Multi-objective optimization method and system based on full-load passenger car framework
By using a multi-objective optimization method based on a fully load-bearing bus frame, and by leveraging parametric and AI models for collaborative optimization, the problems of insufficient lightweighting and difficulty in balancing multiple operating conditions in traditional designs were solved. This enabled lightweight design and performance optimization of the bus frame, shortening the design cycle and reducing costs.
Patent Information
- Application Number
- CN202511753612.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-26
- Publication Date
- 2026-03-17
AI Technical Summary
Traditional bus frame design suffers from low lightweighting levels, long design cycles, and difficulties in balancing multiple operating conditions. Existing finite element analysis methods are insufficient to meet the actual needs of multiple operating conditions.
A multi-objective optimization method based on a fully load-bearing bus frame is adopted. By constructing a parameterized model, defining test conditions, identifying and determining parameters, generating AI model training data, training the AI model, and optimizing the solution, the multi-condition and multi-objective collaborative optimization of the bus frame is achieved.
This achieved a balance between lightweight design and performance in the bus body structure, shortened the design cycle, improved design quality, and reduced R&D costs.
Smart Images
Figure CN121683014A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of passenger cars, in particular to a multi-objective optimization method and system based on a full-load passenger car frame. BACKGROUND
[0002] As a key structure of the vehicle body, the design of the full-load passenger car frame directly affects the safety, comfort and economy of the passenger car. The traditional passenger car frame design has the following problems: Low level of lightweight: In order to ensure strength, conservative design is often used, resulting in excessive weight of the vehicle body and increased energy consumption; long design cycle: relying on experience and physical tests, the iterative process of design-verification-modification is time-consuming and labor-intensive; difficulty in balancing multiple conditions: the passenger car frame needs to meet multiple conditions such as modal, bending, torsion, braking, turning, etc., and it is difficult for traditional methods to achieve global optimization.
[0003] In the prior art, although finite element analysis is used for passenger car frame design, it is mainly for single condition simulation analysis, which is difficult to meet the actual demand of multiple conditions. SUMMARY
[0004] The present application provides a multi-objective optimization method and system based on a full-load passenger car frame, which solves the problems of insufficient lightweight, long design cycle and difficulty in multi-condition optimization in the prior art, realizes lightweight design and performance balance of the passenger car body structure, and shortens the design cycle.
[0005] The present application provides a multi-objective optimization method based on a full-load passenger car frame, comprising the following steps: S1: Constructing a parameterized model: defining the geometric parameters of the passenger car frame as optimization design variables, generating a finite element model input file, establishing a passenger car frame finite element simulation model, and defining the parameter value range; S2: Defining test conditions: determining the typical conditions of the passenger car and their constraint conditions, and calculating the performance indicators under each condition through simulation to verify the simulation model; S3: Identifying and determining parameters: performing sensitivity analysis on the design parameters, and selecting and determining the design parameters; S4: Generating AI model training data: processing the design parameters to generate a design parameter dataset, and calculating the corresponding performance indicators through the simulation model; the design parameter dataset and the performance indicators are both used for subsequent AI model training; S5: Training AI model: constraining the design parameter dataset and the performance indicators, training the AI model using machine learning and deep learning methods, and verifying the accuracy of the AI model; S6: Optimization solution: using a pattern search solver to search for the best design parameters that meet the performance requirements, and verifying through simulation.
[0006] The beneficial effects of the above embodiments are as follows: the multi-objective optimization method achieves the synergistic optimization of multiple working conditions (bending, torsion, braking, etc.) and multiple objectives (lightweighting, strength, stiffness) through the synergistic effect of the finite element simulation model and the AI model, thereby realizing the lightweight design and performance balance of the bus body structure and shortening the design cycle.
