A multi-objective simulation optimization integrated method for automobile front bumper

Abstract

The application relates to the technical field of computer-aided engineering, and particularly discloses a multi-target simulation optimization integrated method for an automobile front bumper, which comprises the following steps: establishing a parameterized template, automatically preprocessing and generating a model, automatically solving and scheduling, automatically post-processing and quality checking, multi-target optimization, verification and rewriting. Through the above steps, each link of the optimization process is integrated into an automatic closed loop, the optimization efficiency and result reliability are significantly improved, and the research and development process is traceable.

Description

Technical Field

[0001] This invention relates to the field of computer-aided engineering technology, and in particular to a multi-objective simulation optimization integration method for automobile front bumpers. Background Technology

[0002] In modern product development, especially in the design of complex products such as automobiles, computer-aided engineering simulation analysis is an indispensable step. Design solutions often need to meet multiple conflicting performance objectives. For example, a car bumper needs to simultaneously meet the energy absorption requirements of low-speed collisions and the anti-dent stiffness requirements under static loading. In existing technologies, a distributed workflow is typically used for such multi-objective optimization problems. Specifically, design engineers, simulation engineers, and optimization engineers work independently in different software environments. For instance, simulation engineers manually modify the model's geometry, material properties, or component thickness in pre-processing software, and then export solution files suitable for different operating conditions. Next, the computational tasks are submitted to a high-performance computing cluster via command line or batch scripts. After the task is completed, the result files are manually downloaded, key performance indicators are extracted using post-processing tools, and the data is summarized using spreadsheets or similar tools. Finally, optimization engineers import this data into independent optimization software for analysis. This decentralized, manually driven process has significant drawbacks. First, it is inefficient, as a large number of experimental design samples require repeated manual preprocessing, task submission, and data organization, consuming a significant amount of manpower and time. Second, it suffers from poor consistency; in the flow of data across multiple operating conditions and versions, inconsistencies in model naming, physical units, and coordinate systems can easily arise, leading to data errors. Third, the process is opaque; the lack of a unified interface between the optimization algorithm and the simulation model results in slow result transmission, and simulation task failures are difficult to detect and trace automatically, requiring manual investigation and handling. Finally, it is difficult to close the loop; the candidate solutions obtained by the optimization algorithm are difficult to automatically write back to the original parametric model for high-fidelity simulation verification, making it difficult to form a closed loop between design and optimization, thus affecting the reliability of the final results. Summary of the Invention

[0003] This invention aims to solve the technical problems of existing simulation optimization processes being fragmented, lacking automation, having poor data consistency, being inefficient, and having difficulty in tracing the optimization process and results. It provides an integrated multi-objective simulation optimization method for automotive front bumpers, which can automatically integrate parametric modeling, multi-condition solution, performance index extraction, data quality control, and multi-objective optimization, thereby shortening the product optimization cycle, improving data quality, and ensuring the traceability of the R&D process.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solutions:

[0005] The present invention discloses a multi-objective simulation optimization integration method for automotive front bumpers. This method is based on a unified simulation process and data management platform and includes the following steps:

[0006] Establish a parametric template: Perform parametric processing on the computer-aided engineering (CAE) model of the product to be optimized, define design variables, and establish the mapping relationship between the design variables and the entities and attributes in the CAE model;

[0007] Automatic preprocessing and model generation: Based on a set of values ​​for the design variables, the CAE model is automatically updated via script to generate a working condition model for simulation calculation;

[0008] Automated solution scheduling: The operating condition model is submitted to high-performance computing resources for simulation calculation;

[0009] Automated post-processing and quality verification: Key performance indicators (KPIs) are automatically extracted from the simulation results, and the KPIs are verified to form valid samples.

[0010] Multi-objective optimization: A multi-objective optimization algorithm is used to perform optimization calculations based on the effective samples to obtain one or more candidate solutions, wherein each candidate solution includes a set of design variable values;

[0011] Verification and write-back: The candidate schemes are recalculated to confirm their performance indicators, and the design variable values ​​included in the confirmed candidate schemes are written back to the computer-aided engineering (CAE) model.

[0012] Furthermore, prior to the multi-objective optimization step, the following steps are also included:

[0013] Based on the effective samples, train the agent model;

[0014] The multi-objective optimization step specifically involves performing optimization calculations on the surrogate model to obtain candidate solutions that include design variable values ​​and key performance indicators predicted by the surrogate model.

