Vehicle body structure collaborative optimization method and system based on multi-part intelligent body

Abstract

The invention belongs to the field of vehicle body intelligent design, and particularly discloses a vehicle body structure collaborative optimization method and system based on a multi-part intelligent agent, and the method comprises the steps: constructing a multi-part intelligent agent system, and defining a design motion space and a performance state space of each intelligent agent; constructing a global reward function and a local reward function of the intelligent design of the vehicle body structure, and constructing a mixed reward function based on the global reward function and the local reward function; under the coordination of the joint optimization intelligent controller, performing iterative multi-stage cooperative training by taking the mixed reward function as an optimization target, and outputting a trained agent strategy network; and performing parallel collaborative optimization on the vehicle body initial design scheme by using the trained agent strategy network to obtain a coordinated multi-part optimization design result. Through the method, the performance balance of the whole vehicle can be realized, and the design efficiency is improved.

Description

Technical Field

[0001] This application belongs to the field of intelligent vehicle body design, and more specifically, relates to a method and system for collaborative optimization of vehicle body structure based on multi-component intelligent agents. Background Technology

[0002] The body-in-white is the skeleton of a car, and its performance directly affects the vehicle's safety, lightweighting level, and durability. Traditional body structure optimization typically employs a "serial" or "isolated" approach, optimizing key components such as sill beams, A / B pillars, and the floor separately. This method has inherent drawbacks. First, the car body is a complete force transmission system; optimizing a single component may disrupt the continuity of the overall load path, leading to the "weakest link effect"—where performance improvement in one component comes at the expense of performance deterioration in another. Furthermore, the coordination between different components is poor; increasing the stiffness of one component may alter the distribution of collision forces throughout the vehicle, thus imposing new and unforeseen performance requirements on related components. In addition, sequential optimization requires significant manual intervention and iteration, relying on engineers' experience to coordinate conflicts between different components, resulting in low efficiency.

[0003] Therefore, in response to the problems of vehicle performance imbalance and low design efficiency caused by the isolated design of components in traditional vehicle body structure optimization, how to achieve overall vehicle performance balance and improve design efficiency is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of this application is to achieve a balanced performance of the whole vehicle and improve design efficiency.

[0005] To achieve the above objectives, in a first aspect, this application provides a method for collaborative optimization of vehicle body structure based on multi-component intelligent agents, comprising: A multi-component intelligent agent system is constructed, and the design action space and performance state space of each intelligent agent are defined; the multi-component intelligent agent system includes a finite element model of the body-in-white, intelligent agent clusters corresponding to different body components, and a joint optimization intelligent controller; Construct a global reward function and a local reward function for intelligent design of vehicle body structure, and construct a hybrid reward function based on the global reward function and the local reward function; Under the coordination of the joint optimization intelligent controller, iterative multi-stage collaborative training is performed with the hybrid reward function as the optimization objective, and the trained agent policy network is output. The trained agent policy network is used to perform parallel collaborative optimization of the initial vehicle body design scheme, resulting in a coordinated multi-component optimized design.

[0006] Optionally, the intelligent agent cluster includes at least a sill beam intelligent agent, an integrated die-cast rear floor intelligent agent, and a B-pillar intelligent agent; The definition of the design action space and performance state space for each agent includes: The design action space of the door sill beam intelligent body is defined by the thickness of the reinforcing plate, and the performance state space of the door sill beam intelligent body is defined by the mass, maximum intrusion, maximum energy absorption value, and contact reaction force at the connection between the door sill beam and the B-column and the floor. The design motion space of the integrated die-cast floor intelligent body is defined by the thickness of the reinforcing ribs, and the performance state space of the integrated die-cast floor intelligent body is defined by its own weight, maximum deformation and stress concentration coefficient at the connection between the integrated die-cast floor and the threshold beam. The action space of the B-pillar intelligent agent is defined by using the segment thickness of the reinforcing plate as a variable, and the performance state space of the B-pillar intelligent agent is defined by the weight of the B-pillar reinforcing plate, the maximum intrusion amount, and the local stiffness of the connection point between the B-pillar and the sill beam.