[0007] Based on the above embodiments, this application can be further improved as follows: In one embodiment of this application, in step S1, the geometric parameters include the cross-sectional parameters of the front A-pillar, the front crossbeam, the gantry side window beam, the roof crossbeam, the side edge beam, the floor crossbeam, and the floor longitudinal beam. Technical effect: Covering the key load-bearing structures of the framework, and combining with the professional finite element tool HYPERMESH to improve model accuracy, ensuring the representativeness of the geometric parameters for overall performance.
[0008] In one embodiment of this application, in step S2, the typical operating conditions include bending, torsion, and braking conditions; the performance indicators include modal analysis, bending stiffness, and torsional stiffness; and the simulation model verification is achieved by comparing physical test data. Technical effect: Simulating extreme operating scenarios and calibrating the model using test data ensures the consistency between simulation results and actual performance, providing a reliable foundation for subsequent optimization.
[0009] In one embodiment of this application, in step S3, the sensitivity analysis calculates the degree of influence of each design parameter on the performance index using the variance threshold method, and filters out the design parameters with the highest sensitivity ranking. Technical effect: Focusing on highly sensitive parameters reduces the dimensionality of optimization variables, lowers computational complexity, and improves the targeting of parameter adjustments.
[0010] In one embodiment of this application, in step S4, the design parameter dataset is generated using Latin hypercube sampling. Technical effect: This ensures that the samples are uniformly distributed in the parameter space, providing high-quality input for AI model training.
[0011] In one embodiment of this application, in step S5, before training the AI model, it is necessary to create charts of design parameter data and performance index data, and use the Minimum Redundancy Maximum Correlation (MRMR) algorithm to sort the input features. The AI model training employs three methods: bagged tree ensemble, Gaussian process regression, and deep neural network, selecting the optimal method based on the root mean square error (RMSE) and coefficient of determination (R-squared). Technical effect: Optimizes the feature input order, integrates the advantages of multiple algorithms, and improves the model prediction accuracy.
[0012] This application also provides an optimization system based on a multi-objective optimization method for a fully load-bearing bus frame, comprising: Parametric modeling module: used to create a finite element simulation model of the bus frame based on the finite element model input file; The working condition simulation module is used to calculate performance indicators and verify the simulation model based on the set typical working conditions and constraints. Parameter filtering module: used to filter and determine design parameters through sensitivity analysis; Data generation module: used to generate design parameter datasets and calculate performance indicators through simulation models; AI training module: used to constrain design parameter datasets and performance metrics, train and validate the accuracy of AI models; Optimization Solver Module: Used to search for optimal design parameters using a pattern search solver and verify them through simulation.
[0013] Technical benefits: It automates the process from model building to optimization and verification, reduces manual intervention, and shortens the design cycle.
[0014] In one embodiment of this application, the parametric modeling module integrates the CATIA 3D modeling interface and the HYPERMESH finite element preprocessing function, and the AI training module supports MRMR algorithm feature sorting and multi-model accuracy comparison. Technical effects: Improves the correlation between the 3D model and the finite element model, optimizes the input features of the AI model, and further reduces prediction errors. Attached Figure Description
[0015] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.
[0016] Figure 1 This is a flowchart illustrating the steps of a multi-objective optimization method based on a fully load-bearing bus frame in an embodiment of this application. Figure 2 This is a schematic diagram of the finite element simulation model in the embodiments of this application; Figure 3 This is a schematic diagram illustrating the sensitivity analysis of parameters in the embodiments of this application; Figure 4 This is a schematic diagram of the design parameters to be optimized in the embodiments of this application; Figure 5 This is a schematic diagram of data samples after Latin hypercube sampling in an embodiment of this application; Figure 6 This is a schematic diagram of the MRMR sorting results in the embodiments of this application; Figure 7 This is a schematic diagram of the performance response results of the AI model's multi-method prediction in the embodiments of this application; Figure 8 This is a schematic diagram showing the comparison results of the accuracy of the AI models in the embodiments of this application; Figure 9 This is a schematic diagram of the optimized parameter analysis results in the embodiments of this application; Figure 10 This is a schematic diagram of simulation model comparison in the embodiments of this application. Detailed Implementation
[0017] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.
