Method and device for determining a construction strategy of a body-in-white in a vehicle

Abstract

The application provides a method and device for determining a construction strategy of a body-in-white in a vehicle, and relates to the technical field of vehicles. The method comprises the following steps: constructing a structure variable set based on a structure dimension and a performance dimension of a body-in-white to be constructed; determining at least one first constraint condition of the body-in-white according to the structure variable set, wherein the first constraint condition is used to constrain the engineering feasibility of the structure of the body-in-white; determining at least one candidate construction strategy of the body-in-white based on the first constraint condition, a target function of the body-in-white and a second constraint condition, wherein the second constraint condition is used to constrain a vehicle with the body-in-white constructed by the candidate construction strategy; and determining a target construction strategy for constructing the body-in-white based on the at least one candidate construction strategy of the body-in-white. The application solves the technical problem of low accuracy of the determination of the construction strategy of the body-in-white.

Description

Technical Field

[0001] This application relates to the field of vehicle control technology, and more specifically, to a method and apparatus for determining the construction strategy of the body-in-white in a vehicle. Background Technology

[0002] As the core load-bearing structure of a vehicle, the rigidity, vibration characteristics, and lightweight level of the body-in-white directly determine the safety, comfort, and energy consumption of the entire vehicle.

[0003] In traditional vehicle development processes, body-in-white structural optimization relies heavily on engineering experience for parameter adjustments, resulting in a high degree of reliance on experience. Consequently, there is a technical problem of low accuracy in determining body-in-white construction strategies.

[0004] There is currently no good solution to the above problems. Summary of the Invention

[0005] This application provides a method and apparatus for determining the construction strategy of the body-in-white in a vehicle, so as to at least solve the technical problem of low accuracy in determining the body-in-white construction strategy.

[0006] According to one aspect of the embodiments of this application, a method for determining the construction strategy of a vehicle body-in-white is provided. The method may include: constructing a set of structural variables based on the structural and performance dimensions of the body-in-white to be constructed; determining at least one first constraint condition for the body-in-white based on the set of structural variables, wherein the first constraint condition is used to constrain the engineering feasibility of the structure of the body-in-white; determining at least one candidate construction strategy for the body-in-white based on the first constraint condition, the objective function of the body-in-white, and a second constraint condition, wherein the objective function is used to indicate whether the performance of the structure of the body-in-white constructed using the candidate construction strategy meets the construction requirements of the body-in-white, and the second constraint condition is used to constrain vehicles possessing body-in-whites constructed using the candidate construction strategy; and determining a target construction strategy for constructing the body-in-white based on the at least one candidate construction strategy for the body-in-white.

[0007] Optionally, based on the first constraint, the objective function of the body-in-white, and the second constraint, at least one candidate construction strategy for the body-in-white is determined, including: evaluating the first constraint, the objective function, and the second constraint using an evolutionary model to obtain the target feasible region of the body-in-white, wherein the evolutionary model is used to determine the candidate structural parameters of the body-in-white, and the candidate structural parameters in the target feasible region satisfy the first constraint, the objective function, and the second constraint; using a surrogate model to determine the candidate performance parameters corresponding to the candidate structural parameters in the target feasible region; and using the candidate structural parameters and the candidate performance parameters to construct a candidate construction strategy, wherein the candidate construction strategy is used to represent the rules for constructing a body-in-white that satisfies the candidate structural parameters and possesses the candidate performance parameters.

[0008] Optionally, the first constraint, objective function, and second constraint are evaluated using an evolutionary model to obtain the target feasible region of the body-in-white, including: using an evolutionary model to determine the initial feasible region of the body-in-white, wherein the candidate structural parameters in the initial feasible region satisfy the objective function and the second constraint; and using an evolutionary model to determine the target feasible region from the initial feasible region, wherein the candidate structural parameters in the target feasible region satisfy at least one first constraint.

[0009] Optionally, a surrogate model is used to determine the candidate performance parameters corresponding to the candidate structural parameters in the target feasible region, including: inputting the candidate structural parameters in the target feasible region into the surrogate model, and using the surrogate model to determine the candidate performance parameters that have a mapping relationship with the candidate structural parameters.

[0010] Optionally, the method further includes: determining structural parameter samples from the set of structural variables, and obtaining performance parameter samples corresponding to the structural parameter samples; initializing the surrogate model using the structural parameter samples and performance parameter samples, wherein the initialized surrogate model is used to determine the candidate performance parameters corresponding to the candidate structural parameters in the target feasible domain.

[0011] Optionally, determining the structural parameter sample from the set of structural variables includes: uniformly dividing the set of structural variables into a target number of structural variable subsets; randomly selecting structural variables from the target number of structural variable subsets; randomly sorting the target number of structural variables to form an initial structural parameter sample; and determining the structural parameter sample from the initial structural parameter sample, wherein the initial structural parameter sample includes structural parameter samples and invalid structural parameter samples.

[0012] Optionally, obtaining the performance parameter samples corresponding to the structural parameter samples includes: performing finite element simulation on the structural parameter samples to obtain the performance parameter samples; and / or, initializing the proxy model using the structural parameter samples and performance parameter samples includes: inputting the structural parameter samples and performance parameter samples into the proxy model, generating a mapping relationship between the structural parameter samples and performance parameter samples using the proxy model; and determining the proxy model with successfully constructed mapping relationship as the initialized proxy model.

[0013] Optionally, based on at least one candidate construction strategy for the body-in-white, determining a target construction strategy for constructing the body-in-white includes: identifying at least one first construction strategy from the candidate construction strategies, wherein the reliability of the first construction strategy is less than the reliability of the candidate construction strategies other than the first construction strategy; generating at least one second construction strategy using the first construction strategy, wherein the reliability of the second construction strategy is greater than the reliability of the first construction strategy; and determining the target construction strategy from the candidate construction strategies that include the first construction strategy, the second construction strategy, and the candidate construction strategies other than the first construction strategy.

[0014] Optionally, the candidate construction strategy includes candidate structural parameters and candidate performance parameters. Identifying at least one first construction strategy from the candidate construction strategies includes: determining a distance information set corresponding to at least one candidate construction strategy, wherein the distance information in the distance information set is used to represent the similarity between the candidate structural parameters and the structural parameter samples used to initialize the surrogate model, and the number of distance information in the distance information set corresponds to the number of structural parameter samples; and determining the candidate construction strategy as the first construction strategy in response to the minimum distance information in the distance information set exceeding a distance threshold.

[0015] Optionally, the first construction strategy includes candidate structural parameters and candidate performance parameters. Using the first construction strategy, at least one second construction strategy is generated, including: performing finite element simulation on the candidate structural parameters to obtain the actual performance parameters corresponding to the candidate structural parameters; determining the actual structural parameters that the body-in-white must meet to have the actual performance parameters; and using the actual performance parameters and the actual structural parameters to construct the second construction strategy.

[0016] Optionally, determining the target construction strategy from the candidate construction strategies other than the first construction strategy, which include the first construction strategy, the second construction strategy, and the candidate construction strategies, includes: updating the initialized proxy model in response to the number of second construction strategies reaching the target number, to obtain an updated proxy model; and using the updated proxy model to determine the target construction strategy from the first construction strategy, the second construction strategy, and the second construction strategy.

[0017] Optionally, the initialized proxy model includes an input end and an output end. Updating the initialized proxy model to obtain an updated proxy model includes: fixing the front layer weight parameters of the input end and updating the remaining layer weight parameters of the output end; responding to the updated remaining layer weight parameters and the original remaining layer weight parameters, satisfying the preset weight change constraints, using the updated remaining layer weight parameters to update the initialized proxy model to obtain the updated proxy model.

[0018] Optionally, the set of structural variables includes a multidimensional design parameter vector. Based on the structural and performance dimensions of the body-in-white to be constructed, the set of structural variables is constructed, including: generating first-level structural information of the body-in-white to be constructed based on the vehicle's macro-architecture parameters, and generating second-level structural information of the body-in-white to be constructed based on the finite element baseline model of the body-in-white, wherein the first level is higher than the second level; establishing a topological coupling relationship between the first-level structural information and the second-level structural information; and integrating the first-level structural information and the second-level structural information into a multidimensional design parameter vector according to the topological coupling relationship.

[0019] Optionally, the method further includes constructing a finite element baseline model of the body-in-white; using the finite element baseline model, determining at least one of the following structural information of the body-in-white: load-bearing structural components, connection relationships, and material properties, wherein the connection relationships are connection rules between at least two load-bearing structural components; and the material properties are used to represent the materials used to generate the load-bearing structural components.

[0020] According to another aspect of the embodiments of this application, a device for determining the construction strategy of a vehicle body-in-white is also provided. The device may include: a construction module for constructing a set of structural variables based on the structural and performance dimensions of the body-in-white to be constructed; a first determination module for determining at least one first constraint condition of the body-in-white based on the set of structural variables, wherein the first constraint condition is used to constrain the engineering feasibility of the structure of the body-in-white; a second determination module for determining at least one candidate construction strategy of the body-in-white based on the first constraint condition, the objective function of the body-in-white, and the second constraint condition, wherein the objective function is used to indicate whether the performance of the structure of the body-in-white constructed using the candidate construction strategy meets the construction requirements of the body-in-white, and the second constraint condition is used to constrain vehicles that possess a body-in-white constructed using the candidate construction strategy; and a third determination module for determining a target construction strategy for constructing the body-in-white based on the at least one candidate construction strategy of the body-in-white.

[0021] According to another aspect of the embodiments of this application, a vehicle is also provided, including: a body-in-white constructed using a target construction strategy determined by the methods in various embodiments of this application.

[0022] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0023] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0024] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0025] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the methods in various embodiments of this application.

[0026] According to another aspect of the embodiments of this application, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of this application.

[0027] In this embodiment, a method for determining the construction strategy of the body-in-white in a vehicle is adopted. This method constructs a set of structural variables based on the structural and performance dimensions of the body-in-white to be constructed. Based on the set of structural variables, at least one first constraint condition is determined for the body-in-white, where the first constraint condition constrains the engineering feasibility of the body-in-white structure. Based on the first constraint condition, the objective function of the body-in-white, and the second constraint condition, at least one candidate construction strategy for the body-in-white is determined. The objective function indicates whether the performance of the body-in-white structure constructed using the candidate construction strategy meets the construction requirements of the body-in-white, and the second constraint condition constrains vehicles capable of constructing body-in-whites using the candidate construction strategy. Based on the at least one candidate construction strategy for the body-in-white, a target construction strategy for constructing the body-in-white is determined. This achieves the goal of systematically quantifying the first and second constraints and collaboratively optimizing the structural performance and vehicle compatibility. This results in the technical effect of achieving performance optimization while meeting hard constraints such as quality and assembly in the final determined target construction strategy. Furthermore, it solves the technical problem in the prior art where body-in-white structure optimization mainly relies on engineering experience for parameter adjustment, leading to low accuracy in determining the body-in-white construction strategy. Attached Figure Description

[0028] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0029] Figure 1 This is a hardware structure block diagram of a computing terminal determined according to an optional construction strategy for the body-in-white in a vehicle, based on an embodiment of this application.

