Automobile sound insulation performance multidisciplinary design optimization method and device and electronic equipment

Abstract

The application relates to a kind of automobile sound insulation performance multidisciplinary design optimization method, device and electronic equipment, method includes: the multiple influence indexes needing optimization design of automobile sound insulation performance are carried out DOE sampling, based on the accuracy determination target proxy model group formed by the target proxy model corresponding to all influence indexes;Determine the optimal point based on target proxy model group, if the difference between the actual index response corresponding to the optimal point and the optimization constraint condition is greater than the preset threshold, the target proxy model group is updated using the encrypted DOE training sample data set corresponding to the optimal point, the encrypted DOE training sample data set is constructed based on the optimal point and the value range of multiple design variables in related parameters, until the difference between the actual index response and the optimization constraint condition is less than or equal to the preset threshold, output the final updated target proxy model group, for automobile sound insulation performance optimization.The application embodiment can realize the joint optimization of automobile sound insulation performance by improving the accuracy of proxy model near the optimal point and selecting the target proxy model with the highest accuracy.

Description

Technical Field

[0001] This application relates to the field of multidisciplinary design optimization, and in particular to a method, apparatus and electronic device for multidisciplinary design optimization of automotive sound insulation performance. Background Technology

[0002] Automotive design is a complex systems engineering project involving multiple disciplines, including collaborative design in areas such as structure, crash, NVH, thermal, battery, and optics. Data transfer between different disciplines is difficult, and experience-based trial-and-error optimization processes lead to long development cycles and high costs. Multidisciplinary Design Optimization (MDO), by exploring and utilizing the synergistic mechanisms of interaction within the system, employs multi-objective strategies and computer-aided techniques to design complex systems and subsystems, effectively shortening the design cycle and achieving optimal overall system performance.

[0003] Currently, multidisciplinary design optimization processes generally include three steps: Design of Experiment (DOE), surrogate model establishment, and optimization design. Applying multidisciplinary design optimization in the automotive field can significantly shorten the R&D cycle and reduce R&D costs. However, in traditional automotive multidisciplinary design optimization, the final results often fail to meet design requirements in actual production. Summary of the Invention

[0004] In order to solve the above-mentioned technical problems, or at least partially solve the above-mentioned technical problems, this application provides a multidisciplinary design optimization method, device and electronic device for automotive sound insulation performance.

[0005] Firstly, this application provides a multidisciplinary design optimization method for automotive sound insulation performance, including:

[0006] Obtain relevant parameters for multidisciplinary design optimization of automotive sound insulation performance, determine multiple influencing indicators that need to be optimized for automotive sound insulation performance, perform DOE sampling of multiple influencing indicators, and determine the initial DOE training sample dataset and DOE validation sample dataset based on the DOE sampling results.

[0007] Based on the accuracy, the target proxy model corresponding to each impact indicator is determined from the multiple candidate proxy model groups corresponding to each impact indicator, and the target proxy model group consisting of the target proxy models corresponding to all impact indicators is obtained. The candidate proxy model group is pre-trained based on the initial DOE training sample dataset and the DOE validation sample dataset.

[0008] Based on the target agent model group, global optimization is performed to obtain the optimal point, and the actual indicator response is determined based on the optimal point;

[0009] If the difference between the actual index response and the optimization constraints is greater than a preset threshold, the target proxy model group is updated using the encrypted DOE training sample dataset corresponding to the optimal point. The encrypted DOE training sample dataset is constructed based on the optimal point and the value range of multiple design variables in the relevant parameters. The updated target proxy model is then used for global optimization until the difference between the actual index response and the optimization constraints is less than or equal to the preset threshold. Finally, the updated target proxy model group is output for use in automotive sound insulation performance optimization.

[0010] Optionally, updating the target agent model group using the encrypted DOE training sample dataset corresponding to the optimal point includes:

[0011] Obtain an encrypted DOE training sample dataset with increased local sampling density within a preset range of the optimal point;

[0012] The target agent model in the target agent model group is trained using the encrypted DOE training sample dataset to obtain the updated target agent model group.

[0013] Optionally, an encrypted DOE training sample dataset with increased local sampling density within a preset range of the optimal point is obtained, including:

[0014] A local sampling space is constructed based on the optimal point and the value range of multiple design variables in the relevant parameters;

[0015] DOE sampling is performed on the local sampling space to obtain a local DOE training sample point set;

[0016] The local DOE training sample point set is analyzed and calculated to obtain the local DOE training sample dataset;

[0017] The local DOE training sample dataset is merged into the initial DOE training sample dataset to obtain the encrypted DOE training sample dataset.

[0018] Optionally, a local sampling space is constructed based on the optimal point and the value ranges of multiple design variables among the relevant parameters, including:

[0019] The value range of multiple design variables is determined based on the aforementioned relevant parameters;

[0020] The size of the initial sampling space is calculated based on the value range of each design variable;

[0021] The radius of the local sampling space is determined based on the size of the initial sampling space;

[0022] An initial local sampling space is generated based on the optimal point and the radius of the local sampling space;

[0023] The local sampling space is determined based on the initial local sampling space and the initial design space.

[0024] Optionally, determining the local sampling space based on the initial local sampling space and the initial design space includes:

[0025] The initial local sampling space is compared with the initial design space to determine whether the initial local sampling space exceeds the initial design space.

[0026] If the initial local sampling space exceeds the initial design space, the intersection of the initial local sampling space and the initial sampling space is determined as the local sampling space for this iteration;

[0027] If the initial local sampling space does not exceed the initial design space, the initial local sampling space is determined as the local sampling space for this iteration.

[0028] Optionally, DOE sampling is performed on multiple influencing metrics, and the initial DOE training sample dataset and DOE validation sample dataset are determined based on the DOE sampling results, including:

[0029] DOE sampling was performed on multiple influencing indicators of automotive sound insulation performance to obtain DOE training sample point set and DOE validation sample point set;

[0030] The sample points in the DOE training sample point set and the DOE validation sample point set are analyzed and calculated to obtain the initial DOE training sample dataset and the DOE validation sample dataset.

[0031] Optionally, the method further includes:

[0032] Multiple surrogate models were trained using the initial DOE training sample dataset and the DOE validation sample dataset, respectively.

[0033] The trained agent model combination is determined as the candidate agent model group.