[0015] Furthermore, after training the agent model, the method also includes:

[0016] Evaluate the accuracy of the proxy model; and,

[0017] If the accuracy does not reach the prediction threshold, adaptive point supplementation is performed to increase the number of samples, and the steps of automatic preprocessing and model generation, automatic solution scheduling, automatic postprocessing and quality verification, and training of the surrogate model are repeated to iteratively update the surrogate model.

[0018] Furthermore, the surrogate model includes at least one of the following: response surface model, radial basis function model, Kriging model, Gaussian process model, or neural network model.

[0019] Furthermore, in the automatic preprocessing and model generation step, after generating the working condition model, an automated model quality check is also performed, which includes at least one of mesh quality check, free edge check, or model penetration check.

[0020] Furthermore, the automated solution scheduling step also includes:

[0021] Monitor the execution status, logs, and exit codes of simulation calculation tasks, and automatically retry tasks that fail due to preset reasons.

[0022] Furthermore, the quality verification includes performing unit conversion on the simulation calculation results and checking at least one of the following: the integrity of the result file, the coverage of the simulation time steps, or the rationality of the energy balance.

[0023] Furthermore, the step of recalculating the candidate scheme in the verification and write-back step includes:

[0024] The recalculated key performance indicators are compared with the key performance indicators predicted by the surrogate model in the candidate scheme, and it is determined whether the error is within the preset tolerance range.

[0025] Furthermore, the multi-objective optimization algorithm includes:

[0026] At least one of the following: non-dominated sorting genetic algorithm NSGA-II, multi-objective particle swarm optimization algorithm MOPSO, or multi-objective differential evolution algorithm MODE.

[0027] Furthermore, the adaptive point compensation employs at least one of the following strategies:

[0028] Supplementary points based on maximum forecast uncertainty;

[0029] Point supplementation based on the maximum expected improvement;

[0030] Point completion based on probability improvement.

[0031] The multi-objective simulation optimization integration method for automobile front bumpers described in this invention has the following beneficial technical effects compared with the prior art:

[0032] 1. Improve efficiency and reduce costs. By integrating parametric modeling, multi-condition solving, post-processing, optimization and verification into an automated process, manual intervention and repetitive operations are significantly reduced, shortening the product development cycle. In particular, when combined with a proxy model, the number of high-cost simulation calculations can be greatly reduced, thus lowering R&D costs.

[0033] 2. Ensure data consistency and reliability. Under a unified platform and data model, the consistency of model, unit and key performance indicator definitions is ensured across multiple working conditions and versions. The built-in automated quality check and the final high-fidelity recalculation verification mechanism eliminate abnormal data and ensure the accuracy and reliability of the final optimization results.

[0034] 3. It enables process traceability and knowledge accumulation, fully recording and preserving the entire chain of correspondence from design variables, model versions, calculation tasks to simulation results and final decisions, achieving complete traceability of the R&D process, which facilitates problem identification, design reuse and knowledge management. Attached Figure Description

[0035] The above and other objects, features, and advantages of the invention will become clearer through a more detailed description of the preferred embodiments illustrated in the accompanying drawings. The same reference numerals denote the same parts throughout the drawings, and the drawings are not intentionally drawn to scale with actual dimensions; the focus is on illustrating the gist of the invention.

[0036] Figure 1 This is a flowchart of an embodiment of the present invention. Detailed Implementation

[0037] To facilitate understanding of the present invention, a more comprehensive description of the present invention will be given below with reference to the accompanying drawings.

[0038] It should be noted that when a component is considered to be "connected" to another component, it can be directly connected to and integrated with the other component, or there may be an intervening component present. The terms "mounted," "one end," "the other end," and similar expressions used in this document are for illustrative purposes only.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0040] This invention provides a specific implementation method for a multi-objective simulation optimization integration method for automotive front bumpers, see [link to implementation details]. Figure 1The multi-objective simulation optimization integration method for automotive front bumpers is implemented based on a unified simulation process and data management platform, and includes the following steps:

[0041] S10. Establish a parameterized template: Perform parameterization processing on the computer-aided engineering (CAE) model of the product to be optimized, define design variables, and establish the mapping relationship between the design variables and the entities and attributes in the CAE model. Specifically, the product to be optimized is a car front bumper. As a complex assembly, the front bumper is divided into variable and fixed areas. For example, the mounting points connecting to the car body are considered fixed areas, while the structural parameters of the front bumper's crossbeam and energy-absorbing box are considered variable areas. Based on this, design variables that significantly affect product performance are defined, such as the bumper crossbeam thickness, crossbeam material, stiffener location, energy-absorbing box wall thickness and dimensions, and connection stiffness. Next, a unique identifier is created for each design variable (e.g., "beam_thickness", "box_wall_thickness", "beam_material", etc.), and a mapping relationship is constructed between these identifiers and specific entity attributes in the model. For example, "beam_thickness" is mapped to the thickness field in the attribute card representing the crossbeam component in the pre-processing software; "beam_material" is mapped to the material ID in the material definition card. These mapping relationships are stored in a template file as the basis for subsequent automated operations.

[0042] In this embodiment, after defining the design variables, it is necessary to further define the optimization objectives and constraints. The optimization objectives are dual: objective one is to minimize the maximum displacement under the anti-dent stiffness analysis condition; objective two is to minimize the maximum intrusion under the low-speed collision condition. A constraint is also set: the total mass of the bumper system must not exceed a preset upper limit.

[0043] S20. Automatic Preprocessing and Model Generation: Based on a set of values ​​for the design variables, the CAE model is automatically updated via script to generate a working condition model for simulation calculations. This step, after generating the working condition model, also includes performing automated model quality checks, which include at least one of mesh quality checks, free edge checks, or model penetration checks. Mesh quality checks ensure that the Jacobian, aspect ratio, and other indicators of the updated elements are within acceptable ranges; free edge checks ensure that the model has no discontinuous holes; and model penetration checks ensure that there is no initial penetration or interference between components. If a serious problem is found during the check, the sample will be marked as "model generation failed" and the reason for the failure will be recorded; subsequent calculation steps will not be executed.

[0044] S30. Automated Solution Scheduling: The aforementioned work case model is submitted to high-performance computing resources for simulation calculation. Specifically, for each successfully generated work case model, it is submitted as a solution task to the high-performance computing cluster. Standardized simulation process nodes are established for each work case. When submitting a task, computing resources, including the number of cores, memory, and computation time limit, are intelligently allocated to the task based on the characteristics of the model (e.g., collision analysis requires a large number of cores and a long computation time) and the current state of the computing environment (e.g., the load of different computing queues), and task priorities are set.

[0045] In a preferred embodiment, the automated solution scheduling step further includes: monitoring the execution status, logs, and exit codes of simulation calculation tasks, and automatically retrying tasks that fail due to preset reasons. In this embodiment, the automatic retry mechanism is as follows: if a task fails due to temporary environmental issues (such as insufficient licenses or transient failure of computing nodes), the exit code and running logs will be checked. If the failure reason is preset to be a retryable type, the task will be automatically resubmitted, for example, the preset number of retries is three. If the task fails due to errors in the model itself (such as non-convergence of the solution), the task status will be marked as computational failure, an error log will be recorded, and the subsequent process of that sample will be terminated.

[0046] S40. Automated Post-processing and Quality Verification: Automatically extract key performance indicators (KPIs) from simulation results and perform quality verification on these KPIs to form valid samples. Quality verification includes unit conversion of simulation results and checking at least one of the following: completeness of the result file, coverage of simulation time steps, or rationality of energy balance. Since different solvers or models may use different unit systems, the post-processing script will uniformly convert all extracted key performance indicators to standard units according to preset rules to ensure data consistency. For example, for concave stiffness analysis, the script reads and extracts the maximum displacement value at a predefined location from the result file (e.g., .odb or .dat); for low-speed collision analysis, it extracts the maximum intrusion of the bumper beam and the total mass of the system from the result file (e.g., d3plot or binout). Quality verification of KPIs can be performed after obtaining the numerical values ​​of key performance indicators. This includes checking the integrity of the result files to ensure the solver terminates normally and generates all necessary result files; checking the coverage of simulation time steps to ensure the simulation calculation proceeds completely to the preset termination time; and checking the rationality of key physical quantities. For example, in collision analysis, checking the energy balance curve ensures total energy conservation and that the proportion of artificial energy (such as hourglass energy or contact energy) to internal energy is below a certain threshold (e.g., 5%). If the calculation result of any working condition of a sample fails the above quality verification, the sample will be marked as an anomalous sample, and the reason for the anomalousness (e.g., energy imbalance) will be recorded. Data from these anomalous samples will not be used for subsequent surrogate model training, thus ensuring the quality of data used for optimization.