[0030] Figure 2 This is a flowchart of a method for determining the construction strategy of the body-in-white in a vehicle according to an embodiment of this application;

[0031] Figure 3 This is a flowchart of a method for collaborative positive optimization of multiple performance of a body-in-white based on a heterogeneous parameter system and deep progressive learning, according to an embodiment of this application.

[0032] Figure 4 This is a schematic diagram of a dual-scale uncertainty determination mechanism based on the coupling of Euclidean distance and distribution density according to an embodiment of this application;

[0033] Figure 5 This is a schematic diagram of a device for determining the construction strategy of the body-in-white in a vehicle according to an embodiment of this application;

[0034] Figure 6 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0035] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0036] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0037] As the core load-bearing structure of the entire vehicle, the stiffness and vibration characteristics of the body-in-white directly determine the vehicle's safety and ride comfort. In the traditional automotive development process, body-in-white structural optimization mainly relies on engineers' experience to adjust parameters, supplemented by time-consuming finite element analysis for successive verification. This approach suffers from long development cycles and strong dependence on experience. In recent years, the combination of data-driven modeling and intelligent optimization algorithms has provided new ideas for structural optimization and has gradually been introduced into the field of body-in-white design.

[0038] While the prior art related to this application has promoted the intelligentization of body-in-white design to some extent, the following unresolved problems remain when applied to practical high-dimensional and complex engineering projects. First, the parameter system construction in existing technologies lacks a unified mechanism, mainly relying on empirical definitions or focusing on a single thickness parameter. Although cross-sectional parameters and connection methods are mentioned, they are mostly discrete enumerations, lacking a systematic construction logic for a structured parameter system. This leads to design variables in the vast variable space of body-in-white design often being either oversimplified or excessive and disordered, causing the curse of dimensionality, resulting in deviations between optimization results and engineering realities, and limiting the potential for performance improvement.

[0039] Secondly, existing technologies suffer from the problem of blind sample supplementation. In high-dimensional design spaces, the predictive reliability of static models built solely based on initial samples decreases significantly as they move away from the sample space. Existing solutions lack a mechanism for quantitatively determining uncertain regions of prediction and for supplementing high-precision samples in a targeted manner, resulting in wasted computational resources in high-confidence regions.

[0040] Secondly, historical knowledge degradation occurs during model updates. Existing static models and ensemble weight update techniques lack mechanisms for preserving historical knowledge when introducing new samples. Traditional full retraining or simple fine-tuning leads to a decline in the model's predictive ability for features in the original design space, thus affecting optimization convergence stability. Finally, there is a misalignment between existing optimization paradigms and engineering practices. Existing techniques often prioritize minimizing mass, which can easily lead to sacrificing stiffness and modal performance in the pursuit of lightweighting. Engineering practices, however, typically explore performance limits under weight constraints rather than simply pursuing weight reduction. Therefore, a multi-performance collaborative forward optimization method for body-in-white that can address these issues is urgently needed.

[0041] According to an embodiment of this application, a method embodiment for determining the construction strategy of the body-in-white in a vehicle is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, steps shown or described may be executed in a different order than that shown here.

[0042] Figure 1 This is a hardware structure block diagram of a computing terminal for an optional method of determining the construction strategy of the body-in-white in a vehicle, according to an embodiment of this application. Figure 1As shown, a computing terminal (e.g., a computer terminal, a mobile smart terminal, a vehicle terminal, or a cloud computing virtual terminal) may include: one or more processors 102, a memory 104, and a transmission device 106 configured to implement communication functions. Each processor 102 may include, but is not limited to, a processing component such as a microprocessor (MCU) or a field programmable gate array (FPGA).

[0043] The aforementioned computing terminal may further include: a display device 110, an input / output device 108, and a Universal Serial Bus (USB) port (this USB port can be used as one of the ports of the computer bus). Figure 1 (not shown in the image), network interface ( Figure 1 (not shown in the image), power supply ( Figure 1 (not shown in the image) and camera ( Figure 1 (Not shown in the image). Those skilled in the art will understand that... Figure 1 The structure of the computing terminal shown is for illustrative purposes only and does not impose strict limitations on the structure of the computing terminal described above. For example, the computing terminal may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0044] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits may be embodied, in whole or in part, as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the vehicle terminal (or mobile device).

[0045] The memory 104 may be configured to store software programs and modules for application software, such as the program instructions / data storage device corresponding to the method for determining the construction strategy of the vehicle body-in-white in this embodiment of the application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned method for determining the construction strategy of the vehicle body-in-white. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to a computing terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0046] The transmission device 106 is configured to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computing terminal. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, configured to communicate with the Internet wirelessly.

[0047] Under the above operating environment, the embodiments of this application provide the following: Figure 2 The method for determining the construction strategy of the body-in-white in the vehicle shown is as follows. Figure 2 This is a flowchart of a method for determining the construction strategy of the body-in-white in a vehicle according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following implementation steps S201 to S204.

[0048] Step S201: Based on the structural and performance dimensions of the body-in-white to be constructed, construct a set of structural variables.

[0049] In this embodiment of the application, the construction strategy determination system constructs a set of structural variables based on the structural and performance dimensions of the body-in-white to be constructed.

[0050] The strategy determination system is the implementing entity of this application.

[0051] Body-in-white refers to the pure body structure of a car after stamping, welding, and other processes have been completed, but before painting and final assembly (such as installing the engine, chassis, interior, and glass). The body-in-white is the load-bearing skeleton of the entire vehicle, directly determining its passive safety and handling stability.

[0052] The structural dimension is a parametric description of the finite element baseline model of the body-in-white, including static parameters such as the geometric features, cross-sectional properties, and connection methods of the physical structure of the body-in-white.

[0053] The performance dimension refers to the benchmark values ​​of the body-in-white structure performance, which represent the mechanical response indicators exhibited by the body-in-white under specific working conditions, such as stiffness, modal frequency, and mass.

[0054] The set of structural variables refers to the set of multidimensional parameters composed of the relevant parameters of the above-mentioned structural and performance dimensions, which are constructed through a heterogeneous logic structured multidimensional parameter system construction mechanism and are used for subsequent optimization calculations.

[0055] It can be seen that the strategy determination system selects and determines the specific parameter items to participate in the subsequent optimization calculation from the overall design space of the body-in-white, forming a combination set that includes structural and performance dimensions, which serves as the basic input data for subsequent constraint definition and strategy generation.

[0056] For example, the strategy determination system constructs a system for determining the thickness of the front windshield pillar (A-pillar) and front door pillar (B-pillar) for a certain vehicle body-in-white as structural dimension variables, and extracts the target bending stiffness and first-order torsional frequency as performance dimension variables. The above parameters are then summarized to form a set of structural variables containing multiple variables.

[0057] Therefore, by constructing a strategy, a systematic set of variables containing structural and performance dimensions is determined, providing a clear and complete data foundation for subsequent optimization calculations under multiple constraints.

[0058] Step S202: Based on the set of structural variables, determine at least one first constraint condition for the body-in-white, wherein the first constraint condition is used to constrain the engineering feasibility of the structure of the body-in-white.

[0059] In this embodiment of the application, the construction strategy determination system determines at least one first constraint condition for the body-in-white based on the set of structural variables.

[0060] The first constraint is used to constrain the engineering feasibility of the body-in-white structure. The first constraint refers to the rules or boundaries established based on engineering experience, manufacturing process limitations, and assembly specifications to limit the range of values ​​of structural variables. The first constraint is a soft constraint, such as engineering rules like topological coupling constraints and process adaptation constraints.

[0061] Engineering feasibility refers to the state in which the body-in-white design scheme is practically feasible in terms of manufacturing, assembly, and safety standards.

[0062] It can be seen that the construction strategy determines the system by identifying key parameters from the set of structural variables, and setting the range of values ​​or interrelationships that these parameters need to meet based on the process specifications and assembly requirements in actual engineering, so as to serve as the preliminary filtering conditions for screening feasible design schemes.

[0063] For example, the strategy determination system sets the thickness ratio of the two structural variables, the A-pillar and the sill beam, to be between 0.8 and 1.2 based on the stamping process requirements, in order to ensure a smooth transition in stiffness and compliance with process specifications.

[0064] Therefore, by setting engineering rules, the reasonable range of design variables is initially limited, ensuring that the subsequent optimization process is carried out within the feasible space that conforms to the basic manufacturing process.

[0065] Step S203: Based on the first constraint, the objective function of the body-in-white, and the second constraint, determine at least one candidate construction strategy for the body-in-white. The objective function is used to indicate whether the performance of the structure of the body-in-white constructed using the candidate construction strategy meets the construction requirements of the body-in-white. The second constraint is used to constrain vehicles that have a body-in-white constructed using the candidate construction strategy.

[0066] In this embodiment of the application, the construction strategy determination system determines at least one candidate construction strategy for the body-in-white based on the first constraint, the objective function of the body-in-white, and the second constraint.

[0067] The objective function represents whether the performance of the body-in-white structure constructed using the candidate construction strategy meets the construction requirements of the body-in-white. Furthermore, the objective function is an optimization objective, referring to a mathematical index that quantifies the degree of matching between the structural performance (such as stiffness and modal characteristics) of the body-in-white under the candidate construction strategy and the preset construction requirements.

[0068] The second constraint is used to constrain vehicles with body-in-whites constructed through candidate construction strategies. This second constraint is a hard constraint on vehicle mass, imposed at the vehicle level (e.g., overall dimensions, vehicle mass, crash safety), to ensure that the body-in-white generated by the candidate construction strategy can be adapted and installed on the target vehicle platform. Examples of the second constraint include that the total mass of the body-in-white does not exceed the upper limit of vehicle mass, that the body-in-white's external dimensions do not interfere with the chassis layout, and that the locations of key mounting points conform to the vehicle coordinate system requirements.

[0069] Candidate construction strategies refer to specific design schemes or optimization paths formed by parametrically adjusting or combining the set of structural variables.

[0070] It can be seen that the system for determining the construction strategy comprehensively considers engineering feasibility constraints (first constraint), performance compliance (objective function), and vehicle compatibility limitations (second constraint). Through optimization algorithms or evaluation mechanisms, it selects or generates a series of preliminary body-in-white design schemes that meet both performance requirements and vehicle platform compatibility requirements, i.e., candidate construction strategies.