[0034] Optionally, based on accuracy, from the multiple candidate proxy model groups corresponding to each impact indicator, a target proxy model corresponding to each impact indicator is determined, resulting in a target proxy model group consisting of target proxy models corresponding to all impact indicators, including:

[0035] For each impact metric, determine the accuracy of the impact metric under each proxy model in the candidate proxy model group;

[0036] The proxy model with the highest accuracy corresponding to each of the aforementioned influencing indicators is determined as the target proxy model corresponding to the aforementioned influencing indicator;

[0037] The target proxy model group is obtained by combining the target proxy models corresponding to all the influencing indicators.

[0038] Optionally, based on the target agent model group, global optimization is performed to obtain the optimal point, and the actual indicator response is determined based on the optimal point, including:

[0039] Based on the target proxy model group, the multiple influencing indicators of automotive sound insulation performance are globally optimized to obtain the optimal point.

[0040] Based on the optimal point, determine the actual index response of the optimal point at the real model.

[0041] Secondly, this application provides a multidisciplinary design optimization device for automotive sound insulation performance, comprising:

[0042] The acquisition module is used to acquire relevant parameters for multidisciplinary design optimization of automotive sound insulation performance, determine multiple influencing indicators that need to be optimized for automotive sound insulation performance, perform DOE sampling of multiple influencing indicators, and determine the initial DOE training sample dataset and DOE validation sample dataset based on the DOE sampling results.

[0043] The first determining module is used to determine the target proxy model corresponding to each influence indicator from multiple candidate proxy model groups corresponding to each influence indicator based on the accuracy, so as to obtain a target proxy model group composed of target proxy models corresponding to all influence indicators. The candidate proxy model group is pre-trained based on the initial DOE training sample dataset and DOE validation sample dataset.

[0044] The optimization module is used to perform global optimization based on the target agent model group to obtain the optimal point, and to determine the actual indicator response based on the optimal point;

[0045] The update module is used to update the target proxy model group using the encrypted DOE training sample dataset corresponding to the optimal point if the difference between the actual index response and the optimization constraints is greater than a preset threshold. The encrypted DOE training sample dataset is constructed based on the optimal point and the value range of multiple design variables in the relevant parameters. The updated target proxy model is then used for global optimization until the difference between the actual index response and the optimization constraints is less than or equal to the preset threshold. Finally, the updated target proxy model group is output for use in automotive sound insulation performance optimization.

[0046] Optionally, the update module includes:

[0047] The acquisition submodule is used to acquire an encrypted DOE training sample dataset with increased local sampling density within a preset range of the optimal point;

[0048] The training submodule is used to train each target agent model in the target agent model group using the encrypted DOE training sample dataset to obtain the updated target agent model group.

[0049] Optionally, the acquisition submodule includes:

[0050] A construction unit is used to construct a local sampling space based on the optimal point and the value range of multiple design variables in the relevant parameters;

[0051] A sampling unit is used to perform DOE sampling on the local sampling space to obtain a local DOE training sample point set;

[0052] The computing unit is used to analyze and calculate the local DOE training sample point set to obtain the local DOE training sample dataset.

[0053] The merging unit is used to merge the local DOE training sample dataset into the initial DOE training sample dataset to obtain the encrypted DOE training sample dataset.

[0054] Optionally, the building unit includes:

[0055] The first determining subunit is used to determine the value range of multiple design variables based on the relevant parameters;

[0056] The calculation sub-unit is used to calculate the size of the initial sampling space based on the value range of each design variable;

[0057] The second determining subunit is used to determine the radius of the local sampling space based on the size of the initial sampling space;

[0058] A sub-unit is generated to generate an initial local sampling space based on the optimal point and the radius of the local sampling space;

[0059] The third determining subunit is used to determine the local sampling space based on the initial local sampling space and the initial design space.

[0060] Optionally, the third determining subunit is further configured to:

[0061] The initial local sampling space is compared with the initial design space to determine whether the initial local sampling space exceeds the initial design space.

[0062] If the initial local sampling space exceeds the initial design space, the intersection of the initial local sampling space and the initial sampling space is determined as the local sampling space for this iteration;

[0063] If the initial local sampling space does not exceed the initial design space, the initial local sampling space is determined as the local sampling space for this iteration.

[0064] Optionally, the acquisition module includes:

[0065] The sampling submodule is used to perform DOE sampling on multiple influencing indicators of automotive sound insulation performance to obtain DOE training sample point sets and DOE validation sample point sets.

[0066] The analysis and calculation submodule is used to analyze and calculate the sample points in the DOE training sample point set and the DOE verification sample point set to obtain the initial DOE training sample dataset and the DOE verification sample dataset.

[0067] Optionally, the device further includes:

[0068] The training module is used to train multiple surrogate models using the initial DOE training sample dataset and the DOE validation sample dataset, respectively.

[0069] The second determining module is used to determine the trained agent model combination as the candidate agent model group.

[0070] Optionally, the first determining module includes:

[0071] The first determining submodule is used to determine the accuracy of each influencing indicator under each proxy model in the candidate proxy model group for each influencing indicator;

[0072] The second determining submodule is used to determine the proxy model with the highest accuracy corresponding to each of the aforementioned influencing indicators as the target proxy model corresponding to the influencing indicator;

[0073] The combination submodule is used to combine the target proxy models corresponding to all the influencing indicators to obtain the target proxy model group.

[0074] Optionally, the optimization module includes:

[0075] The optimization submodule is used to globally optimize multiple influencing indicators of automotive sound insulation performance based on the target proxy model group to obtain the optimal point.

[0076] The third determining submodule is used to determine the actual index response of the optimal point at the real model based on the optimal point.

[0077] Thirdly, this application provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0078] Memory, used to store computer programs;

[0079] When a processor executes a program stored in memory, it implements the multidisciplinary design optimization method for automotive sound insulation performance as described in any of the first aspects.

[0080] Fourthly, this application provides a computer-readable storage medium storing a program for a multidisciplinary design optimization method for automotive sound insulation performance, wherein when the program for the multidisciplinary design optimization method for automotive sound insulation performance is executed by a processor, it implements the steps of the multidisciplinary design optimization method for automotive sound insulation performance as described in any of the first aspects.