[0047] S50. Multi-objective optimization: A multi-objective optimization algorithm is employed to perform optimization calculations based on the effective samples to obtain one or more candidate solutions, wherein each candidate solution includes a set of design variable values. Specifically, on the surrogate model that has passed accuracy verification, a multi-objective optimization algorithm is executed to perform optimization calculations to obtain candidate solutions containing design variable values ​​and key performance indicators predicted by the surrogate model. On the surrogate model that has passed accuracy verification, a multi-objective optimization algorithm is executed to perform optimization calculations. Since the computational cost of the surrogate model is extremely low (milliseconds), a large number of iterative searches can be performed. The multi-objective optimization algorithm includes at least one of the following: Non-dominated sorting genetic algorithm (NSGA-II), multi-objective particle swarm optimization algorithm (MOPSO), or multi-objective differential evolution algorithm (MODE). In this embodiment, the non-dominated sorting genetic algorithm (NSGA-II) is used for optimization. In this embodiment, the optimization algorithm aims to minimize the maximum displacement and maximum intrusion predicted by the surrogate model, while using the total mass predicted by the surrogate model as a constraint. After evolution, the algorithm converges and outputs a set of Pareto non-dominated solutions, which contains different design schemes that form a trade-off front between the two objectives. For example, some solutions perform well in concavity stiffness but are slightly worse at collision energy absorption, while others are the opposite, and some solutions are relatively balanced in both performance aspects.

[0048] S60. Verification and Write-back: The candidate schemes are recalculated to confirm their performance indicators, and the design variable values ​​included in the confirmed candidate schemes are written back to the computer-aided engineering (CAE) model. In this embodiment, the step of recalculating the candidate schemes includes comparing the recalculated key performance indicators with the key performance indicators predicted by the surrogate model in the candidate schemes, and determining whether the error is within the preset tolerance range. In this embodiment, based on the obtained Pareto front, three representative candidate schemes are selected for final verification according to engineering preferences: Scheme A (emphasizing stiffness, minimum displacement), Scheme B (emphasizing energy absorption, minimum intrusion), and Scheme C (balanced scheme). For each set of design variables of these three candidate schemes, a high-fidelity recalculation process is automatically triggered. That is, the design variables of these three schemes are used as input, and steps S20 to S40 are executed repeatedly, i.e., models are automatically generated for them, submitted to the high-performance computing cluster for real finite element calculations, and real performance indicators are automatically extracted. Subsequently, the calculated true key performance indicators are compared with the key performance indicators predicted by the surrogate model in step S50. For example, for scheme C, the surrogate model predicts a maximum displacement of 4.5 mm, while the high-fidelity recalculation result is 4.6 mm, with an error of (4.6-4.5) / 4.6 ≈ 2.2%. In this embodiment, the preset tolerance range is 5%. After comparison, the prediction errors of all three candidate schemes are less than 5%, and the recalculation results all meet the total mass constraint, which proves the effectiveness of the surrogate model and the reliability of the optimization results. Finally, the scheme C with the most balanced performance is adopted (for example, its design variables are: beam thickness 2.2 mm, energy-absorbing box wall thickness 2.0 mm, material A). The design variable values ​​of this scheme are automatically written back to the parameterized basic model library, and a new version of the computer-aided engineering model file is generated. At the same time, a complete optimization report is automatically generated, which includes the evolution history of the design variables, the Pareto front plot, the performance comparison table of candidate schemes, and the final decision record. At this point, the entire optimization process is in a closed loop, and all process data is properly archived, achieving complete traceability.

[0049] In a preferred embodiment, the optimization of the front bumper's anti-dent stiffness and low-speed collision performance employs a "small initial sample + adaptive point filling" sample generation and surrogate model construction strategy. Specifically, after step S10 and before step S20, multiple initial sample points are generated using the Latin hypercube sampling method, orthogonal design method, or optimal space filling method, with the Latin hypercube sampling method being preferred. Each initial sample represents a specific combination of design variable values, and a unique identifier is assigned to each initial sample. Invalid combinations that do not meet geometric feasibility or variable combination prohibition rules are pre-checked and filtered out.