[0071] For example, after generating a candidate construction strategy, the construction strategy determination system first verifies whether the candidate construction strategy meets the first constraint condition of the thickness ratio of adjacent plates, then calculates whether the bending stiffness under the candidate construction strategy reaches the objective function, and finally checks whether the total mass of the vehicle body corresponding to the candidate construction strategy exceeds the second constraint condition. If all conditions are met, the candidate construction strategy is retained.

[0072] Therefore, through the synergistic evaluation of multi-dimensional constraints and performance targets, it is ensured that the generated candidate construction strategies not only meet structural performance requirements but also have engineering adaptability at the vehicle level.

[0073] Step S204: Based on at least one candidate construction strategy for the body-in-white, determine the target construction strategy for constructing the body-in-white.

[0074] In this embodiment of the application, the construction strategy determination system determines the target construction strategy for constructing the body-in-white based on at least one candidate construction strategy of the body-in-white.

[0075] The target construction strategy refers to the preferred single or combination of preferred solutions ultimately selected from the set of candidate construction strategies to guide the actual manufacturing or further detailed design of the body-in-white.

[0076] As can be seen, the construction strategy determination system selects and locks a final target construction strategy for the actual construction of the body-in-white from the multiple candidate construction strategies generated in the aforementioned steps, based on preset decision criteria or optimization logic.

[0077] For example, after obtaining three candidate construction strategies, A, B, and C, the comprehensive score of each strategy is calculated through a multi-objective decision-making algorithm based on the dual emphasis on lightweighting and torsional stiffness at the current stage of the project, and finally the preferred strategy A is selected as the target construction strategy.

[0078] Therefore, by selecting the final target from multiple feasible options, the scientific and rational nature of the body-in-white construction scheme was ensured, providing a clear basis for subsequent production and manufacturing.

[0079] In summary, this application constructs a set of structural variables based on the structural and performance dimensions of the body-in-white to be constructed. This construction strategy systematically establishes a variable set encompassing both structural and performance dimensions, providing a clear and complete data foundation for subsequent optimization calculations under multiple constraints. Based on the set of structural variables, at least one first constraint is determined for the body-in-white, which constrains the engineering feasibility of the body-in-white's structure. This pre-defined engineering rule initially limits the reasonable range of design variables, ensuring that subsequent optimization processes are conducted within a feasible space that conforms to basic manufacturing processes. Based on the first constraint, the objective function of the body-in-white, and the second constraint, at least one... A candidate construction strategy is proposed, wherein the objective function represents whether the performance of the body-in-white structure constructed using the candidate construction strategy meets the construction requirements of the body-in-white, and the second constraint condition is used to constrain vehicles that possess body-in-whites constructed using the candidate construction strategy. Thus, through the synergistic evaluation of multi-dimensional constraints and performance objectives, it is ensured that the generated candidate construction strategy meets structural performance requirements while possessing engineering adaptability at the vehicle level. Based on at least one candidate construction strategy for the body-in-white, a target construction strategy for constructing the body-in-white is determined. Thus, by selecting the final target from multiple feasible solutions, the scientificity and rationality of the body-in-white construction scheme are ensured, providing a clear execution basis for subsequent production and manufacturing.

[0080] In this embodiment, a method for determining the construction strategy of the vehicle body-in-white is adopted. This method constructs a set of structural variables based on the structural and performance dimensions of the body-in-white to be constructed. Based on the set of structural variables, at least one first constraint condition is determined for the body-in-white, where the first constraint condition constrains the engineering feasibility of the body-in-white structure. Based on the first constraint condition, the objective function of the body-in-white, and the second constraint condition, at least one candidate construction strategy for the body-in-white is determined. The objective function indicates whether the performance of the body-in-white structure constructed using the candidate construction strategy meets the construction requirements of the body-in-white, and the second constraint condition constrains vehicles capable of constructing body-in-whites using the candidate construction strategy. Based on the at least one candidate construction strategy for the body-in-white, a target construction strategy for constructing the body-in-white is determined. This achieves the goal of systematically quantifying multi-dimensional engineering constraints and collaboratively optimizing structural performance and vehicle compatibility. This results in a high-quality design scheme that meets all system constraints during the forward design process of the body-in-white, improving design efficiency and overall performance. Furthermore, it solves the technical problem in existing technologies where body-in-white structure optimization mainly relies on engineering experience for parameter adjustment, leading to low accuracy in determining the body-in-white construction strategy.

[0081] Optionally, based on the first constraint, the objective function of the body-in-white, and the second constraint, at least one candidate construction strategy for the body-in-white is determined, including:

[0082] An evolutionary model is used to evaluate the first constraint, the objective function, and the second constraint to obtain the target feasible region of the body-in-white. The evolutionary model is used to determine the candidate structural parameters of the body-in-white. The candidate structural parameters in the target feasible region satisfy the first constraint, the objective function, and the second constraint.

[0083] Using a surrogate model, candidate performance parameters corresponding to candidate structural parameters in the target feasible domain are determined;

[0084] Candidate construction strategies are constructed using candidate structural parameters and candidate performance parameters. The candidate construction strategy is used to represent the rules for constructing a body-in-white that satisfies the candidate structural parameters and has the candidate performance parameters.

[0085] In this embodiment of the application, firstly, the strategy determination system uses an evolutionary model to evaluate the first constraint, the objective function, and the second constraint to obtain the target feasible region of the body-in-white.

[0086] The evolutionary model is used to determine candidate structural parameters for the body-in-white. The evolutionary model refers to a multi-objective evolutionary algorithm based on swarm intelligence and genetic algorithms, which searches for optimal solutions by simulating the natural selection process.

[0087] The evolutionary model is used to determine candidate structural parameters of the body-in-white under the premise of satisfying the hard constraints of body mass and topological coupling constraints. Optionally, the evolutionary model can use the Nondominated Sorting Genetic Algorithm II (NSGA-II) as the core of multi-objective optimization. This evolutionary model searches by simulating natural selection and genetic mechanisms. Key implementation details include: setting the population size to 100 individuals, the maximum number of generations to 100, until the convergence condition is met or the maximum number of iterations is reached; setting the crossover probability to 0.8~0.9 and the mutation probability to 0.01~0.1 to balance global exploration capability and local development accuracy; using Pareto dominance relation to perform nondominated sorting of multiple objectives such as mass, bending stiffness, torsional stiffness, and modal frequency, and maintaining the diversity of the solution set through crowding distance, thereby selecting the Pareto optimal solution set as candidate structural parameters.

[0088] Candidate structural parameters in the target feasible region satisfy the first constraint, the objective function, and the second constraint. The target feasible region refers to the region in the multidimensional parameter space that comprises all candidate structural parameters that simultaneously satisfy the first constraint, the objective function, and the second constraint.

[0089] It can be seen that the construction strategy determination system uses the iterative search and evaluation mechanism of the evolutionary model to select parameter combinations that simultaneously meet all preset constraints and performance requirements (objective functions), thereby defining the parameter space range that meets engineering requirements, i.e., the target feasible region.

[0090] For example, during the iteration process, the evolutionary model eliminates individuals whose total mass exceeds the upper limit (violating the second constraint) or whose thickness ratio of adjacent plates exceeds the process range (violating the first constraint), and finally retains those parameter combinations that meet the process requirements, vehicle size limits, and stiffness performance (satisfying the objective function), forming the target feasible region.

[0091] Therefore, evolutionary search can efficiently locate the parameter space that satisfies multiple constraints, providing a high-quality candidate solution set for subsequent performance evaluation.

[0092] Then, the strategy determination system uses a proxy model to determine the candidate performance parameters corresponding to the candidate structural parameters in the target feasible domain.

[0093] Among them, the surrogate model refers to a mathematical model trained on historical data (such as a deep feedforward neural network surrogate model) used to approximate a computationally expensive physical simulation model.

[0094] Candidate performance parameters refer to the body-in-white performance indicators (such as bending stiffness, torsional stiffness, first-order bending frequency, and first-order torsional frequency) that are predicted by the surrogate model and correspond to the candidate structural parameters.

[0095] It can be seen that the strategy determination system uses a trained surrogate model to quickly predict the performance of candidate structural parameters within the target feasible domain, thereby obtaining the corresponding candidate performance parameters and avoiding direct, time-consuming simulations.

[0096] For example, the strategy determination system inputs a candidate structural parameter vector from the target feasible domain into a pre-trained neural network surrogate model. After forward propagation calculation, the surrogate model outputs the predicted values ​​of the bending stiffness and torsional stiffness corresponding to the candidate structural parameter, which are the candidate performance parameters.

[0097] Therefore, by using the proxy model to achieve rapid performance evaluation, the computational cost is significantly reduced and the efficiency of candidate policy generation is improved.

[0098] Finally, the strategy determination system uses candidate structural parameters and candidate performance parameters to construct candidate construction strategies.

[0099] Among them, the candidate construction strategy is used to represent the rules for constructing a body-in-white that satisfies the candidate structural parameters and has candidate performance parameters.

[0100] It can be seen that the construction strategy determination system integrates structural parameters and performance parameters to form a complete candidate construction strategy, thereby clearly expressing a body-in-white construction scheme with specific performance characteristics.

[0101] For example, the construction strategy determination system binds the candidate structural parameter "A-pillar thickness 0.8mm" with the corresponding candidate performance parameter "bending stiffness 8000N / mm" to generate a candidate construction strategy entry. This candidate construction strategy entry represents the scheme rule for constructing a body-in-white with an A-pillar thickness of 0.8mm to obtain the stiffness of 8000N / mm.

[0102] This establishes a clear mapping between parameters and performance, forming a complete set of evaluable and comparable candidate solutions, thus providing a basis for the final strategy selection.

[0103] Optionally, an evolutionary model is used to evaluate the first constraint, the objective function, and the second constraint to obtain the target feasible region of the body-in-white, including:

[0104] An evolutionary model is used to determine the initial feasible region of the body-in-white, where the candidate structural parameters in the initial feasible region satisfy the objective function and the second constraint.

[0105] Using an evolutionary model, a target feasible region is determined from the initial feasible region, wherein the candidate structural parameters in the target feasible region satisfy at least one first constraint condition.

[0106] In this embodiment of the application, firstly, the strategy determination system uses an evolutionary model to determine the initial feasible domain of the body-in-white.

[0107] In this context, the candidate structural parameters in the initial feasible region satisfy the objective function and the second constraint. Specifically, the initial feasible region refers to the preliminary parameter space selected in the early stage of the evolutionary model search based solely on performance target requirements (objective function) and basic vehicle adaptation constraints (second constraint), without considering detailed process details.

[0108] Candidate structural parameters refer to specific combinations of geometric or material variables in the design of the body-in-white.

[0109] It can be seen that the construction strategy determination system utilizes an evolutionary model to initially screen candidate structural parameters that meet the basic performance and size requirements based on the objective function and the second constraint, thus forming an initial feasible region.

