The second fusion weights of the first target vehicle body frame performance prediction model are calculated based on the first mean square error and the second mean square error, including: The second fusion weight of the first target vehicle body frame performance prediction model is calculated based on the following formula: in, This represents the first fusion weight of the first vehicle body frame performance prediction model for the i-th vehicle body frame performance index. This represents the second fusion weight of the second vehicle body frame performance prediction model for the i-th vehicle body frame performance index. This represents the first mean square error of the predicted value of the i-th body frame performance index by the first body frame performance prediction model. This represents the second mean square error of the predicted value of the i-th body frame performance index by the second body frame performance prediction model.

13. A performance prediction method, characterized in that, include: Obtain the target design variables for each member of the target vehicle body frame; The target body frame performance prediction model is used to predict the multidimensional body frame performance index of the target body frame based on the target design variables, and the prediction result of the multidimensional body frame performance index is obtained. The target vehicle body frame performance prediction model includes a first vehicle body frame performance prediction model and a second vehicle body frame performance prediction model; the first vehicle body frame performance prediction model and the second vehicle body frame performance prediction model are trained by the performance prediction model training method according to any one of claims 1-12.

14. A performance prediction model training device, characterized in that, include: The model training sample set generation module is used to generate a model training sample set based on the reference design variables of each member of the vehicle body frame. The model topology determination module is used to determine the model topology of the first body frame performance prediction model and the second body frame performance prediction model; wherein, the first body frame performance prediction model and the second body frame performance prediction model are used to predict multi-dimensional body frame performance indicators. The first target body frame performance prediction model training module is used to iteratively update the model parameters of the first body frame performance prediction model according to the model training sample set using a local search update algorithm, and to select the first target body frame performance prediction model according to the iterative update results. The training module for the second target body frame performance prediction model is used to iteratively update the model parameters of the second body frame performance prediction model based on the model training sample set using a global search update algorithm, and to select the second target body frame performance prediction model based on the iterative update results. The target body frame performance prediction model generation module is used to generate a target body frame performance prediction model based on the first target body frame performance prediction model and the second target body frame performance prediction model.

15. A performance prediction device, characterized in that, include: The target design variable acquisition module is used to acquire the target design variables of each member of the target vehicle body frame; The multi-dimensional body frame performance index prediction module is used to predict the multi-dimensional body frame performance index of the target body frame according to the target design variables through the target body frame performance prediction model, and obtain the multi-dimensional body frame performance index prediction result. The target vehicle body frame performance prediction model includes a first vehicle body frame performance prediction model and a second vehicle body frame performance prediction model; the first vehicle body frame performance prediction model and the second vehicle body frame performance prediction model are trained by the performance prediction model training method according to any one of claims 1-12.

16. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to said at least one processor; wherein, The memory stores a computer program that is executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the performance prediction model training method according to any one of claims 1-12, or to perform the performance prediction method according to claim 13.

17. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the performance prediction model training method of any one of claims 1-12, or to perform the performance prediction method of claim 13.

18. A computer program product, characterized in that, Includes a computer program / instruction, wherein when the computer program / instruction is executed by a processor, it implements the performance prediction model training method of any one of claims 1-12, or performs the performance prediction method of claim 13.

Images