[0007] Optionally, the construction process of the global reward function includes: The torsional stiffness score, bending stiffness score, side impact score, frontal impact score, and rear impact score of the vehicle body are obtained through finite element analysis. Obtain the total mass of the body-in-white and calculate the compatibility penalty values ​​between the components to be optimized; The torsional stiffness score, bending stiffness score, side impact score, frontal impact score, rear impact score, total mass of the white body, and compatibility penalty value are multiplied by their respective weighting coefficients and then weighted and combined to obtain the global reward function value.

[0008] Optionally, the method for obtaining the compatibility penalty value includes: Obtain the maximum equivalent stress on adjacent components at the connection interface of each component and the material yield strength of the adjacent components; Determine the ratio of the maximum equivalent stress to the material yield strength, calculate the stress concentration value of each connection interface based on the difference between the ratio and the preset stress threshold, and sum the stress concentration values ​​of all connection interfaces to obtain the interface stress concentration penalty sub-item; The total mass of a specified local area of ​​the vehicle body and the total mass of the body-in-white are obtained to determine the mass ratio. The mass distribution imbalance penalty item is calculated based on the difference between the mass ratio and a preset mass distribution threshold. The interface stress concentration penalty item and the mass distribution imbalance penalty item are multiplied by their respective weighting coefficients and then weighted and combined to obtain the compatibility penalty value.

[0009] Optionally, the construction process of the hybrid reward function includes: Obtain the performance reward and quality penalty corresponding to the target component intelligent agent, wherein the performance reward includes at least the component's stiffness, deformation, intrusion amount, or connection reaction force; The performance reward and quality penalty are multiplied by their respective weight coefficients and then weighted and combined to obtain the local reward function of the target component agent. The global reward function and the sum of all local reward functions are multiplied by their respective weight coefficients and then weighted and summed to obtain the hybrid reward function.

[0010] Optionally, the agent policy network is trained using a multi-stage training strategy; The multi-stage training strategy includes a first training stage and a second training stage; In the first training phase, the weight of the local reward function in the hybrid reward function is increased to guide each agent to prioritize improving the performance of its corresponding components. In the second training phase, the weight of the global reward function in the hybrid reward function is increased to guide the agent to optimize the overall performance of the vehicle and achieve multi-component collaborative optimization.

[0011] Optionally, the collaborative policy training of the joint optimization intelligent controller adopts a multi-agent reinforcement learning algorithm, and updates the policy network of each agent according to the local reward function and the global reward function value; During the policy network update process, agents communicate with each other through communication channels to exchange interface information in order to achieve collaboration among agents.

[0012] Secondly, this application also provides a vehicle body structure collaborative optimization system based on multi-component intelligent agents, including: The system construction module is used to build a multi-component intelligent agent system and define the design action space and performance state space of each intelligent agent; the multi-component intelligent agent system includes a finite element model of the body-in-white, intelligent agent clusters corresponding to different body components, and a joint optimization intelligent controller; The function construction module is used to construct the global reward function and the local reward function for intelligent design of the vehicle body structure, and to construct a hybrid reward function based on the global reward function and the local reward function; The training module is used to perform iterative multi-stage collaborative training under the coordination of the joint optimization intelligent controller, with the hybrid reward function as the optimization objective, and output the trained agent policy network. The collaborative optimization module is used to perform parallel collaborative optimization of the initial vehicle body design scheme using the trained agent policy network, so as to obtain the coordinated multi-component optimized design result.

[0013] Thirdly, this application provides an electronic device, comprising: at least one memory for storing a program; and at least one processor for executing the program stored in the memory, wherein when the program stored in the memory is executed, the processor is configured to execute the method described in the first aspect or any possible implementation thereof.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.

[0015] Fifthly, this application provides a computer program product that, when run on a processor, causes the processor to perform the method described in the first aspect or any possible implementation thereof.

[0016] It is understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0017] Overall, the technical solutions conceived in this application have the following beneficial effects compared with the prior art: (1) This application achieves global balance and coordinated improvement of vehicle-level performance through a multi-agent collaborative mechanism, effectively avoiding load path breakage and performance bottleneck caused by isolated component design; adopts a multi-agent reinforcement learning framework with centralized training and distributed execution. After training, each agent can independently and in parallel output an optimization scheme that is coordinated with the whole vehicle, without relying on the manual coordination and repeated iteration of engineers, which greatly shortens the design cycle of the vehicle body structure and significantly improves R&D efficiency.