[0110] For example, during the iteration, the evolutionary model eliminates individuals with a total mass exceeding 180kg or a bending stiffness below 8000N / mm, retaining the combination of structural parameters that both meet performance indicators and vehicle size constraints, thus forming the initial feasible domain.

[0111] Thus, the initial screening narrowed the search space, providing a high-quality candidate base for subsequent refined process constraint checks.

[0112] Then, the strategy determination system uses an evolutionary model to determine the target feasible region from the initial feasible region.

[0113] Among them, the candidate structural parameters in the target feasible region satisfy at least one first constraint condition.

[0114] It can be seen that the construction strategy determination system uses an evolutionary model to evaluate the process feasibility of candidate structural parameters in the initial feasible region, and eliminates invalid solutions that violate assembly continuity, stiffness gradient mutation, process mismatch or body mass exceeding constraints, thereby obtaining the final target feasible region.

[0115] For example, the strategy determination system examines each candidate parameter from the initial feasible domain. If the thickness ratio of the A-column to the sill beam of a certain parameter is 0.7 (violating the soft process constraint), then this candidate parameter is eliminated, and only the structural parameters that satisfy all preset process rules are retained to form the target feasible domain.

[0116] Therefore, by introducing process constraints, the parameters within the final feasible domain not only meet performance standards but also possess operability for actual manufacturing.

[0117] Optionally, using a surrogate model, candidate performance parameters corresponding to candidate structural parameters in the target feasible region are determined, including:

[0118] The candidate structural parameters in the target feasible region are input into the surrogate model, and the surrogate model is used to determine the candidate performance parameters that have a mapping relationship with the candidate structural parameters.

[0119] In this embodiment of the application, the construction strategy determination system inputs the candidate structural parameters in the target feasible domain into the surrogate model, and uses the surrogate model to determine the candidate performance parameters that have a mapping relationship with the candidate structural parameters.

[0120] It can be seen that the construction strategy determination system takes the candidate structural parameters in the target feasible domain as input data and passes them into the surrogate model. Through the mapping calculation inside the surrogate model, it directly outputs the corresponding structural performance prediction value, that is, the candidate performance parameter.

[0121] For example, the strategy determination system selects candidate structural parameter vectors from the target feasible region, inputs the candidate structural parameter vectors into a trained neural network, and after forward propagation calculation, outputs the predicted values ​​of bending stiffness and torsional stiffness corresponding to the candidate structural parameters, which are the candidate performance parameters.

[0122] Therefore, by using a proxy model to achieve rapid performance prediction, the time-consuming physical simulation calculation is avoided, and the generation efficiency of candidate strategies is significantly improved.

[0123] Optionally, the method further includes:

[0124] From the set of structural variables, determine the structural parameter samples and obtain the performance parameter samples corresponding to the structural parameter samples;

[0125] The surrogate model is initialized using structural parameter samples and performance parameter samples. The initialized surrogate model is used to determine the candidate performance parameters corresponding to the candidate structural parameters in the target feasible domain.

[0126] In this embodiment of the application, firstly, the strategy determination system determines structural parameter samples from the set of structural variables, and obtains performance parameter samples corresponding to the structural parameter samples.

[0127] Among them, the structural parameter sample, or initial sample point, refers to a number of specific combinations of body-in-white structural parameters selected from the design variable space. For example, the structural parameter sample generates 200 sets of parameter combinations, each of which includes variables such as thickness and cross-sectional dimensions.

[0128] Performance parameter samples refer to the body-in-white structural performance data that correspond one-to-one with the structural parameter samples, obtained through high-fidelity physical simulation (such as finite element analysis). Performance parameter samples include, for example, numerical results such as bending stiffness, torsional stiffness, and modal frequencies obtained after performing finite element simulations on the aforementioned 200 sets of parameter combinations.

[0129] It can be seen that the construction strategy determines the system to build the original dataset used to train the agent model, that is, to generate structural parameter samples by sampling and to obtain the corresponding performance parameter samples through high-cost simulation, thus forming parameter-performance mapping pairs.

[0130] For example, the construction strategy determination system randomly or according to rules selects 100 sets of structural parameters from the 36-dimensional design space, inputs them into the finite element software for simulation, and records the stiffness and frequency results corresponding to each set of parameters, thereby obtaining a structural parameter-performance parameter dataset containing 100 samples.

[0131] This provides basic data support for training the surrogate model and establishes an initial mapping relationship between input parameters and output performance.

[0132] Then, the strategy determination system uses structural parameter samples and performance parameter samples to initialize the surrogate model, whereby the initialized surrogate model is used to determine the candidate performance parameters corresponding to the candidate structural parameters in the target feasible domain.

[0133] Initialization refers to the process of training or fitting the internal parameters (such as neural network weights, radial basis function coefficients, etc.) of the surrogate model using structural parameter samples and performance parameter samples. For example, it includes inputting sample data into the network and adjusting the weights through the backpropagation algorithm to minimize the error between the network output and the real performance samples.

[0134] It can be seen that the strategy determination system uses the acquired sample data to train the surrogate model, enabling the surrogate model to learn the mapping rules between structural parameters and performance parameters, thereby forming an initial surrogate model that can perform fast performance prediction.

[0135] For example, the 100 sets of structural and performance parameters obtained in the aforementioned steps are input into the neural network of the initial structure. Through multiple rounds of iterative training, the network weights are adjusted until the prediction error of the surrogate model on the validation set is lower than a set threshold, thus completing the initialization of the surrogate model.

[0136] Thus, by training the agent model to enable performance prediction, a tool is provided for rapidly evaluating candidate strategies within the target feasible domain.

[0137] Optionally, a sample of structural parameters is determined from the set of structural variables, including:

[0138] The set of structural variables is uniformly divided into a target number of subsets of structural variables;

[0139] Randomly select structural variables from the target number of structural variables subsets;

[0140] Randomly sort the target number of structural variables to form an initial structural parameter sample;

[0141] From the initial structural parameter sample, a structural parameter sample is determined, wherein the initial structural parameter sample includes structural parameter samples and invalid structural parameter samples.

[0142] In this embodiment of the application, firstly, the strategy is constructed to determine that the system uniformly divides the set of structural variables into a target number of subsets of structural variables.

[0143] It can be seen that the strategy determination system divides the large set of structural variables into several subsets of equal size, laying the foundation for subsequent grouped random sampling and ensuring the balance of variable selection.

[0144] For example, the 36 design variables are divided into 6 subsets according to region, such as the front panel region, A / B column region, floor region, etc. Each subset contains 6 specific variable parameters within that region.

[0145] Therefore, by uniformly dividing the data, it is ensured that variables of different categories or regions have an equal probability of being selected during the sampling process, thus avoiding bias in variable selection.

[0146] Secondly, the construction strategy determines that the system randomly selects structural variables from a subset of structural variables of the target number.

[0147] It can be seen that the construction strategy determines the system by randomly selecting structural variables from each uniformly divided subset, and by combining structural variables from different subsets, constructing a parameter combination basis with diversity.

[0148] For example, the strategy determination system randomly selects variables V1 and V2 in the first subset, randomly selects variable V3 in the second subset, and so on, selecting several variables from each subset to form a preliminary batch of candidate parameter combinations.

[0149] Therefore, by randomly sampling across subsets, the distribution diversity of the initial samples in the parameter space is increased, which helps to cover a wider design area.

[0150] Then, the strategy determination system randomly sorts the target number of structural variables to form an initial structural parameter sample.

[0151] It can be seen that the construction strategy determination system shuffles the randomly selected structural variables to generate a preliminary set of parameter samples. This initial set of structural parameter samples represents a specific combination of structural parameters.

[0152] For example, the extracted set of structural variables {V1,V2,V3,V4,V5,V6} is randomly rearranged to obtain the sequence {V4,V1,V6,V2,V5,V3}. The randomly rearranged sequence {V4,V1,V6,V2,V5,V3} is then mapped to specific parameter values ​​to form an initial structural parameter sample.

[0153] Thus, random sorting eliminates potential correlations or biases that may arise from the order of variable selection, enhancing the randomness and disorder of the sample.

[0154] Finally, the strategy determination system determines the structural parameter samples from the initial structural parameter samples.

[0155] The initial structural parameter sample includes structural parameter samples and invalid structural parameter samples. Invalid structural parameter samples refer to combinations of structural parameters in the initial structural parameter sample that do not meet basic physical rules, process limitations, or logical constraints (such as negative thickness, dimensions exceeding limits, etc.).

[0156] It can be seen that the construction strategy determination system performs validity checks on the initial structural parameter samples, eliminates invalid samples that do not conform to basic engineering logic or constraints, and thus obtains the final structural parameter samples for subsequent processing.

[0157] For example, the construction strategy determines that the system checks each parameter in the initial sample. If a cross-sectional dimension is found to be less than 0 or greater than the maximum allowable value, the structural parameter sample is determined to be an invalid structural parameter sample and discarded. Only samples in which all parameters are within a reasonable range are retained as structural parameter samples.

[0158] Therefore, by filtering out invalid samples, the quality and usability of subsequent data are improved, and calculation errors or model training biases caused by invalid parameters are avoided.

[0159] Optionally, obtain the performance parameter samples corresponding to the structural parameter samples, including:

[0160] Finite element simulation was performed on the structural parameter samples to obtain the performance parameter samples;

[0161] And / or, using structural parameter samples and performance parameter samples, initialize the surrogate model, including:

[0162] Input the structural parameter samples and performance parameter samples into the surrogate model, and use the surrogate model to generate the mapping relationship between the structural parameter samples and the performance parameter samples;

[0163] The proxy model that successfully establishes the mapping relationship is identified as the initialized proxy model.

[0164] In this embodiment of the application, the construction strategy determination system can perform finite element simulation on structural parameter samples to obtain performance parameter samples.

[0165] Finite element simulation refers to the use of computer numerical simulation methods to establish a finite element model of the body-in-white based on structural parameter samples, and to calculate the mechanical response of the finite element model of the body-in-white under specific loads and boundary conditions.

[0166] It can be seen that the strategy determination system obtains the real structural performance data corresponding to the structural parameter samples through high-fidelity finite element simulation calculations, thereby forming label data for training the surrogate model.

[0167] For example, the strategy determination system inputs the selected structural parameter samples into finite element software, establishes a finite element model of the body-in-white, applies bending or torsional loads, and extracts the displacement or stress cloud diagrams of each node after solving, thereby calculating the overall bending stiffness and torsional stiffness as performance parameter samples.

[0168] This provides high-precision performance ground truth, ensuring the accuracy of the training data for the surrogate model.

[0169] And / or, the strategy determination system initializes the surrogate model using structural parameter samples and performance parameter samples. The initial dataset can be input into the deep feedforward neural network surrogate model, and the surrogate model can be trained using this initial data sample set. That is, the surrogate model can initially learn the mapping relationship between structural parameters and structural performance.