[0081] The technical solutions provided in this application have the following advantages compared with the prior art:

[0082] This application embodiment constructs a target proxy model group based on multiple target proxy models with the highest accuracy, which greatly improves the accuracy of the target proxy model group and ensures the accuracy of each influencing index when the target proxy model group performs multidisciplinary design optimization. Moreover, by updating the target proxy model group using an encrypted DOE training sample dataset with increased local sampling density near the optimal point, the accuracy of the updated target proxy model group near the optimal point is improved, effectively improving the global accuracy of the proxy model. This avoids the situation where a scheme that has been optimized to meet the target through the proxy model is found to be unsatisfactory in actual experiments. This application embodiment can achieve joint optimization of automotive sound insulation performance by improving the accuracy of the proxy model near the optimal point and selecting the target proxy model with the highest accuracy. Attached Figure Description

[0083] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0084] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0085] Figure 1 A flowchart of a multidisciplinary design optimization method for automotive sound insulation performance provided in this application embodiment;

[0086] Figure 2A flowchart illustrating another multidisciplinary design optimization method for automotive sound insulation performance provided in this application embodiment;

[0087] Figure 3 A schematic diagram comparing the FDVGA algorithm with typical traditional optimization algorithms provided in the embodiments of this application;

[0088] Figure 4 A schematic diagram showing the results of some influencing indicators after the first round of optimization before updating the target proxy model provided in this application embodiment.

[0089] Figure 5 A schematic diagram showing the comparison between the results of some influencing indicators output by the proxy model and the actual reverse results after the first round of optimization before updating the target proxy model provided in this embodiment of the application;

[0090] Figure 6 A schematic diagram showing the results of some influencing indicators after the final optimization of the target agent model provided in this application embodiment;

[0091] Figure 7 This is a schematic diagram showing the effect comparison of local DOE training provided in the embodiments of this application;

[0092] Figure 8 A structural diagram of a multidisciplinary design optimization device for automotive sound insulation performance provided in this application embodiment;

[0093] Figure 9 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0094] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0095] Current multidisciplinary design optimization processes generally include three steps: Design of Experiment (DOE), surrogate model establishment, and optimization design.

[0096] Design of Experiments (DOE) is based on probability theory and mathematical statistics. It uses a small number of representative samples for experimentation or high-precision simulation analysis to quickly explore the entire design space and obtain ideal experimental results economically and scientifically. Surrogate models utilize the actual experimental sample data obtained from DOE. Through surrogate model algorithms, an approximate model between design parameters and product performance is established, replacing experimental or high-precision simulation analysis models. This efficiently predicts response values, especially for complex, large-scale, high-dimensional, nonlinear, and costly optimization problems with multiple extrema. The application of surrogate models enables optimization algorithms, such as evolutionary algorithms, that require extensive experimentation. The optimization design process combines optimization algorithms and surrogate models, using the approximate calculations of the surrogate model to solve for the optimal solution of one or more objectives within the design space. Among optimization algorithms, evolutionary algorithms, unlike mathematical optimization algorithms, do not depend on the properties of the problem, making them applicable to a wider range of problems. They can solve nonlinear, non-differentiable, non-convex, discrete, and combinatorial optimization problems. Compared to gradient optimization methods, they are less prone to getting trapped in local optima and are widely used in multidisciplinary optimization.

[0097] The DOE method includes full factorial design, partial factorial design, Plackett-Burman design, central composite design, Latin hypercube design, grid design, and other methods.

[0098] High-frequency noise in automobiles is one of the main NVH (noise, vibration, and harshness) problems. Especially for electric vehicles, which are an important future direction of automobile development, the lack of the masking effect of traditional internal combustion engine noise and the high-frequency characteristics of electromagnetic noise that are more easily perceived by the human ear have placed higher demands on the high-frequency noise insulation performance of automobiles.

[0099] Applying multidisciplinary design optimization in the automotive field can significantly shorten the R&D cycle and reduce R&D costs. Traditional optimization methods for high-frequency noise insulation performance in automobiles involve first identifying weaknesses, optimizing key components to achieve performance targets, and then reducing excess performance redundancy based on experience to control costs. This method often fails to achieve the optimal goal of lowest cost or lightest weight. Furthermore, the final result of traditional multidisciplinary automotive design optimization often fails to meet design requirements in actual production.

[0100] Therefore, this application provides a multidisciplinary design optimization method, apparatus, and electronic device for automotive sound insulation performance. The design objective of this embodiment is to achieve the lowest cost while optimizing multiple high-frequency influence indicators.

[0101] Based on the above, such as Figure 1 and Figure 2 As shown in the embodiments of this application, the multidisciplinary design optimization method for automotive sound insulation performance may include the following steps:

[0102] Step S101: Obtain relevant parameters for multidisciplinary design optimization of automotive sound insulation performance, determine multiple influencing indicators that need to be optimized for automotive sound insulation performance, perform DOE sampling of multiple influencing indicators, and determine the initial DOE training sample dataset and DOE validation sample dataset based on the DOE sampling results.

[0103] In this embodiment, the relevant parameters for multidisciplinary design optimization of automotive sound insulation performance include: design variables, design space A. design_space DOE sampling methods, sampling scale, objectives, and performance constraints, etc.

[0104] For example, since numerous factors affect the overall sound insulation performance of a vehicle, including flow resistance, porosity, thermal characteristic length, viscous characteristic length, density, sound-absorbing layer thickness, coverage, and sheet metal thickness, a total of 205 design variables were identified. After correlation analysis, the number of design variables was reduced to 101. The design space was determined based on physical constraints and the feasible space in actual engineering. Considering the uniformity of sampling, the Latin hypercube sampling method was used in this example. The 101 design variables were coded and automatically calculated using acoustic performance solving software, ultimately yielding 78 quantified influence indicators.

[0105] In this embodiment, the goal of multidisciplinary design optimization of automotive sound insulation performance is to minimize cost, and the performance constraints are set based on the requirements of influencing indicators. Considering the dimensions of the comprehensive design space and the experimental time cost, the sampling scale for DOE training can be determined as follows: the number of sampling points must be greater than 5 × the design dimension (number of design variables). For example, the sampling scale can be set to 1000 sets of sample points, and DOE validation sampling can be performed on 20 sets of sample points. DOE sampling of multiple influencing indicators is carried out, and the initial DOE training sample dataset and DOE validation sample dataset are determined based on the sampling results.

[0106] Step S102: Based on the accuracy, determine the target proxy model corresponding to each impact indicator from the multiple candidate proxy model groups corresponding to each impact indicator, and obtain the target proxy model group composed of the target proxy models corresponding to all impact indicators.