[0050] In a preferred embodiment, after step S40 and before step S50, it is also necessary to train and evaluate the accuracy of the surrogate model: based on the effective samples, the surrogate model is trained; and the accuracy of the surrogate model is evaluated. If the accuracy does not reach the prediction threshold, adaptive point supplementation is performed to increase the number of samples, and the steps of automatic preprocessing and model generation, automatic solution scheduling, automatic post-processing and quality verification, and training the surrogate model are repeated to iteratively update the surrogate model. In this embodiment, the surrogate model includes at least one of the following: response surface model, radial basis function model, Kriging model, Gaussian process model, or neural network model. The Kriging model is preferred as the surrogate model because it usually exhibits high accuracy when dealing with small sample and nonlinear problems. Three independent Kriging models are established for two optimization objectives (maximum displacement and maximum intrusion) and one constraint (total mass). These models can quickly predict the corresponding response values ​​with design variables as input. Specifically, a preliminary Kriging surrogate model is trained using the aforementioned sample data. After training, the accuracy of the surrogate model is evaluated using cross-validation. If the evaluation results show that the model's global accuracy is insufficient—for example, if the root mean square error (RMSE) for predicting the maximum intrusion exceeds a preset threshold (e.g., 10%)—this indicates that directly using this model for optimization may result in a Pareto front that deviates significantly from the actual situation. Therefore, adaptive point supplementation is needed in the process judgment to improve model accuracy.

[0051] In a preferred embodiment, adaptive point addition employs at least one of the following strategies: point addition based on maximum prediction uncertainty; point addition based on maximum expected improvement; and point addition based on probability improvement. This embodiment preferably uses a hybrid strategy combining point addition based on maximum prediction uncertainty and point addition based on maximum expected improvement, which can simultaneously improve both the global and local accuracy of the model. Specifically, point addition based on maximum prediction uncertainty: In addition to providing predicted values, the Kriging model can also provide the confidence interval or prediction variance of the predicted value at a given point. Regions with larger prediction variances represent higher uncertainty in that region, typically corresponding to sparse sample points. This strategy aims to find several points (e.g., 5 points) with the largest prediction variances throughout the entire design space as new sample points, thereby helping to explore the entire design space and improve the global accuracy of the model. Point addition based on maximum expected improvement: This strategy aims to more accurately characterize the current Pareto optimal frontier. By calculating the expected value of each point in the design space becoming a new better solution, it determines where to add samples. In this embodiment, it can be simplified to adding several (e.g., 5) new sample points near the Pareto front predicted by the current surrogate model. This strategy helps to perform a fine search near the current optimal solution, improving the quality of optimization results. Through the above hybrid strategy, new sample points that need to be simulated are generated. In this embodiment, by introducing an adaptive point addition strategy, computational resources can be greatly saved and the development cycle shortened.

[0052] In a preferred embodiment, simulation and model updates are performed on the new sample points: the automated simulation calculation process of steps S20 to S40 is fully executed for the new sample points to obtain their true performance index values. After the calculation is completed, the new sample data (design variable-key performance index pairs) is added to the original multiple sample datasets to form an augmented dataset containing all samples (the samples generated above and the new samples generated in this step). Using the augmented dataset of all samples, the step of training the surrogate model is re-executed, that is, the Kriging surrogate model is retrained.

[0053] In a preferred embodiment, the updated model is further re-evaluated, that is, the updated Kriging model is cross-validated again. If the evaluation results show that the model's accuracy has been significantly improved, for example, the root mean square error (RMSE) has decreased to 4%, meeting the accuracy requirement of less than 5%, then the model accuracy is considered satisfactory. It should be noted that if the accuracy still does not meet the requirements, adaptive point addition and subsequent simulations and model updates can be repeated, forming an iterative cycle of "modeling-evaluation-point addition-simulation-remodeling" until the model accuracy meets the requirements.

[0054] The multi-objective simulation optimization integration method for automobile front bumpers in this embodiment has the following beneficial technical effects compared with the prior art:

[0055] 1. Improve efficiency and reduce costs. By integrating parametric modeling, multi-condition solving, post-processing, optimization and verification into an automated process, manual intervention and repetitive operations are significantly reduced, shortening the product development cycle. In particular, when combined with a proxy model, the number of high-cost simulation calculations can be greatly reduced, thus lowering R&D costs.

[0056] 2. Ensure data consistency and reliability. Under a unified platform and data model, the consistency of model, unit and key performance indicator definitions is ensured across multiple working conditions and versions. The built-in automated quality check and the final high-fidelity recalculation verification mechanism eliminate abnormal data and ensure the accuracy and reliability of the final optimization results.