[0170] Specifically, this may include: the strategy determination system inputs structural parameter samples and performance parameter samples into the surrogate model, and uses the surrogate model to generate a mapping relationship between the structural parameter samples and the performance parameter samples. Then, the strategy determination system identifies the surrogate model with the successfully constructed mapping relationship as the initialized surrogate model.

[0171] It can be seen that the strategy determination system uses structural parameter samples and performance parameter samples to train the surrogate model. By minimizing the error between the predicted and actual values, the surrogate model learns the inherent mapping relationship between the two. Then, after confirming that the surrogate model training is complete, the surrogate model's state is marked as "initialized," enabling the surrogate model to perform subsequent performance prediction tasks.

[0172] For example, the strategy determination system inputs structural parameter samples into a deep feedforward neural network, calculates the difference between the network output and the performance parameter samples, updates the network weights through backpropagation, and trains iteratively for multiple rounds until the network can accurately map the input structural parameters to the expected performance parameters. When the surrogate model's prediction accuracy on the independent test set meets engineering requirements (e.g., relative error less than 5%), training stops, and the surrogate model is saved as the initialized surrogate model, ready to receive candidate structural parameters from the target feasible domain of the evolutionary model.

[0173] Thus, a rapid prediction model between parameters and performance was established, replacing time-consuming finite element simulation, and providing a basic tool for rapid evaluation of candidate strategies within the target feasible region.

[0174] Optionally, based on at least one candidate construction strategy for the body-in-white, a target construction strategy for constructing the body-in-white is determined, including:

[0175] Identify at least one first construction strategy from the candidate construction strategies, wherein the reliability of the first construction strategy is less than the reliability of the candidate construction strategies other than the first construction strategy.

[0176] Using a first construction strategy, at least one second construction strategy is generated, wherein the reliability of the second construction strategy is greater than that of the first construction strategy;

[0177] The target construction strategy is determined from the candidate construction strategies, which include the first construction strategy, the second construction strategy, and the candidate construction strategies other than the first construction strategy.

[0178] In this embodiment of the application, firstly, the construction strategy determination system identifies at least one first construction strategy from the candidate construction strategies.

[0179] The reliability of the first construction strategy is lower than that of all other candidate construction strategies. The first construction strategy refers to the option with the lowest reliability or confidence level in its performance prediction among the candidate construction strategies.

[0180] As can be seen, the construction strategy determination system evaluates the predictive reliability of all candidate construction strategies and identifies the schemes with lower reliability among all candidate construction strategies as those that need further verification.

[0181] For example, the construction strategy determination system calculates the prediction confidence interval of each candidate construction strategy in the surrogate model and finds that the confidence interval of scheme X is wider, indicating that the prediction uncertainty of scheme X is higher. Therefore, scheme X is identified as the first construction strategy.

[0182] This accurately identified unreliable schemes in the prediction results, providing a clear target for subsequent high-precision verification.

[0183] Then, the build strategy determination system uses the first build strategy to generate at least one second build strategy.

[0184] The reliability of the second construction strategy is greater than that of the first construction strategy. The second construction strategy refers to the new samples generated based on the first construction strategy through perturbation, mutation, or local search.

[0185] It can be seen that the construction strategy determines the system to generate new alternatives with higher reliability around the first construction strategy with low reliability, in order to explore a better or more reliable design space.

[0186] For example, the construction strategy determination system takes the first construction strategy scheme X as the center, generates several perturbation schemes (such as schemes Y1, Y2, Y3) in the parameter neighborhood of the first construction strategy scheme X, evaluates the reliability of these new perturbation schemes, and selects the scheme with a higher reliability than scheme X as the second construction strategy.

[0187] Thus, by generating highly reliable alternatives, the candidate solution set is enriched and the overall solution quality is improved.

[0188] Finally, the construction strategy determination system determines the target construction strategy from the candidate construction strategies, which include the first construction strategy, the second construction strategy, and the candidate construction strategies other than the first construction strategy.

[0189] It can be seen that the construction strategy determines the system to integrate the original high-reliability scheme, the newly generated high-reliability scheme, and the original low-reliability scheme, forming a larger decision space, and ultimately selecting the target construction strategy from it.

[0190] For example, schemes BZ, Y1-Y3 and X are combined into a list, sorted according to a predetermined multi-objective evaluation criterion (such as stiffness / weight ratio), and the scheme ranked first is selected as the objective construction strategy.

[0191] Thus, by expanding the candidate pool and re-optimizing, the final target construction strategy was ensured to achieve a balance between performance and reliability.

[0192] Optionally, the candidate construction strategies include candidate structural parameters and candidate performance parameters, and at least one first construction strategy is identified from the candidate construction strategies, including:

[0193] Determine at least one distance information set corresponding to a candidate construction strategy, wherein the distance information in the distance information set is used to represent the similarity between the candidate structural parameters and the structural parameter samples used to initialize the surrogate model, and the number of distance information in the distance information set corresponds to the number of structural parameter samples;

[0194] If the minimum distance information in the distance information set exceeds the distance threshold, the candidate construction strategy is determined as the first construction strategy.

[0195] Among them, candidate structural parameters refer to the combination of variables generated by the evolutionary model or sampling strategy during the optimization process, which represent the specific design state of the body-in-white. They typically include plate thickness, cross-sectional geometry, material properties, or connection process parameters, and are used as input to the surrogate model to predict the structural performance of the candidate structural parameters.

[0196] Candidate performance parameters refer to quantitative indicators that reflect the mechanical behavior or physical characteristics of the body-in-white, obtained by quickly predicting them through a surrogate model based on candidate structural parameters. These indicators include bending stiffness, torsional stiffness, first-order modal frequency, or mass, and are used to evaluate the performance of design schemes.

[0197] In this embodiment of the application, firstly, the construction strategy determination system determines the distance information set corresponding to at least one candidate construction strategy.

[0198] In this distance information set, distance information represents the similarity between candidate structural parameters and the structural parameter samples used to initialize the surrogate model. The number of distance information items in the distance information set corresponds to the number of structural parameter samples. The distance information set refers to the set of distance values ​​calculated between a candidate construction strategy and all historical structural parameter samples used to train the surrogate model, for example, Euclidean distance.

[0199] The process of acquiring the distance information set may include, for example, the following steps: first, parameter normalization to eliminate the influence of different variables on distance calculation; then, direct range normalization (Min-Max Normalization) for continuous variables; and finally, one-hot encoding and normalization for discrete variables to form a normalized variable vector. Then, distance calculations are performed to compute the candidate solutions generated from the optimization. Normalized Euclidean distance with all samples in the current sample database The calculation formula is: .

[0200] Structural parameter samples refer to historical data points that have been verified by finite element simulation and are used for the initial training of the surrogate model. They are all samples in the current sample database.

[0201] It can be seen that the construction strategy determination system calculates the distance between the structural parameters of the current candidate construction strategy and all historical samples in the training dataset, thereby constructing a distance information set that reflects the distribution and similarity of the candidate construction strategy in the parameter space.

[0202] For example, the construction strategy determination system selects a candidate construction strategy with parameters such as [A column thickness 0.8, B column thickness 0.9...]. The parameters of the candidate construction strategy are used to calculate the Euclidean distance between each of the 200 sets of historical parameters in the training set, generating an array containing 200 distance values. Each value in the array represents the degree of deviation between the candidate construction strategy and the corresponding historical sample.

[0203] Therefore, by quantifying the distance between candidate strategies and historical data, an intuitive geometric basis is provided for assessing the reliability of predictions.

[0204] Then, when the minimum distance information in the distance information set exceeds the distance threshold, the construction strategy determination system determines the candidate construction strategy as the first construction strategy.

[0205] The distance threshold refers to a pre-defined critical value used to determine whether the parameter space is outside the coverage area of ​​known data. For example, in constructing a strategy determination system, an adaptive distance threshold is set based on an adaptive threshold for distribution density. The algorithm is dynamically adjusted based on the sample distribution density in the parameter space. The K-nearest neighbor method is used to calculate the candidate scheme's performance across all samples. The mean of the nearest normalized distances Further calculations of all one sample The average of the nearest normalized distances The threshold is calculated using the following formula. . Specifically, , , .in, To design the dimensions of variables, Based on the threshold, The adjustment coefficient, ranging from 0.01 to 0.1, is used to adjust the sensitivity of the uncertainty judgment threshold and can be determined through small-scale preliminary experiments. Based on the above formula, a larger threshold is used in densely sampled areas to avoid redundant calculations, while a smaller threshold is used in sparsely sampled areas to encourage exploration of unknown regions.

[0206] As can be seen, the construction strategy determination system filters out candidate construction strategies that are far from known training data and have high prediction uncertainty by comparing the minimum neighborhood distance of candidate strategies with a preset threshold. The candidate construction strategies with high uncertainty are marked as the first construction strategy, and it is determined that the first construction strategy is in the prediction uncertainty region of the surrogate model.

[0207] For example, the construction strategy determination system checks the distance information set of candidate strategy A and finds that its minimum distance is 0.35. Assuming that the preset distance threshold is 0.3, since 0.35 is greater than 0.3, it determines that the candidate construction strategy is in the prediction uncertainty region, so it is identified as the first construction strategy, and triggers finite element calculation to add new samples to the database.

[0208] Therefore, by using simple threshold comparison, low-reliability design areas that require special attention can be quickly and effectively identified.

[0209] Optionally, the first construction strategy includes candidate structural parameters and candidate performance parameters. Using the first construction strategy, at least one second construction strategy is generated, including:

[0210] Finite element simulations were performed on the candidate structural parameters to obtain the actual performance parameters corresponding to the candidate structural parameters.

[0211] Determine the actual structural parameters that a body-in-white must meet to achieve actual performance parameters;

[0212] A second construction strategy is constructed using actual performance parameters and actual structural parameters.

[0213] In this embodiment of the application, firstly, the strategy determination system is constructed to perform finite element simulation on the candidate structural parameters to obtain the actual performance parameters corresponding to the candidate structural parameters.

[0214] Finite element simulation refers to the analytical method of using numerical simulation to perform high-precision calculations of the mechanical response of the body-in-white structure under specific working conditions.

[0215] Actual performance parameters refer to the quantitative results that reflect the true mechanical properties of the body-in-white (such as stiffness and modal characteristics) obtained directly through finite element simulation. Compared with the predicted values ​​of the surrogate model, actual performance parameters have higher accuracy.

[0216] It can be seen that the construction strategy determination system uses high-precision physical simulation to obtain the true performance of candidate construction strategies identified as having low reliability, so as to correct the possible biases in the surrogate model prediction.

[0217] For example, for a design scheme that is determined to be a first construction strategy, the design scheme of the first construction strategy includes specific plate thickness parameters. The finite element solver is called to perform simulation calculations on the design scheme of the first construction strategy, and the torsional stiffness and modal frequency values ​​corresponding to the design scheme of the first construction strategy are output. These values ​​are the actual performance parameters.