[0107] In this embodiment of the application, multiple surrogate models can be trained in advance using the initial DOE training sample dataset and the DOE verification sample dataset, and the combination of the trained surrogate models can be determined as the candidate surrogate model group.

[0108] Multiple surrogate models can be different types of surrogate model algorithms. For example, the candidate algorithms in this embodiment can be Least Square, Kriging, Marginal Gaussian Process (MGP), Kriging surrogate model based on partial least squares (KPLS), or Support Vector Regression (SVM regression), etc. Multiple surrogate models can also be surrogate model algorithms that are partially or completely the same type.

[0109] Least squares is a candidate algorithm for surrogate models due to its fast training speed and low time cost, and can be used for almost any problem. Among commonly used algorithms, the classic Kriging algorithm is too time-consuming to train for high-dimensional, large-sample problems, so the KPLS algorithm, which is more suitable for large-scale problems, has been replaced. This algorithm uses the PLS (partial least squares) method to construct the covariance kernel, reducing the dimensionality of the algorithm's internal parameters and significantly improving the training speed of the surrogate model.

[0110] In this embodiment of the application, each candidate proxy model group includes multiple proxy models, and the form of the candidate proxy model group is shown in Formula 1:

[0111]

[0112] The first subscript of f represents the index of the dependent variable (generally including optimization indicators and constraint indicators) that needs to be predicted by the surrogate model. The dependent variable is the influencing indicator. The second subscript is the index of the surrogate model. k is the total number of dependent variables that need to be predicted by the surrogate model. l is the number of surrogate models.

[0113] In this step, for each influencing indicator, the accuracy of the influencing indicator under each proxy model in the candidate proxy model group can be determined; then, the proxy model with the highest accuracy corresponding to each influencing indicator is determined as the target proxy model corresponding to the influencing indicator; then, the target proxy models corresponding to all influencing indicators are combined to obtain the target proxy model group.

[0114] In other words, the accuracy of each influencing indicator under different surrogate models is calculated separately, and the surrogate model with the highest accuracy (smallest error) is determined as the target surrogate model corresponding to that influencing indicator. The target surrogate models corresponding to all influencing indicators are combined to obtain the target surrogate model group. This avoids the situation where only a single surrogate model is used in a single multidisciplinary design optimization design project, which cannot guarantee that each specific influencing indicator has sufficient accuracy.

[0115] The target surrogate model set takes the form shown in Formula 2, where the i-th optimal surrogate model is the surrogate model with the smallest error for the dependent variable (i.e., the influencing index), i.e., there exists a formula as follows:

[0116] The relationship is shown in Equation 3.

[0117]

[0118]

[0119] Where ε() represents the error index value of the target surrogate model within the parentheses.

[0120] Accuracy can be calculated using mean absolute error (MAE), mean absolute percentage error (MAPE), or multiple correlation coefficient (R²). 2 The accuracy indicators are evaluated using metrics such as , and the expression for the above accuracy indicators is shown in Formula 4-6:

[0121]

[0122]

[0123]

[0124] Where n is the number of sample points. ym represents the predicted value of the m-th sample point in the surrogate model, and ym represents the actual experimental response value of the m-th sample point. This is the mean of the true responses of all test samples.

[0125] In one embodiment of this application, for all influencing indicators, the mean absolute percentage error of the KPLS algorithm is smaller than that of the SVM regression algorithm and the partial least squares algorithm, i.e., the accuracy is the highest. The accuracy of some influencing indicators under different surrogate models is shown in Table 1. Therefore, the target surrogate model group in this embodiment is a combination of candidate surrogate models for the KPLS algorithm for 78 influencing indicators. In more common cases, the target surrogate models for each influencing indicator are different, and the target surrogate model group should be combined according to the surrogate model with the best actual accuracy for each influencing indicator.

[0126] Table 1. Accuracy of different proxy models for global optimization of vehicle sound insulation performance

[0127]

[0128]

[0129] Step S103: Based on the target agent model group, perform global optimization to obtain the optimal point, and determine the actual indicator response based on the optimal point;

[0130] In this step, multiple influencing indicators of automotive sound insulation performance can be globally optimized based on the target proxy model group to obtain the optimal point; based on the optimal point, the actual indicator response of the optimal point at the real model is determined.

[0131] In this step, a target agent model group is used to replace the high-frequency simulation software. Considering the characteristics of multiple decision variables and high objective dimension in this embodiment, a global optimization algorithm is applied to perform global optimization with the goal of minimizing cost, thereby obtaining an approximate optimal solution. The approximate optimal solution corresponds to the optimal point in the corresponding iteration round, which is used to design variables. That is, the optimal point. The embodiments of this application, by using a global optimization algorithm, can avoid getting trapped in local optima.

[0132] Global optimization algorithms include: Non-dominated sorting genetic algorithm (NSGA-II), Non-dominated sorting genetic algorithm (NSGA-III), Population-based stochastic optimization algorithm (PSO), Differential evolution algorithm (DE), Multi-objective evolutionary algorithm (MOED / D), and fuzzy decision variable-based genetic algorithm (FDVGA).

[0133] Among them, the FDVGA algorithm is more suitable for high-dimensional problems with large-scale decision variables compared to other algorithms, such as Figure 3 As shown, the test results of the standard test function DTLZ3 with 500 design dimensions show that the solution set of this algorithm is uniformly distributed on the optimal surface, while the traditional algorithms NSGA-II and MOED / D get trapped in local optima and the solution set does not converge.

[0134] Finally, the optimal point can be substituted into high-frequency simulation software for analysis and calculation to obtain the actual index response y′ of the current optimal point. opt .

[0135] Step S104: If the difference between the actual index response and the optimization constraint is greater than a preset threshold, the target proxy model group is updated using the encrypted DOE training sample dataset corresponding to the optimal point. The encrypted DOE training sample dataset is constructed based on the optimal point and the value range of multiple design variables in the relevant parameters. The updated target proxy model is then used for global optimization until the difference between the actual index response and the optimization constraint is less than or equal to the preset threshold. Finally, the updated target proxy model group is output for use in automotive sound insulation performance optimization.

[0136] If the difference between the actual index response and the optimization constraints is less than or equal to a preset threshold (e.g., difference ≤ 1dB), then the actual index response of the current optimal point achieves the optimization objective, the optimization is complete, and the optimal solution y is obtained. opt =y′ opt , optimal point x opt =x′ opt .