[0018] Example: like Figure 1 As shown, a multi-objective optimization method based on a fully load-bearing bus frame includes the following steps: S1: Construct a parametric model: Define the geometric parameters of the bus frame and use them as optimization design variables, generate a finite element model input file, establish a finite element simulation model of the bus frame, and define the parameter value range; The geometric parameters include the front A-pillar cross-section parameters, the front crossbeam cross-section parameters, the gantry side window beam cross-section parameters, the roof crossbeam cross-section parameters, the side edge beam cross-section parameters, the floor crossbeam cross-section parameters, and the floor longitudinal beam cross-section parameters.
[0019] A specific example of this step is as follows: (1) Taking a certain type of 12-meter fully load-bearing bus frame as an example, the profile cross-sectional dimensions (rectangular tube: height × width × wall thickness) of the seven key areas below the body are defined as follows: front A-pillar cross-section parameters, front crossbeam cross-section parameters, gantry side window beam cross-section parameters, roof crossbeam cross-section parameters, side beam cross-section parameters, floor crossbeam cross-section parameters, and floor longitudinal beam cross-section parameters; the basic parameter table is shown in Table 1 below:
[0020] (2) Complete the basic 3D data and finite element model: Complete the basic 3D data, use CATIA as the 3D simulation modeling tool to establish a parametric basic model of the vehicle skeleton, ensure the geometric correlation of each component, use HYPERMESH for preprocessing (mesh generation, material property definition and boundary condition setting), and generate a finite element simulation model, such as... Figure 2 As shown; (3) Define each parameter within a reasonable range of values, including the upper and lower limits of the value.
[0021] S2: Define test conditions: Determine the typical operating conditions and constraints of the bus, and verify the simulation model by simulating the performance indicators under each operating condition. The typical operating conditions include bending, torsion, braking, and turning conditions. The performance indicators include modal analysis, bending stiffness, and torsional stiffness. The simulation model is verified by comparing physical test data.
[0022] A specific example of this step is as follows: Define typical operating conditions and their constraints: Overall modal analysis of the vehicle frame, with the constraint requirement being the free state of the entire vehicle frame; Under bending conditions, the constraints are as follows: the front airbag connection point is constrained by 23 degrees of freedom, the center point of the front and rear airbag connection points is constrained by 123 degrees of freedom, and a force of 10000N is applied to the floor between the front and rear axles in three sections: front, middle and rear.
[0023] Under torsional conditions, the constraints are: Y-axis acceleration 0.8g, Z-axis acceleration -1g; Under braking conditions, the constraints are: X-axis acceleration -1g, Z-axis acceleration -2g.
[0024] Through simulation analysis, key performance indicators, including modes and stiffness, were extracted from the simulation results. In this example, the overall vehicle bending stiffness is 15258 N / mm, the front torsional stiffness is 4132 N·m / °, and the rear torsional stiffness is 4236 N·m / °; the first four overall modes are 7.6 Hz, 9.5 Hz, 13.4 Hz, and 16.3 Hz. Comparison with actual values in this example confirms the accuracy of the simulation model.
[0025] S3: Identify and determine parameters: Perform sensitivity analysis on design parameters, screen and determine design parameters; The sensitivity analysis uses the variance threshold method to calculate the degree of influence of each design parameter on performance indicators, and selects the design parameters with the highest sensitivity. The purpose is to identify the design parameters that play a key role in performance.
[0026] A specific example of this step is as follows: (1) Generate samples for the entire design: Generate samples within the full range of the profile section parameters in 7 regions (sample data is large and omitted) to ensure coverage of all possible parameter combinations; (2) Simulation to obtain performance indicators: Input the above parameter samples into the simulation model, and run the simulation to obtain modal and stiffness performance indicators; (3) Obtaining the sensitivity of indicators to parameters: Based on the variance threshold, sensitive values are identified, and statistical methods are used to analyze the degree of influence of each performance indicator on each design parameter, that is, the strength of the indicator change caused by parameter change. The analysis results are as follows: Figure 3 As shown; (4) Selecting influential parameters: Based on the sensitivity results, parameters that significantly affect performance indicators are selected to provide key directions for subsequent design optimization. In this example, the top 8 parameters with the highest sensitivity impact are selected as design parameters to be optimized, such as... Figure 4 As shown.