[0057] 3. It enables process traceability and knowledge accumulation, fully recording and preserving the entire chain of correspondence from design variables, model versions, calculation tasks to simulation results and final decisions, achieving complete traceability of the R&D process, which facilitates problem identification, design reuse and knowledge management.

[0058] In this application, unless otherwise expressly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "on top of," and "over" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0059] In the description of this specification, the references to terms such as "preferred embodiment," "another embodiment," "other embodiment," or "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0060] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A multi-objective simulation optimization integration method for automobile front bumpers, characterized in that, The multi-objective simulation optimization integration method for automotive front bumpers is implemented based on a unified simulation process and data management platform, and includes the following steps: Establish a parametric template: Perform parametric processing on the computer-aided engineering (CAE) model of the product to be optimized, define design variables, and establish the mapping relationship between the design variables and the entities and attributes in the CAE model; Automatic preprocessing and model generation: Based on a set of values ​​for the design variables, the CAE model is automatically updated via script to generate a working condition model for simulation calculation; Automated solution scheduling: The operating condition model is submitted to high-performance computing resources for simulation calculation; Automated post-processing and quality verification: Key performance indicators (KPIs) are automatically extracted from the simulation results, and the KPIs are verified to form valid samples. Multi-objective optimization: A multi-objective optimization algorithm is used to perform optimization calculations based on the effective samples to obtain one or more candidate solutions, wherein each candidate solution includes a set of design variable values; Verification and write-back: The candidate schemes are recalculated to confirm their performance indicators, and the design variable values ​​included in the confirmed candidate schemes are written back to the computer-aided engineering (CAE) model.

2. The multi-objective simulation optimization integration method for automobile front bumpers according to claim 1, characterized in that, Prior to the multi-objective optimization step, the following is also included: Based on the effective samples, train the agent model; The multi-objective optimization step specifically involves performing optimization calculations on the surrogate model to obtain candidate solutions that include design variable values ​​and key performance indicators predicted by the surrogate model.

3. The multi-objective simulation optimization integration method for automobile front bumpers according to claim 2, characterized in that, After training the agent model, the following is also included: Evaluate the accuracy of the proxy model; and, If the accuracy does not reach the prediction threshold, adaptive point supplementation is performed to increase the number of samples, and the steps of automatic preprocessing and model generation, automatic solution scheduling, automatic postprocessing and quality verification, and training of the surrogate model are repeated to iteratively update the surrogate model.

4. The multi-objective simulation optimization integration method for automobile front bumpers according to claim 2 or 3, characterized in that, The surrogate model includes at least one of the following: response surface model, radial basis function model, Kriging model, Gaussian process model, or neural network model.

5. The multi-objective simulation optimization integration method for automobile front bumpers according to any one of claims 1-3, characterized in that, In the automatic preprocessing and model generation step, after generating the working condition model, an automated model quality check is also performed. The model quality check includes at least one of mesh quality check, free edge check, or model penetration check.

6. The multi-objective simulation optimization integration method for automobile front bumpers according to any one of claims 1-3, characterized in that, The automated solution scheduling step also includes: Monitor the execution status, logs, and exit codes of simulation calculation tasks, and automatically retry tasks that fail due to preset reasons.

7. The multi-objective simulation optimization integration method for automobile front bumpers according to any one of claims 1-3, characterized in that, The quality verification includes performing unit conversion on the simulation calculation results and checking at least one of the following: the integrity of the result file, the coverage of the simulation time steps, or the rationality of the energy balance.

8. The multi-objective simulation optimization integration method for automobile front bumpers according to claim 2, characterized in that, The verification and write-back step includes the following steps for recalculating the candidate scheme: The recalculated key performance indicators are compared with the key performance indicators predicted by the surrogate model in the candidate scheme, and it is determined whether the error is within the preset tolerance range.

9. The multi-objective simulation optimization integration method for automobile front bumpers according to claim 1, characterized in that, The multi-objective optimization algorithm includes: At least one of the following: non-dominated sorting genetic algorithm NSGA-II, multi-objective particle swarm optimization algorithm MOPSO, or multi-objective differential evolution algorithm MODE.

10. The multi-objective simulation optimization integration method for automobile front bumpers according to claim 3, characterized in that, The adaptive point compensation employs at least one of the following strategies: Supplementary points based on maximum forecast uncertainty; Point supplementation based on the maximum expected improvement; Point completion based on probability improvement.

Images