[0137] This application uses the relative error between the actual index response and the constraint conditions, i.e., the actual index meets the standard and has little redundancy, as the condition for exiting the iteration. This relative accuracy can be given according to engineering specifications or experience, which is easier to implement, more efficient, and more in line with the optimization goal itself. It avoids the difficulty in intuitively determining the accuracy requirements that meet the actual engineering needs when judging whether to exit the iteration by comparing the accuracy of the predicted value of the surrogate model with the actual index response for new engineering problems.

[0138] Some of the influencing indicators obtained in step S104 are as follows: Figure 4 As shown, after optimization, the actual index responses of each influencing index all exhibited failure to meet performance constraints at points Y24 to Y34. And as... Figure 5 As shown, comparing the actual simulation results of the target surrogate model reveals that the output impact index of the target surrogate model has met the target, indicating that the target surrogate model lacks local accuracy at the optimal point. The current optimal point has not yet met the optimization requirements and further optimization is necessary.

[0139] In traditional multidisciplinary optimization design for automobiles, the Design of Excellence (DOE) stage often focuses on the randomness and uniformity of the overall sample across the entire design range. This effectively improves the global accuracy of the subsequent surrogate model and prevents convergence to local optima during optimization. However, a drawback is that high global accuracy does not necessarily mean sufficient accuracy near the optimum. Often, solutions optimized using surrogate models fail to meet the targets in actual testing.

[0140] Therefore, in this embodiment of the application, when the difference between the actual index response and the optimization constraint at the optimal point is greater than a preset threshold, the target proxy model group is updated by obtaining the encrypted DOE training sample dataset corresponding to the optimal point. The encrypted DOE training sample dataset corresponding to the optimal point is a DOE training sample dataset with increased local DOE sampling density within a preset range near the optimal point, so as to improve the accuracy near the optimal point through the encrypted DOE training sample dataset.

[0141] For example, the target agent model group composed of KPLS algorithm can be retrained using an encrypted DOE training sample dataset of 1050 sets of data to obtain an updated target agent model group.

[0142] Then, the updated target proxy model is used for global optimization until the difference between the actual index response and the optimization constraints is less than or equal to a preset threshold. Finally, the updated target proxy model group is output for use in optimizing automotive sound insulation performance.

[0143] For example, in the next iteration after updating the target agent model group, the updated target agent model group can be used to perform global optimization with the goal of minimizing cost, applying the FDVGA evolutionary algorithm to obtain a new optimal point. This new optimal point is then substituted into high-frequency simulation software for analysis and calculation to obtain new actual index responses. After checking and comparing, it is found that the values ​​of all influencing indicators in the obtained optimal point have met the standards and have minimal redundancy. For some indicators, such as... Figure 6 As shown, optimization is complete.

[0144] This application embodiment constructs a target proxy model group based on multiple target proxy models with the highest accuracy, which greatly improves the accuracy of the target proxy model group and ensures the accuracy of each influencing index when the target proxy model group performs multidisciplinary design optimization. Moreover, by updating the target proxy model group using an encrypted DOE training sample dataset with increased local sampling density near the optimal point, the accuracy of the updated target proxy model group near the optimal point is improved, effectively improving the global accuracy of the proxy model. This avoids the situation where a scheme that has been optimized to meet the target through the proxy model is found to be unsatisfactory in actual experiments. This application embodiment can achieve joint optimization of automotive sound insulation performance by improving the accuracy of the proxy model near the optimal point and selecting the target proxy model with the highest accuracy.

[0145] In another embodiment of this application, step S104 updates the target agent model group using the encrypted DOE training sample dataset corresponding to the optimal point, including:

[0146] Step S201: Obtain an encrypted DOE training sample dataset with increased local sampling density within a preset range of the optimal point;

[0147] In this step, a preset range can be determined first around the optimal point, and then the local sampling density can be increased within the preset range. This is equivalent to increasing the local sampling density around the optimal point in the initial DOE training sample dataset to obtain the encrypted DOE training sample dataset.

[0148] Step S202: Train each target agent model in the target agent model group using the encrypted DOE training sample dataset to obtain the updated target agent model group.

[0149] The target agent model group is updated by training each target agent model in the target agent model group using an encrypted DOE training sample dataset with increased local sampling density near the optimal point.

[0150] This application obtains an encrypted DOE training sample dataset by increasing the local sampling density near the optimal point, and uses the encrypted DOE training sample dataset to update the target proxy model group, thereby improving the accuracy of the updated target proxy model group near the optimal point and effectively improving the global accuracy of the proxy model. This avoids situations where a solution that has been optimized to meet the target through the proxy model is found to be unsatisfactory in actual experiments.

[0151] In another embodiment of this application, step S201, obtaining an encrypted DOE training sample dataset with increased local sampling density within a preset range of the optimal point, includes:

[0152] Step S301: Construct a local sampling space based on the optimal point and the value range of multiple design variables in the relevant parameters;

[0153] In this step, the optimal point x′ can be determined. opt Determine the local sampling space nearby.

[0154] Step S302: Perform DOE sampling on the local sampling space to obtain a local DOE training sample point set;

[0155] In this step, a local DOE training sample point set can be constructed by applying a selected sampling method within the local sampling space. For example, a local sampling space A can be used to construct the local DOE training sample point set. opt_space The Latin hypercube sampling method is applied in the space. In this example, the number of sampling points is 5% of the number of initial DOE training sampling points, that is, 50 sets of sampling points are used to form a local DOE training sample point set.

[0156] Step S303: Analyze and calculate the local DOE training sample point set to obtain the local DOE training sample dataset;

[0157] The local DOE training sample point set is used for experiments or high-precision analysis models (such as high-frequency simulation software) to calculate the local DOE training sample dataset.

[0158] Step S304: Merge the local DOE training sample dataset into the initial DOE training sample dataset to obtain the encrypted DOE training sample dataset.