[0218] Therefore, obtaining real data through high-precision simulation provides reliable factual basis for subsequent model correction and the generation of better strategies.

[0219] Then, the strategy determination system determines the actual structural parameters that the body-in-white must meet to achieve the actual performance parameters.

[0220] The actual structural parameters refer to the set of variables, such as specific geometric dimensions and material properties, that need to be adjusted or confirmed for the body-in-white structure in order to achieve or approximate the aforementioned actual performance parameters.

[0221] It can be seen that the construction strategy determines the logical connection between the system and the performance results to trace back or confirm the required structural features, and clarifies the specific structural design elements that can support the current high performance.

[0222] For example, the strategy determination system is based on the torsional stiffness value obtained from simulation, and analyzes which main load-bearing structures contribute to the torsional stiffness value, thereby determining the set of key structural parameters that support the performance level, such as the dimensions corresponding to the moment of inertia of the key longitudinal beam.

[0223] This clarifies the structural roots behind high performance and provides specific structural guidance for building new strategies with high reliability.

[0224] Finally, the construction strategy determines the system to construct a second construction strategy using actual performance parameters and actual structural parameters.

[0225] It can be seen that the strategy determination system integrates high-precision simulation data with corresponding structural design parameters to form a validated, high-quality design strategy.

[0226] For example, the construction strategy determination system combines the precise stiffness value obtained from the simulation with the corresponding adjusted plate thickness parameters, records it as a new, reliable design rule, and marks it as the second construction strategy.

[0227] Therefore, by integrating high-precision data to generate high-quality strategies, the data quality and decision reliability of subsequent optimization processes are improved.

[0228] Optionally, the target build strategy is determined from candidate build strategies other than the first build strategy, which includes the first build strategy, the second build strategy, and candidate build strategies, including:

[0229] In response to the second construction strategy reaching the target number, the initialized proxy model is updated to obtain the updated proxy model;

[0230] Using the updated proxy model, the target construction strategy is determined from the first construction strategy, the second construction strategy, and the third construction strategy.

[0231] In this embodiment of the application, firstly, when the number of second construction strategies reaches the target number, the construction strategy determination system updates the initialized proxy model to obtain the updated proxy model.

[0232] The target number refers to the minimum cumulative number of second construction strategies preset to trigger model updates, ensuring the statistical significance of the updated data. For example, the target number is set to 50, meaning that an update is triggered when 50 high-quality samples of the second construction strategy are collected.

[0233] The initialized surrogate model refers to a neural network or other predictive model that has been trained on a small number of samples in the initial stage and has not yet absorbed new validation data.

[0234] The updated surrogate model refers to a surrogate model that has been trained and iterated with new data and whose weight parameters have been adjusted, thereby achieving higher prediction accuracy.

[0235] It can be seen that after accumulating a sufficient amount of high-quality validation data, the strategy determination system uses this data to retrain or fine-tune the existing prediction model in order to improve the accuracy of the surrogate model's prediction of future designs.

[0236] For example, when the system detects that the number of second construction strategies that have been generated has reached a preset 50, it calls an update command to input the data of these 50 new strategies into the current agent model, adjusts the model parameters through the backpropagation algorithm, and generates a new agent model with better performance.

[0237] Therefore, by regularly introducing high-quality data to update the surrogate model, the prediction accuracy and generalization ability of the surrogate model are effectively improved.

[0238] Then, the strategy determination system uses the updated agent model to determine the target strategy from the first, second, and third strategy construction strategies.

[0239] As can be seen, the strategy determination system uses an optimized and improved proxy model to comprehensively evaluate and screen all candidate strategies (including low-reliability and high-reliability strategies) to determine the final preferred solution.

[0240] For example, the build strategy determination system applies the updated proxy model to evaluate all first and second build strategies, and selects the solution that meets the design requirements and has better performance as the final target build strategy based on the predicted performance indicators and constraints.

[0241] Therefore, by using a more accurate proxy model for final decision-making, the performance and reliability of the selected target construction strategy can be ensured to meet the actual engineering requirements.

[0242] Optionally, the initialized proxy model includes an input and an output. Updating the initialized proxy model yields an updated proxy model, which includes:

[0243] The weight parameters of the first layer at the fixed input are updated, and the weight parameters of the remaining layers at the output are updated.

[0244] In response to the updated and unupdated remaining layer weight parameters, satisfying the preset weight change constraints, the initialized proxy model is updated using the updated remaining layer weight parameters to obtain the updated proxy model.

[0245] The input end refers to the network part of the surrogate model, which consists of the input layer and the hidden layer near the input end. It is mainly used to extract common basic features from the original structural parameters. The weight parameters of the input end remain fixed during the model update process to retain the general understanding of the overall data distribution.

[0246] The output end refers to the network part in the surrogate model that consists of the remaining hidden layers and the output layer. It is mainly used to map the extracted features into specific performance indicators. The weight parameters of the output end are locally fine-tuned according to the new samples during the model update process to adapt to the new performance mapping relationship.

[0247] In this embodiment, the construction strategy determines the front-layer weight parameters at the fixed input end of the system and updates the remaining layer weight parameters at the output end.

[0248] Among them, the front layer weight parameters at the input end refer to the network layer parameters in the deep neural network that are close to the input layer and used to extract basic general features.

[0249] The remaining layer weight parameters refer to the network layer parameters in a deep neural network that are close to the output layer and are used to map specific performance outputs.

[0250] Updating to fine-tuning training refers to the process of adjusting the network weight values ​​based on the error of new samples using the backpropagation algorithm.

[0251] It can be seen that the system constructs a strategy to retain the general feature extraction capability by freezing the input layer parameters and only adjusting the output layer parameters to adapt to new data, thereby achieving local fine-tuning of the surrogate model.

[0252] For example, when the strategy determination system trains the model using new finite element simulation data, it prohibits any changes to the weights of the first and second layers, and only allows the weights of the third and fourth layers to be calculated and adjusted based on the error of the new data.

[0253] Thus, while absorbing new knowledge, old knowledge is protected, preventing the surrogate model from losing its predictive ability for the original data distribution.

[0254] When the updated remaining layer weight parameters and the original remaining layer weight parameters satisfy the preset weight change constraints, the strategy determination system uses the updated remaining layer weight parameters to update the initialized proxy model, thus obtaining the updated proxy model.

[0255] Among them, the preset weight change constraint refers to the mathematical condition or regularization term that limits the magnitude of weight changes, used to prevent drastic weight fluctuations. For example, the preset weight change constraint can be a regularization term introduced into the loss function. Limit the range of weight changes to prevent the degradation of historical knowledge. ,in: , These are the weight parameters before and after the network update. The regularization coefficient, ranging from 0.01 to 0.1, is used to balance the absorption of new sample knowledge with the retention of historical knowledge. It can be determined by monitoring the change in prediction error of the surrogate model on the original sample before and after the update.

[0256] It can be seen that the strategy determination system checks whether the magnitude of the agent model parameter update is within a controllable range. If it meets the constraints, the new weights are formally applied, and the agent model update is completed.

[0257] For example, the strategy determination system calculates the difference between the new third-layer weights and the old third-layer weights. If the penalty term of the loss function introduced by the difference between the new third-layer weights and the old third-layer weights is less than a set value, the update is confirmed to be valid, and the new weights are filled into the proxy model to generate the updated proxy model.

[0258] This further ensures the stability of the proxy model updates and avoids the risk of drastic fluctuations in model performance due to individual abnormal data.

[0259] Optionally, the set of structural variables includes a multidimensional design parameter vector. Based on the structural and performance dimensions of the body-in-white to be constructed, the set of structural variables is constructed, including:

[0260] Based on the vehicle's macroscopic architecture parameters, the first-level structural information of the body-in-white to be constructed is generated, and based on the finite element baseline model of the body-in-white, the second-level structural information of the body-in-white to be constructed is generated, wherein the first level is higher than the second level.

[0261] Establish the topological coupling relationship between the first-level structural information and the second-level structural information;

[0262] Based on the topological coupling relationship, the structural information of the first level and the structural information of the second level are integrated into a multi-dimensional design parameter vector.

[0263] Among them, the multidimensional design parameter vector is a structured multidimensional parameter system, which refers to a numerical array or matrix composed of all design variables involved in the optimization in the body-in-white structure arranged in a specific logical order, and is used as a unified input for the surrogate model or a search space representation for the optimization algorithm.

[0264] In this embodiment of the application, firstly, the strategy determination system generates the first-level structural information of the body-in-white to be constructed based on the macroscopic architecture parameters of the vehicle, and generates the second-level structural information of the body-in-white to be constructed based on the finite element reference model of the body-in-white.

[0265] Among them, the macro-architectural parameters of a vehicle refer to the key global parameters that determine the basic size and layout of the entire vehicle, such as wheelbase, track width, body length, width and height, and front and rear overhang lengths.

[0266] The first level of structural information refers to the high-level structural description that reflects the overall geometric contour and layout of the body-in-white, directly determined by macroscopic architectural parameters. For example, in the overall vehicle geometry dimension, this includes determining the length, width, height, wheelbase, front and rear track widths, front and rear overhangs, and vehicle mass. In the connection process dimension, it includes determining the type of silicone sealant as one of the design variables. The macroscopic architectural parameters satisfy the constraints of overall vehicle space layout and structural continuity.

[0267] The finite element baseline model refers to a numerical simulation model of the body-in-white that includes nodes, elements, material properties, and connection relationships, established based on key region response driving parameters and detailed engineering design.

[0268] The second level of structural information refers to the low-level structural details extracted from the finite element baseline model, reflecting the geometric characteristics of specific components. Examples include thickness parameters for areas such as the A-pillar, B-pillar, rear windshield pillar (C-Pillar, or C-pillar for short), sill beam, longitudinal beam, floor, front bulkhead, rear bulkhead, and roof crossbeam, as well as the cross-sectional geometric parameters of critical paths such as the rear C-loop / C-frame (C-loop for short), rear longitudinal beam, and door frame. The first level is higher than the second level.

[0269] It can be seen that the construction strategy determination system extracts structural features from two levels: the overall layout detail design of the vehicle and the detailed design of specific components, forming a basic data structure with hierarchical relationships.

[0270] For example, the strategy determination system first reads the overall vehicle layout data to determine the outline dimensions of the body-in-white as the first level of information, and at the same time reads the cross-sectional and thickness data of each key beam and column from the finite element model as the second level of information.

[0271] This enables the hierarchical extraction of structural information from macro to micro levels, laying the foundation for establishing logical connections between levels.

[0272] Then, the construction strategy determines the topological coupling relationship between the first-level structural information and the second-level structural information of the system.