[0027] S4: Generate AI model training data: Process the design parameters to generate a design parameter dataset, and calculate the corresponding performance indicators through the simulation model; both the design parameter dataset and the performance indicators are used for subsequent AI model training. Select design parameters that significantly impact performance metrics and split them: For the parameters identified in the previous step that significantly impact performance metrics, categorize them according to their lower and upper limits. Use Latin hypercube sampling to generate a large set of parameters, ensuring the samples cover all parameter combinations in the target design space. In this example, Latin hypercube sampling is used to divide the data into 1000 samples as follows: Figure 5 As shown.
[0028] Parameter samples are input into the simulation analysis model, and performance index data is obtained after running the model. This data is one of the key inputs for training the AI model. The validity of the generated parameter samples and performance indices is verified to ensure that the data quality meets the requirements for AI model training.
[0029] S5: Training the AI model: Constrain the design parameters, dataset, and performance metrics, train the AI model using machine learning and deep learning methods, and verify the accuracy of the AI model; Choose an AI model. Before training the AI model, it is necessary to create charts of design parameter data and performance index data. The minimum redundancy maximum correlation (MRMR) algorithm is used to rank the input features. The AI model training employs three methods: bagged tree ensemble, Gaussian process regression, and deep neural networks. The optimal method is selected based on the root mean square error (RMSE) and coefficient of determination (R-squared). Details are as follows: (1) Loading of input design parameter data: Organize the training data and constrain the input data of design parameters and the response data of simulation analysis results. Split the data into training data (e.g. 80%) and test data (e.g. 20%) for machine learning and deep learning.
[0030] (2) Data Analysis: Create charts of design parameter input data and performance response data to understand the correspondence between design parameters and performance result statistics. Use the MRMR method to plot feature contribution. Next, use the Minimum Redundancy Maximum Correlation (MRMR) algorithm to rank the regression features, such as... Figure 6 As shown.
[0031] (3) Train AI models using three methods: bagged tree ensemble, Gaussian process regression machine learning, and deep neural network deep learning, based on the training data. Use test data to predict the performance response results of the AI models using the three methods respectively, and plot the root mean square error (RMSE) and coefficient of determination (R-squared) of all available models to compare the accuracy of the AI models, such as... Figure 7 , Figure 8 As shown. Save this AI model for process optimization.
[0032] S6: Optimization Solution: Use the pattern search solver to search for the optimal design parameters that meet the performance requirements, and verify them through simulation.
[0033] Specifically: (1) Select the parameters to be optimized: Define an array of optimization variables using the parameters, lower bound and upper bound extracted from the parameter table.
[0034] (2) Define design objectives: Define the body frame modes and stiffness as design objectives.
[0035] (3) Add constraints: Optimizing the variable array requires additional constraints to ensure that the optimized data is within a reasonable range of values.
[0036] In this example, a Gaussian regression process model is used, and the set of optimization data with the highest sensitivity is selected as the variable array. The performance targets of first-order mode 8Hz, second-order mode 9Hz, third-order mode 12Hz and vehicle stiffness 15000N / mm are defined.
[0037] (4) Parameter optimization: The results of the optimization variable array are analyzed using a pattern search solver to obtain the optimal solution. Simulation verification of the solution: The optimal parameters are calculated in the simulation analysis, and the optimal AI model solution is compared with the simulation analysis results to verify that they are the best design parameters, such as... Figure 9 , Figure 10 As shown.