[0159] Since the local DOE training sample dataset is obtained by locally increasing the sampling density of the local sampling space near the optimal point, merging this local DOE training sample dataset into the initial DOE training sample dataset results in an encrypted DOE training sample dataset with increased sampling density near the optimal point. Therefore, the encrypted DOE training sample dataset can be used to update the target proxy model group, improving the accuracy of the updated target proxy model group near the optimal point, effectively improving the global accuracy of the proxy model. This achieves a significant improvement in the accuracy near the existing optimal solution while maintaining the overall accuracy of the proxy model. Figure 7 As shown. Due to the stability of the overall surrogate model's accuracy, even if the previous round of optimization before the update gets stuck in a local optimum due to the surrogate model's accuracy issues, the current round of optimization after the update can still optimize to near the global optimum, avoiding situations where a solution that has been optimized to meet the target through the surrogate model is found to be unsatisfactory in actual experiments.

[0160] In another embodiment of this application, step S301 constructs a local sampling space based on the optimal point and the value ranges of multiple design variables among the relevant parameters, including:

[0161] Step S401: Determine the value range of multiple design variables based on the relevant parameters;

[0162] Step S402: Calculate the size of the initial sampling space based on the value range of each design variable;

[0163] The ranges of values ​​for each design variable can be combined to obtain the size of the initial sampling space, as shown in Formula 7:

[0164] x range =[x 1max -x 1min ,x 2max -x 2min ,…,x pmax -x pmin 7)

[0165] Where p is the number of design variables, x pmin x is the lower bound of this design variable. pmax Design variable x p The upper boundary.

[0166] Step S403: Determine the radius of the local sampling space based on the size of the initial sampling space;

[0167] The radius r of the local sampling space is determined based on the size of the initial sampling space, and is obtained using Equation 8:

[0168] r = α·x range 0 < α < 0.5 8)

[0169] Wherein, α is a constant, which can be taken as an empirical value based on the actual problems of the car's sound insulation performance. For example, α can be set to 0.1.

[0170] Since the value ranges of multiple design variables are preset, the size of the initial sampling space is determined based on the value ranges of multiple design variables, and thus the radius of the local sampling space is stable. In other words, this application selects a stable sampling radius to optimize the sound insulation performance of automobiles based solely on the size of the initial sampling space and engineering experience, which has better robustness and avoids defining the size of the local sampling space in an initial sampling point area, thus avoiding overfitting problems.

[0171] Step S404: Generate an initial local sampling space based on the optimal point and the radius of the local sampling space;

[0172] In this step, the initial local sampling space can be generated according to Formula 9, that is, using the existing optimal point x′ opt Centered on r, with r as the upper and lower bounds, the initial local sampling space is generated as shown in the following equation:

[0173] A′ opt_space =x′ opt ±r 9)

[0174] For example, the existing optimal point x′ can be used. opt Centered on this, the upper bound is increased by 0.1 times the initial sampling space size, and the lower bound is decreased by 0.1 times the initial sampling space size to generate the initial local sampling space.

[0175] Step S405: Determine the local sampling space based on the initial local sampling space and the initial design space.

[0176] In one embodiment of this application, the initial local sampling space can be compared with the initial design space to determine whether the initial local sampling space exceeds the initial design space; if the initial local sampling space exceeds the initial design space, the intersection of the initial local sampling space and the initial sampling space is determined as the local sampling space of this iteration; if the initial local sampling space does not exceed the initial design space, the initial local sampling space is determined as the local sampling space of this iteration.

[0177] In other words, the initial local sampling space can be compared with the initial design space to check if the initial local sampling space exceeds the initial design space. If so, the initial local sampling space is reduced to within the initial design space (e.g., x). k The upper bound of the initial local sampling space is > x kmax Then xk The upper bound is changed to x kmax That is, the intersection of the initial local sampling space and the design space is taken as the local sampling space A for this iteration. opt_space As shown in Equation 10 below:

[0178] A opt_space =A′ opt_space ∩A design_space 10)

[0179] This application embodiment calculates the size of the initial sampling space based on the value range of each design variable, determines the radius of the local sampling space based on the size of the initial sampling space, then determines the initial local sampling space based on the optimal point and the radius of the local sampling space, and finally determines the local sampling space based on the initial local sampling space and the initial design space. This realizes the construction of a local sampling space based on the range of each influencing index and the optimal point, so that the constructed local sampling space is near the optimal point, which facilitates the local increase of sampling density in the local sampling space near the optimal point.

[0180] In another embodiment of this application, step S101 performs DOE sampling of multiple influencing indicators, and determines the initial DOE training sample dataset and DOE validation sample dataset based on the DOE sampling results, including:

[0181] Step S501: Perform DOE sampling on multiple influencing indicators of automotive sound insulation performance to obtain DOE training sample point set and DOE verification sample point set.

[0182] In this step, DOE sampling is performed on multiple influencing indicators of automotive sound insulation performance, and DOE training sample point set and DOE validation sample point set are generated after sampling.

[0183] Step S502: Analyze and calculate the sample points in the DOE training sample point set and the DOE verification sample point set to obtain the initial DOE training sample dataset and the DOE verification sample dataset.

[0184] Experiments or high-precision analysis model calculations are performed on the DOE training sample point set and DOE validation sample point set to obtain the DOE training sample dataset and DOE validation sample dataset containing accurate experimental results.

[0185] In practical applications, since this example is a large-scale, high-dimensional optimization problem, the cost of using real vehicle testing is too high to be feasible. Therefore, high-frequency simulation software is used, and the DOE training sample point set and DOE validation sample point set are substituted for analysis and calculation. The entire process is automated through multidisciplinary optimization commercial software to obtain the initial DOE training sample dataset and DOE validation sample dataset.

[0186] In another embodiment of this application, a multidisciplinary design optimization device for automotive sound insulation performance is also provided, such as... Figure 8 As shown, it includes:

[0187] The acquisition module 11 is used to acquire relevant parameters for multidisciplinary design optimization of automotive sound insulation performance, determine multiple influencing indicators that need to be optimized for automotive sound insulation performance, perform DOE sampling of multiple influencing indicators, and determine the initial DOE training sample dataset and DOE validation sample dataset based on the DOE sampling results.

[0188] The first determining module 12 is used to determine the target proxy model corresponding to each impact indicator from multiple candidate proxy model groups corresponding to each impact indicator based on the accuracy, so as to obtain a target proxy model group composed of target proxy models corresponding to all impact indicators.