[0273] The topological coupling relationship refers to the constraints between the global structural information of the first level and the local structural information of the second level, including geometric coordination, force transmission continuity, and process adaptation. Specifically, based on structural connection relationships and engineering specifications, parameter value constraints are established in the dimensions of geometric dimension coordination, force transmission path continuity, and process adaptation. This includes geometric coordination constraints for the overall vehicle layout, force transmission continuity constraints for adjacent load-bearing structures, and thickness-connection process adaptation constraints.

[0274] It can be seen that the construction strategy determines the system to identify and define the mutual constraints and association rules between global architecture parameters and local structural parameters.

[0275] For example, the rule is set that if the overall length of the vehicle in the first level changes, the length of the corresponding front longitudinal beam and the position of the rear panel in the second level must be adjusted proportionally or by a fixed offset to maintain structural continuity.

[0276] Therefore, by establishing coupling rules between levels, the logical consistency and engineering feasibility of the structural design are ensured.

[0277] Finally, the construction strategy determines that the system integrates the structural information of the first level and the structural information of the second level into a multi-dimensional design parameter vector based on the topological coupling relationship.

[0278] It can be seen that the construction strategy determination system integrates multi-source structural information with hierarchical coupling into a unified standard input format.

[0279] For example, the strategy determination system places the overall vehicle length, width and height parameters at the beginning and the thickness parameters of each panel after topology rule verification at the end, combining them into a complete vector containing all design variables.

[0280] This completed the standardized integration of heterogeneous structural data, forming a unified multidimensional design parameter vector that can be used for subsequent optimization algorithm processing.

[0281] Optionally, the method further includes:

[0282] Construct a finite element baseline model of the body-in-white;

[0283] Using a finite element baseline model, at least one of the following structural information of the body-in-white is determined: load-bearing structural components, connection relationships, and material properties, wherein the connection relationships are the connection rules between at least two load-bearing structural components; and the material properties are used to represent the materials used to generate the load-bearing structural components.

[0284] In this embodiment of the application, firstly, the strategy determination system constructs a finite element baseline model of the body-in-white.

[0285] Among them, the finite element reference model refers to the discretized mathematical model used for numerical simulation based on the actual geometry of the body-in-white. It is usually composed of nodes, elements (such as shell elements and beam elements) and mesh.

[0286] It can be seen that the construction strategy determines the foundation for establishing a digital physical model for subsequent performance calculations and simulation analysis.

[0287] For example, the engineer imports the computer-aided design (CAD) model of the body-in-white, sets the mesh size constraints in the software, generates a finite element mesh model containing tens of thousands of nodes, and saves it as a baseline model file.

[0288] This provides a standardized simulation analysis platform, ensuring the repeatability and accuracy of subsequent calculation results.

[0289] Then, the strategy determination system uses a finite element baseline model to determine at least one of the following structural information of the body-in-white: load-bearing structural components, connection relationships, and material properties.

[0290] The connection relationship refers to the connection rules between at least two load-bearing structural components. For example, the connection relationship refers to the physical connection logic and positional information formed between two or more load-bearing structural components through welding, gluing, bolting, or other methods.

[0291] Material properties are used to describe the materials used to create load-bearing structural components. Material properties refer to the mechanical properties of the metallic materials (such as high-strength steel and aluminum alloys) that make up the load-bearing structural components, such as elastic modulus, Poisson's ratio, density, and stress-strain curves.

[0292] It can be seen that the construction strategy determines the key physical components of the body-in-white, as well as the interconnection methods and material properties of the body-in-white, by extracting or defining them from the benchmark model.

[0293] This clarifies the physical object being simulated, as well as its boundary conditions and material basis, providing necessary data support for accurately calculating structural performance.

[0294] Figure 3 This is a flowchart of a multi-performance collaborative positive optimization method for body-in-white based on heterogeneous parameter systems and deep progressive learning, according to an embodiment of this application. Figure 3 As shown, the method includes the following steps S301 to S308.

[0295] Step S301 involves constructing the structural baseline model and calculating initial performance values. In this embodiment, a passenger vehicle body-in-white from a certain platform is selected as the research object, and a finite element baseline model is established using shell elements with a mesh size of 5mm. Subsequently, bending, torsion, and modal analysis conditions and corresponding boundary conditions are set, and the initial performance values ​​of the body-in-white are calculated, specifically including bending stiffness, torsional stiffness, first-order bending frequency, and first-order torsional frequency.

[0296] Step S302 involves constructing a structured multidimensional parameter system for heterogeneous logic. This embodiment employs a differentiated parameter construction mechanism driven by global constraints and responses in key regions, integrating topological coupling parameters based on structural connections to establish value constraints, ultimately forming a multidimensional parameter vector X with a total of 36 design variables. Specifically, in the vehicle geometry dimension, the vehicle length, width, height, wheelbase, front and rear track widths, front and rear overhang lengths, and body mass parameters are determined based on the overall vehicle layout scheme. In the connection process dimension, the type of silicone sealant used in areas such as the windshield and side windows is selected as a design variable. In the key thickness and cross-section dimensions, based on the sensitivity coefficient calculated using the single-parameter perturbation method, the top 20% of parameters in terms of performance response gradient are selected from the candidate parameters. These parameters cover the thickness values ​​of areas such as the A-pillar, B-pillar, C-pillar, sill beam, longitudinal beam, floor, front bulkhead, rear bulkhead, and roof crossbeam, as well as the cross-sectional geometric parameters of key paths such as the C-ring, rear longitudinal beam, and door frame. Furthermore, the above parameters are correlated and constrained through engineering topology rules, and embedded into the constraint processing module of the optimization algorithm in the form of inequality functions. For example, the thickness ratio constraint of adjacent structural components (such as B-pillar and door sill beam) is set to 0.8:1.2 to ensure a smooth transition in stiffness, and the deviation tolerance of the total length of the vehicle, wheelbase and the sum of front and rear overhangs is set to ±10mm to ensure overall layout compatibility.

[0297] Step S303 involves constructing the initial sample database. In this embodiment, Latin hypercube sampling is used to generate 200 initial samples in a 36-dimensional design space, and finite element analysis is performed on the parameters of each sample to obtain the corresponding structural performance response, thereby forming the initial sample dataset.

[0298] Step S304 initializes the deep feedforward neural network surrogate model. A multi-layer fully connected deep feedforward neural network surrogate model with 36 input nodes, 4 hidden layers, and 4 output nodes is constructed. The Rectified Linear Unit (ReLU) is used as the activation function, and the Adaptive Moment Estimation (Adam) is used as the optimizer for initial training. The initial sample dataset is divided into training, validation, and test sets in a 70%:15%:15% ratio. Early stopping is used to prevent overfitting of the surrogate model and to ensure the determination coefficient of the test set. Not less than 0.95.

[0299] Step S305 defines the multi-objective optimization problem for the forward design paradigm. It establishes a performance-first, quality-constrained forward optimization approach, setting optimization objectives and conditions with maximizing structural performance as the goal and overall vehicle mass as a hard constraint. A multi-objective evolutionary algorithm with an embedded surrogate model is used for rapid evaluation. Combining the topological coupling constraints and the mapping relationship between design parameters and vehicle mass established in step S302, feasible region filtering is performed and solutions that violate engineering logic are directly eliminated during the iteration process, ensuring that the optimization iteration always converges efficiently within the engineering feasible space.

[0300] Step S306 is a two-scale uncertainty determination mechanism based on the coupling of Euclidean distance and distribution density. First, the multidimensional parameter vector... All variables are normalized, and the normalized Euclidean distance between the candidate solution and all samples in the existing dataset is calculated. Subsequently, the mean of the 10 nearest normalized distances of the candidate solution across all samples is calculated using the K-nearest neighbor method (K=10 in this embodiment). And further calculate all The average of the 10 nearest normalized distances of a sample Set dynamic threshold If the minimum normalized Euclidean distance of the candidate solutions is... Exceeding this dynamic threshold If the result is positive, the scheme is determined to be in the prediction uncertainty region of the surrogate model. At this time, finite element simulation is called to obtain high-precision results, and the new sample is added to the database.

[0301] Step S307 is a progressive model update strategy to prevent forgetting. A model update is triggered when the cumulative number of new samples reaches a set threshold. During the update process, the weight parameters of the first two layers (i.e., the input layer and the first hidden layer) closest to the input are locked, and gradient updates of the weight parameters of these two layers are prohibited to preserve general feature extraction capabilities. Only the weight parameters of the remaining three layers at the output are locally fine-tuned, and a learning rate lower than the initial training rate is used to update the weight parameters of the remaining layers. A regularization term is added to the loss function to limit the magnitude of weight changes, thereby preventing the degradation of historical knowledge.

[0302] Step S308 involves iterative convergence and scheme output. Steps S305 to S307 are repeated until the preset convergence condition is met. Finally, the Pareto front solution set is output and engineering verification is performed.

[0303] Figure 4 This is a schematic diagram of a dual-scale uncertainty determination mechanism based on the coupling of Euclidean distance and distribution density according to an embodiment of this application, as shown below. Figure 4As shown, step S401 is executed first to generate candidate solutions using algorithm optimization. Then, step S402 is executed to perform Min-Max normalization on the parameters. Next, step S403 is executed to calculate the minimum normalized Euclidean distance of the candidate solutions. In step S404, an adaptive threshold is calculated based on the distribution density. Then proceed to step S405 to determine whether... > If the determination is otherwise, proceed to step S406, determine the parameter space region, directly output the predicted value, and proceed to step S409 for subsequent steps. If the determination is yes, proceed to step S407, classify the parameter space region as an uncertain region and trigger finite element calculation. After the finite element calculation is completed, proceed to step S408, supplement the corresponding sample into the dataset, and then proceed to step S409 for subsequent steps.

[0304] As can be seen, this application constructs a parameter system that conforms to engineering practice. Based on a differentiated parameter construction mechanism that combines global constraint-driven, key region response-driven, and topological coupling constraints, a multi-dimensional parameter system is formed, with clear value constraint relationships between each parameter. This multi-dimensional parameter system not only ensures the stability of the vehicle architecture but also accurately taps into the structural performance potential, effectively overcoming the limitations of variable selection caused by single thickness optimization or discrete module libraries in existing technologies, making the optimization results closer to engineering practice.

[0305] This application achieves efficient utilization of computational resources. By introducing a dual-scale uncertainty determination mechanism based on the coupling of Euclidean distance and distribution density, uncertain regions of the predictive surrogate model are dynamically identified during the optimization process, and high-precision samples are supplemented only in these regions. This dual-scale uncertainty determination mechanism achieves precise allocation of computational resources, significantly avoiding meaningless global search waste and improving optimization efficiency compared to full-space sampling and Pareto solution set full backfilling strategies.