[0038] A multi-objective optimization system based on a fully load-bearing bus frame, used to implement the above-mentioned multi-objective optimization method, includes: Parametric modeling module: used to create a finite element simulation model of the bus frame based on the finite element model input file; The working condition simulation module is used to calculate performance indicators and verify the simulation model based on the set typical working conditions and constraints. Parameter filtering module: used to filter and determine design parameters through sensitivity analysis; Data generation module: used to generate design parameter datasets and calculate performance indicators through simulation models; AI training module: used to constrain design parameter datasets and performance metrics, train and validate the accuracy of AI models; Optimization Solver Module: Used to search for optimal design parameters using a pattern search solver and verify them through simulation.
[0039] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: 1. This multi-objective optimization method can accurately identify and set design parameters to achieve optimal performance, realize lightweight design and performance balance of bus body structure, and improve design quality; at the same time, it can quickly predict performance indicators and finally confirm them through simulation analysis, thus shortening the design cycle. 2. This multi-objective optimization method reduces the number of physical experiments, lowers the consumption of human and computational resources, and reduces R&D costs; 3. This multi-objective optimization method provides the optimal solution based on the sensitivity analysis results, offering designers a scientific basis for decision-making.
[0040] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.
Claims
1. A multi-objective optimization method based on a full load passenger car body frame, characterized by, The method comprises the following steps: S1: defining the geometric parameters of the passenger car frame as optimization design variables, generating a finite element model input file, establishing a passenger car frame finite element simulation model, and defining the parameter value range; S2: determining the typical working conditions of the passenger car and the constraint conditions, and verifying the simulation model by calculating the performance indicators under each working condition; S3: performing sensitivity analysis on the design parameters, and screening and determining the design parameters; S4: processing the design parameters to generate a design parameter dataset, and calculating the corresponding performance indicators through the simulation model; S5: constraining the design parameter dataset and the performance indicators, and training an AI model using machine learning and deep learning methods; S6: searching for the best design parameters that meet the performance requirements using a pattern search solver.
2. The multi-objective optimization method of claim 1, wherein: In step S1, the geometric parameters include front wall A column section parameters, front wall beam section parameters, gantry side window beam section parameters, roof beam section parameters, side wall beam section parameters, floor beam section parameters, and floor longitudinal beam section parameters.
3. The multi-objective optimization method of claim 1, wherein: In step S2, the typical working conditions include bending working conditions, torsion working conditions, and braking working conditions, and the performance indicators include modal, bending stiffness, and torsional stiffness.
4. The multi-objective optimization method of claim 1, wherein: In step S3, the sensitivity analysis calculates the influence of each design parameter on the performance indicators through the variance threshold method, and screens the design parameters with high sensitivity.
5. The multi-objective optimization method of claim 1, wherein: In step S4, the design parameter dataset is generated using Latin hypercube sampling.
6. The multi-objective optimization method of claim 1, wherein: In step S5, a chart of design parameter data and performance indicator data is created before training the AI model, and the minimum redundancy maximum correlation algorithm is used to sort the input features. The AI model training uses three methods: bagged tree integration, Gaussian process regression, and deep neural network. The optimal method is selected based on the root mean square error and the determination coefficient.
7. A multi-objective optimization system based on a full load passenger car body frame, characterized by, The multi-objective optimization method according to any one of claims 1-6 comprises: A parameterized modeling module for establishing a passenger car frame finite element simulation model according to a finite element model input file; A working condition simulation module for calculating performance indicators and verifying the simulation model according to the set typical working conditions and constraint conditions; A parameter screening module for screening and determining the design parameters through sensitivity analysis; A data generation module for generating a design parameter dataset and calculating performance indicators through the simulation model; An AI training module for constraining the design parameter dataset and the performance indicators, training and verifying the AI model accuracy; An optimization solving module for searching for the best design parameters using a pattern search solver and verifying through simulation.
8. The multi-objective optimization system of claim 1, wherein: The parameterized modeling module integrates the CATIA three-dimensional modeling interface and the HYPERMESH finite element pre-processing function, and the AI training module supports MRMR algorithm feature sorting and multiple model accuracy comparison.