[0189] Optimization module 13 is used to perform global optimization based on the target agent model group to obtain the optimal point, and to determine the actual indicator response based on the optimal point;

[0190] The update module 14 is used to update the target proxy model group using the encrypted DOE training sample dataset corresponding to the optimal point if the difference between the actual index response and the optimization constraint is greater than a preset threshold. The encrypted DOE training sample dataset is constructed based on the optimal point and the value range of multiple design variables in the relevant parameters. The updated target proxy model is used for global optimization until the difference between the actual index response and the optimization constraint is less than or equal to the preset threshold. The final updated target proxy model group is then output for use in automotive sound insulation performance optimization.

[0191] Optionally, the update module includes:

[0192] The acquisition submodule is used to acquire an encrypted DOE training sample dataset with increased local sampling density within a preset range of the optimal point;

[0193] The training submodule is used to train each target agent model in the target agent model group using the encrypted DOE training sample dataset to obtain the updated target agent model group.

[0194] Optionally, the acquisition submodule includes:

[0195] A construction unit is used to construct a local sampling space based on the optimal point and the value range of multiple design variables in the relevant parameters;

[0196] A sampling unit is used to perform DOE sampling on the local sampling space to obtain a local DOE training sample point set.

[0197] The computing unit is used to analyze and calculate the local DOE training sample point set to obtain the local DOE training sample dataset.

[0198] The merging unit is used to merge the local DOE training sample dataset into the initial DOE training sample dataset to obtain the encrypted DOE training sample dataset.

[0199] Optionally, the building unit includes:

[0200] The first determining subunit is used to determine the value range of multiple design variables based on the relevant parameters;

[0201] The calculation sub-unit is used to calculate the size of the initial sampling space based on the value range of each design variable;

[0202] The second determining subunit is used to determine the radius of the local sampling space based on the size of the initial sampling space;

[0203] A sub-unit is generated to generate an initial local sampling space based on the optimal point and the radius of the local sampling space;

[0204] The third determining subunit is used to determine the local sampling space based on the initial local sampling space and the initial design space.

[0205] Optionally, the third determining subunit is further configured to:

[0206] The initial local sampling space is compared with the initial design space to determine whether the initial local sampling space exceeds the initial design space.

[0207] If the initial local sampling space exceeds the initial design space, the intersection of the initial local sampling space and the initial sampling space is determined as the local sampling space for this iteration;

[0208] If the initial local sampling space does not exceed the initial design space, the initial local sampling space is determined as the local sampling space for this iteration.

[0209] Optionally, the acquisition module includes:

[0210] The sampling submodule is used to perform DOE sampling on multiple influencing indicators of automotive sound insulation performance to obtain DOE training sample point sets and DOE validation sample point sets.

[0211] The analysis and calculation submodule is used to analyze and calculate the sample points in the DOE training sample point set and the DOE verification sample point set to obtain the initial DOE training sample dataset and the DOE verification sample dataset.

[0212] Optionally, the device further includes:

[0213] The training module is used to train multiple surrogate models using the initial DOE training sample dataset and the DOE validation sample dataset, respectively.

[0214] The second determining module is used to determine the trained agent model combination as the candidate agent model group.

[0215] Optionally, the first determining module includes:

[0216] The first determining submodule is used to determine the accuracy of each influencing indicator under each proxy model in the candidate proxy model group for each influencing indicator;

[0217] The second determining submodule is used to determine the proxy model with the highest accuracy corresponding to each of the aforementioned influencing indicators as the target proxy model corresponding to the influencing indicator;

[0218] The combination submodule is used to combine the target proxy models corresponding to all the influencing indicators to obtain the target proxy model group.

[0219] Optionally, the optimization module includes:

[0220] The optimization submodule is used to globally optimize multiple influencing indicators of automotive sound insulation performance based on the target proxy model group to obtain the optimal point.

[0221] The third determining submodule is used to determine the actual index response of the optimal point at the real model based on the optimal point.

[0222] In another embodiment of this application, an electronic device is also provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0223] Memory, used to store computer programs;

[0224] When the processor executes the program stored in the memory, it implements the multidisciplinary design optimization method for automotive sound insulation performance as described in any of the foregoing method embodiments.

[0225] The electronic device provided in this invention, through its processor executing a program stored in memory, constructs a target proxy model group based on multiple target proxy models with the highest accuracy. This significantly improves the accuracy of the target proxy model group, ensuring the accuracy of various influencing indicators when the target proxy model group performs multidisciplinary design optimization. Furthermore, by updating the target proxy model group using an encrypted DOE training sample dataset with increased local sampling density near the optimal point, the accuracy of the updated target proxy model group near the optimal point is improved, effectively enhancing the global accuracy of the proxy model. This avoids situations where a scheme optimized through the proxy model fails to meet the target in actual testing. This embodiment of the application achieves joint optimization of automotive sound insulation performance by improving the accuracy of the proxy model near the optimal point and selecting the target proxy model with the highest accuracy.

[0226] The communication bus 1140 mentioned in the above-mentioned electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus 1140 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0227] The communication interface 1120 is used for communication between the above-mentioned electronic device and other devices.

[0228] The memory 1130 may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0229] The processor 1110 mentioned above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0230] In another embodiment of this application, a computer-readable storage medium is also provided, on which a program for a multidisciplinary design optimization method for automotive sound insulation performance is stored. When the program for the multidisciplinary design optimization method for automotive sound insulation performance is executed by a processor, it implements the steps of the multidisciplinary design optimization method for automotive sound insulation performance described in any of the foregoing method embodiments.

[0231] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0232] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A multidisciplinary design optimization method for automotive sound insulation performance, characterized in that, include: Obtain relevant parameters for multidisciplinary design optimization of automotive sound insulation performance, determine multiple influencing indicators that need to be optimized for automotive sound insulation performance, perform DOE sampling of multiple influencing indicators, and determine the initial DOE training sample dataset and DOE validation sample dataset based on the DOE sampling results. Based on the accuracy, the target proxy model corresponding to each impact indicator is determined from the multiple candidate proxy model groups corresponding to each impact indicator, and the target proxy model group consisting of the target proxy models corresponding to all impact indicators is obtained. The candidate proxy model group is pre-trained based on the initial DOE training sample dataset and the DOE validation sample dataset. Based on the target agent model group, global optimization is performed to obtain the optimal point, and the actual indicator response is determined based on the optimal point; If the difference between the actual index response and the optimization constraint is greater than a preset threshold, the target proxy model group is updated using the encrypted DOE training sample dataset corresponding to the optimal point. The encrypted DOE training sample dataset is constructed based on the optimal point and the value range of multiple design variables in the relevant parameters. The updated target proxy model is then used for global optimization until the difference between the actual index response and the optimization constraint is less than or equal to the preset threshold. Finally, the updated target proxy model group is output for use in automotive sound insulation performance optimization. Update the target agent model group using the encrypted DOE training sample dataset corresponding to the optimal point, including: Obtain an encrypted DOE training sample dataset with increased local sampling density within a preset range of the optimal point; Obtaining an encrypted DOE training sample dataset with increased local sampling density within a preset range of the optimal point includes: constructing a local sampling space based on the optimal point and the value range of multiple design variables among the relevant parameters; performing DOE sampling on the local sampling space to obtain a local DOE training sample point set; analyzing and calculating the local DOE training sample point set to obtain a local DOE training sample dataset; and merging the local DOE training sample dataset into the initial DOE training sample dataset to obtain an encrypted DOE training sample dataset. The target agent model in the target agent model group is trained using the encrypted DOE training sample dataset to obtain the updated target agent model group.