[0306] This application possesses the ability to preserve historical knowledge during model updates. By employing a progressive update strategy that combines weight freezing, local fine-tuning, and the introduction of a regularization term into the loss function, it effectively solves the problem of historical knowledge degradation that easily occurs during model updates in traditional techniques. Thus, while absorbing knowledge from new samples, it retains the predictive ability for the original sample space, thereby ensuring the convergence stability of the optimization process and the long-term usability of the model.

[0307] This application establishes a forward design paradigm geared towards engineering practice. Unlike traditional optimization methods that aim to minimize mass, this application sets mass as a hard constraint, focusing on maximizing structural stiffness and modal performance. This paradigm aligns with the forward development logic of exploring performance limits under weight constraints, allowing the output results to be directly applied to engineering practice. It avoids performance degradation caused by excessive pursuit of lightweighting, demonstrating significant engineering practicality.

[0308] Figure 5 This is a schematic diagram of a device for determining the construction strategy of the body-in-white in a vehicle according to an embodiment of this application, as shown below. Figure 5 As shown, the vehicle body-in-white construction strategy determination device 500 may include: a construction module 501, used to construct a set of structural variables based on the structural and performance dimensions of the body-in-white to be constructed; a first determination module 502, used to determine at least one first constraint condition for the body-in-white based on the set of structural variables, wherein the first constraint condition is used to constrain the engineering feasibility of the structure of the body-in-white; a second determination module 503, used to determine at least one candidate construction strategy for the body-in-white based on the first constraint condition, the objective function of the body-in-white, and the second constraint condition, wherein the objective function is used to indicate whether the performance of the structure of the body-in-white constructed using the candidate construction strategy meets the construction requirements of the body-in-white, and the second constraint condition is used to constrain vehicles that possess a body-in-white constructed through the candidate construction strategy; and a third determination module 504, used to determine a target construction strategy for constructing the body-in-white based on at least one candidate construction strategy for the body-in-white.

[0309] Embodiments of this application also provide a vehicle, including: a body-in-white constructed using a target construction strategy determined by the methods in various embodiments of this application.

[0310] Figure 6 This is a schematic diagram of an electronic device according to an embodiment of this application, such as... Figure 6 As shown, an electronic device 600 is also provided, including: a memory 104 storing an executable program; and a processor 102 for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0311] Embodiments of this application also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0312] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0313] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0314] Embodiments of this application also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of this application.

[0315] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0316] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection can be through some interfaces; the indirect coupling or communication connection between units or modules can be electrical or other forms.

[0317] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0318] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0319] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0320] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for determining the construction strategy of the body-in-white in a vehicle, characterized in that, include: Based on the structural and performance dimensions of the body-in-white to be constructed, a set of structural variables is constructed; Based on the set of structural variables, at least one first constraint condition is determined for the body-in-white, wherein the first constraint condition is used to constrain the engineering feasibility of the structure of the body-in-white; Based on the first constraint, the objective function of the body-in-white, and the second constraint, at least one candidate construction strategy for the body-in-white is determined, wherein the objective function is used to indicate whether the performance of the structure of the body-in-white constructed using the candidate construction strategy meets the construction requirements of the body-in-white, and the second constraint is used to constrain the vehicle having the body-in-white constructed by the candidate construction strategy. Based on at least one of the candidate construction strategies of the body-in-white, a target construction strategy for constructing the body-in-white is determined.

2. The method according to claim 1, characterized in that, The step of determining at least one candidate construction strategy for the body-in-white based on the first constraint, the objective function of the body-in-white, and the second constraint includes: An evolutionary model is used to evaluate the first constraint, the objective function, and the second constraint to obtain the target feasible region of the body-in-white. The evolutionary model is used to determine the candidate structural parameters of the body-in-white. The candidate structural parameters in the target feasible region satisfy the first constraint, the objective function, and the second constraint. Using a surrogate model, candidate performance parameters corresponding to the candidate structural parameters in the target feasible domain are determined; Using the candidate structural parameters and the candidate performance parameters, a candidate construction strategy is constructed, wherein the candidate construction strategy represents a rule for constructing a body-in-white that satisfies the candidate structural parameters and possesses the candidate performance parameters.

3. The method according to claim 2, characterized in that, The step of evaluating the first constraint, the objective function, and the second constraint using an evolutionary model to obtain the target feasible region of the body-in-white includes: Using the evolutionary model, the initial feasible region of the body-in-white is determined, wherein the candidate structural parameters in the initial feasible region satisfy the objective function and the second constraint condition; Using the evolutionary model, the target feasible region is determined from the initial feasible region, wherein the candidate structural parameters in the target feasible region satisfy at least one of the first constraints.

4. The method according to claim 2, characterized in that, The step of using a surrogate model to determine the candidate performance parameters corresponding to the candidate structural parameters in the target feasible domain includes: The candidate structural parameters in the target feasible domain are input into the surrogate model, and the surrogate model is used to determine the candidate performance parameters that have a mapping relationship with the candidate structural parameters.

5. The method according to claim 4, characterized in that, The method further includes: From the set of structural variables, determine the structural parameter samples and obtain the performance parameter samples corresponding to the structural parameter samples; The surrogate model is initialized using the structural parameter sample and the performance parameter sample, wherein the initialized surrogate model is used to determine the candidate performance parameters corresponding to the candidate structural parameters in the target feasible domain.

6. The method according to claim 5, characterized in that, Determining the structural parameter sample from the set of structural variables includes: The set of structural variables is uniformly divided into a target number of subsets of structural variables; Randomly select the structural variables from the target number of the structural variable subsets respectively; The target number of structural variables are randomly sorted to form an initial structural parameter sample; The structural parameter sample is determined from the initial structural parameter sample, wherein the initial structural parameter sample includes the structural parameter sample and invalid structural parameter samples.

7. The method according to claim 5, characterized in that, The step of obtaining the performance parameter sample corresponding to the structural parameter sample includes: Finite element simulation was performed on the structural parameter sample to obtain the performance parameter sample; And / or, The initialization of the surrogate model using the structural parameter samples and the performance parameter samples includes: The structural parameter samples and the performance parameter samples are input into the proxy model, and the proxy model is used to generate a mapping relationship between the structural parameter samples and the performance parameter samples; The proxy model that successfully establishes the mapping relationship is identified as the initialized proxy model.

8. The method according to claim 1, characterized in that, The determination of a target construction strategy for constructing the body-in-white based on at least one of the candidate construction strategies includes: At least one first construction strategy is identified from the candidate construction strategies, wherein the reliability of the first construction strategy is less than the reliability of the candidate construction strategies other than the first construction strategy among the candidate construction strategies; Using the first construction strategy, at least one second construction strategy is generated, wherein the reliability of the second construction strategy is greater than the reliability of the first construction strategy; The target construction strategy is determined from the candidate construction strategies, which include the first construction strategy, the second construction strategy, and the candidate construction strategies other than the first construction strategy.

9. The method according to claim 8, characterized in that, The candidate construction strategies include candidate structural parameters and candidate performance parameters, and identifying at least one first construction strategy from the candidate construction strategies includes: A distance information set corresponding to at least one of the candidate construction strategies is determined, wherein the distance information in the distance information set is used to represent the similarity between the candidate structural parameters and the structural parameter samples used to initialize the proxy model, and the number of distance information in the distance information set corresponds to the number of structural parameter samples; In response to the minimum distance information in the distance information set exceeding a distance threshold, the candidate construction strategy is determined to be the first construction strategy.

10. The method according to claim 8, characterized in that, The first construction strategy includes candidate structure parameters and candidate performance parameters. Generating at least one second construction strategy using the first construction strategy includes: Finite element simulation was performed on the candidate structural parameters to obtain the actual performance parameters corresponding to the candidate structural parameters; Determine the actual structural parameters that the body-in-white must meet to achieve the actual performance parameters; The second construction strategy is constructed using the actual performance parameters and actual structural parameters.

11. The method according to claim 8, characterized in that, Determining the target construction strategy from candidate construction strategies other than the first construction strategy, which include the first construction strategy, the second construction strategy, and the candidate construction strategies, includes: In response to the second construction strategy reaching the target number, the initialized proxy model is updated to obtain the updated proxy model; Using the updated proxy model, the target construction strategy is determined from the first construction strategy, the second construction strategy, and the third construction strategy.

12. The method according to claim 11, characterized in that, The initialized proxy model includes an input end and an output end. Updating the initialized proxy model to obtain the updated proxy model includes: The weight parameters of the previous layer at the input are fixed, and the weight parameters of the remaining layer at the output are updated. In response to the updated remaining layer weight parameters and the original remaining layer weight parameters satisfying the preset weight change constraints, the initialized proxy model is updated using the updated remaining layer weight parameters to obtain the updated proxy model.

13. The method according to claim 1, characterized in that, The set of structural variables includes a multidimensional design parameter vector. Based on the structural and performance dimensions of the body-in-white to be constructed, the set of structural variables is constructed, including: Based on the macroscopic architecture parameters of the vehicle, the first-level structural information of the body-in-white to be constructed is generated, and based on the finite element reference model of the body-in-white, the second-level structural information of the body-in-white to be constructed is generated, wherein the first level is higher than the second level. Establish the topological coupling relationship between the structural information of the first level and the structural information of the second level; Based on the topological coupling relationship, the structural information of the first level and the structural information of the second level are integrated into the multidimensional design parameter vector.

14. The method according to any one of claims 1 to 13, characterized in that, The method further includes: Construct the finite element baseline model of the body-in-white; Using the finite element baseline model, at least one of the following structural information of the body-in-white is determined: load-bearing structural components, connection relationships, and material properties, wherein the connection relationships are connection rules between at least two load-bearing structural components; and the material properties are used to represent the material used to generate the load-bearing structural components.

15. A device for determining the construction strategy of the body-in-white in a vehicle, characterized in that, include: The construction module is used to construct a set of structural variables based on the structural and performance dimensions of the body-in-white to be constructed; The first determining module is used to determine at least one first constraint condition for the body-in-white based on the set of structural variables, wherein the first constraint condition is used to constrain the engineering feasibility of the structure of the body-in-white. The second determining module is used to determine at least one candidate construction strategy for the body-in-white based on the first constraint, the objective function of the body-in-white, and the second constraint. The objective function is used to indicate whether the performance of the structure of the body-in-white constructed using the candidate construction strategy meets the construction requirements of the body-in-white. The second constraint is used to constrain the vehicle having the body-in-white constructed by the candidate construction strategy. The third determining module is used to determine a target construction strategy for constructing the body-in-white based on at least one of the candidate construction strategies.

16. A vehicle, characterized in that, include: A body-in-white constructed using the target construction strategy determined by the method described in any one of claims 1 to 14.

17. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method according to any one of claims 1 to 14.

18. A computer program product comprising a computer program that, when executed by a processor, implements the method of any one of claims 1 to 14.

Images