2. The multidisciplinary design optimization method for automotive sound insulation performance according to claim 1, characterized in that, A local sampling space is constructed based on the optimal point and the value ranges of multiple design variables in the relevant parameters, including: The value range of multiple design variables is determined based on the aforementioned relevant parameters; The size of the initial sampling space is calculated based on the value range of each design variable; The radius of the local sampling space is determined based on the size of the initial sampling space; An initial local sampling space is generated based on the optimal point and the radius of the local sampling space; The local sampling space is determined based on the initial local sampling space and the initial design space.

3. The multidisciplinary design optimization method for automotive sound insulation performance according to claim 2, characterized in that, Determining the local sampling space based on the initial local sampling space and the initial design space includes: The initial local sampling space is compared with the initial design space to determine whether the initial local sampling space exceeds the initial design space. If the initial local sampling space exceeds the initial design space, the intersection of the initial local sampling space and the initial design space is determined as the local sampling space for this iteration. If the initial local sampling space does not exceed the initial design space, the initial local sampling space is determined as the local sampling space for this iteration.

4. The multidisciplinary design optimization method for automotive sound insulation performance according to claim 1, characterized in that, DOE sampling was performed on multiple influencing metrics. Based on the DOE sampling results, the initial DOE training sample dataset and DOE validation sample dataset were determined, including: DOE sampling was performed on multiple influencing indicators of automotive sound insulation performance to obtain DOE training sample point set and DOE validation sample point set; The sample points in the DOE training sample point set and the DOE validation sample point set are analyzed and calculated to obtain the initial DOE training sample dataset and the DOE validation sample dataset.

5. The multidisciplinary design optimization method for automotive sound insulation performance according to claim 4, characterized in that, The method further includes: Multiple surrogate models were trained using the initial DOE training sample dataset and the DOE validation sample dataset, respectively. The trained agent models are identified as the candidate agent model group.

6. The multidisciplinary design optimization method for automotive sound insulation performance according to claim 1, characterized in that, Based on accuracy, from the multiple candidate surrogate model groups corresponding to each impact indicator, the target surrogate model corresponding to each impact indicator is determined, resulting in a target surrogate model group consisting of target surrogate models corresponding to all impact indicators, including: For each impact metric, determine the accuracy of the impact metric under each proxy model in the candidate proxy model group; The proxy model with the highest accuracy corresponding to each of the aforementioned influencing indicators is determined as the target proxy model corresponding to the aforementioned influencing indicator; The target proxy model group is obtained by combining the target proxy models corresponding to all the influencing indicators.

7. The multidisciplinary design optimization method for automotive sound insulation performance according to claim 1, characterized in that, Based on the target agent model group, global optimization is performed to obtain the optimal point. Based on the optimal point, the actual indicator response is determined, including: Based on the target proxy model group, the multiple influencing indicators of automotive sound insulation performance are globally optimized to obtain the optimal point. Based on the optimal point, determine the actual index response of the optimal point at the real model.

8. A multidisciplinary design optimization device for automotive sound insulation performance, characterized in that, include: The acquisition module is used to acquire relevant parameters for multidisciplinary design optimization of automotive sound insulation performance, determine multiple influencing indicators that need to be optimized for automotive sound insulation performance, perform DOE sampling of multiple influencing indicators, and determine the initial DOE training sample dataset and DOE validation sample dataset based on the DOE sampling results. The first determining module is used to determine the target proxy model corresponding to each influence indicator from multiple candidate proxy model groups corresponding to each influence indicator based on the accuracy, so as to obtain a target proxy model group composed of target proxy models corresponding to all influence indicators. The candidate proxy model group is pre-trained based on the initial DOE training sample dataset and DOE validation sample dataset. The optimization module is used to perform global optimization based on the target agent model group to obtain the optimal point, and to determine the actual indicator response based on the optimal point; The update module is used to update the target surrogate model group using an encrypted DOE training sample dataset corresponding to the optimal point if the difference between the actual index response and the optimization constraints is greater than a preset threshold. The encrypted DOE training sample dataset is constructed based on the optimal point and the value ranges of multiple design variables in the relevant parameters. Global optimization is performed using the updated target surrogate model group until the difference between the actual index response and the optimization constraints is less than or equal to the preset threshold. The final updated target surrogate model group is then output for use in automotive sound insulation performance optimization. Updating the target surrogate model group using the encrypted DOE training sample dataset corresponding to the optimal point includes: obtaining data within a preset range that increases the local sampling density of the optimal point. An encrypted DOE training sample dataset is obtained by increasing the local sampling density within a preset range of the optimal point, including: constructing a local sampling space based on the optimal point and the value range of multiple design variables in the relevant parameters; performing DOE sampling on the local sampling space to obtain a local DOE training sample point set; analyzing and calculating the local DOE training sample point set to obtain a local DOE training sample dataset; merging the local DOE training sample dataset into the initial DOE training sample dataset to obtain an encrypted DOE training sample dataset; and training each target agent model in the target agent model group using the encrypted DOE training sample dataset to obtain the updated target agent model group.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When a processor executes a program stored in a memory, it implements the multidisciplinary design optimization method for automotive sound insulation performance as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program for a multidisciplinary design optimization method for automotive sound insulation performance, which, when executed by a processor, implements the steps of the multidisciplinary design optimization method for automotive sound insulation performance as described in any one of claims 1-7.

Images