[0018] (2) The system of this application has good scalability and can easily incorporate more intelligent agents of body parts. It can solve the problem of whole vehicle collaborative optimization with higher dimensions and more complex coupling relationships without redesigning the system architecture, and provides an engineeringable technical path for large-scale intelligent design of body structure. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the vehicle body structure collaborative optimization method based on multi-component intelligent agents provided in this application embodiment; Figure 2 This is a schematic diagram of a multi-component intelligent body system for intelligent design of vehicle body structure according to an embodiment of this application; Figure 3 This is a schematic diagram illustrating the design and training process of a multi-component intelligent agent system according to an embodiment of this application; Figure 4This is a schematic diagram of the vehicle body structure collaborative optimization system based on multi-component intelligent agents provided in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0021] In this article, the term "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The symbol " / " in this article indicates that the related objects are in an "or" relationship; for example, A / B means A or B.

[0022] The terms "first" and "second," etc., used in the specification and claims herein are used to distinguish different objects, not to describe a specific order of objects. For example, "first response message" and "second response message," etc., are used to distinguish different response messages, not to describe a specific order of response messages.

[0023] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0024] In the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more, for example, multiple processing units means two or more processing units, multiple elements means two or more elements, etc.

[0025] The embodiments of this application are described below with reference to the accompanying drawings.

[0026] Reference Figure 1 This application provides a method for collaborative optimization of vehicle body structure based on multi-component intelligent agents, including: S101. Construct a multi-component intelligent agent system and define the design action space and performance state space for each intelligent agent; the multi-component intelligent agent system includes a finite element model of the body-in-white, intelligent agent clusters corresponding to different body components, and a joint optimization intelligent controller; S102. Construct a global reward function and a local reward function for intelligent design of the vehicle body structure, and construct a hybrid reward function based on the global reward function and the local reward function; S103. Under the coordination of the joint optimization intelligent controller, iterative multi-stage collaborative training is carried out with the hybrid reward function as the optimization objective, and the trained agent policy network is output. S104. The trained agent policy network is used to perform parallel collaborative optimization of the initial vehicle body design scheme to obtain the coordinated multi-component optimization design result.

[0027] Specifically, in this embodiment, a multi-component intelligent agent system is first constructed through S101. This system consists of three core parts: first, a parameterized high-fidelity finite element model of the white body, which serves as a simulation environment for intelligent agent interaction and performance evaluation; second, an intelligent agent cluster composed of multiple structural optimization intelligent agents corresponding one-to-one with different body parts, with each intelligent agent responsible for the design decision of a specific component; and third, a joint optimization intelligent controller, which is used to coordinate the collaborative training and strategy updates of each intelligent agent.

[0028] After the system is built, each agent independently defines its design action space and performance state space. The design action space is a set of design variables that the agent can adjust, such as the set of thicknesses of the internal reinforcement plates of the door sill beam, the set of thicknesses of the floor reinforcement ribs after one-piece die casting, and the set of segmented thicknesses of the B-pillar reinforcement plates; the performance state space is a set of indicators used by the agent to perceive the current performance state of the components, including mass, maximum intrusion, maximum energy absorption, maximum deformation, contact reaction force at the connection, stress concentration factor, local stiffness, etc.

[0029] A multi-component joint optimization intelligent controller may include the following modules: The structural state fusion module collects all local states observed by the agents. The local states involved in this invention include maximum stress, maximum displacement, mass, etc. The vehicle-component hybrid hierarchical evaluation module calculates reward values ​​based on local and global performance indicators. The cooperative strategy training module uses a centralized training and distributed execution framework to train agents to cooperate. The communication channel module allows agents to exchange limited information related to cooperative optimization during training.

[0030] Secondly, after the system is built, a reward function system to guide the collaborative optimization of intelligent agents is constructed through S102. The global reward function is used to evaluate the overall performance of the vehicle and is calculated by weighted fusion of torsional stiffness score, bending stiffness score, side impact score, frontal impact score, rear impact score, total mass of the body-in-white, and compatibility penalty value between components.

[0031] The compatibility penalty value further includes an interface stress concentration penalty and a mass distribution imbalance penalty. The former is used to prevent excessive stress or even tearing at the connection due to component stiffness mismatch, while the latter is used to prevent excessive material accumulation in local areas, leading to unreasonable mass distribution. The local reward function is used to evaluate the performance of individual components and is composed of a weighted combination of component performance rewards and mass penalties. The performance rewards include at least indicators such as stiffness, deformation, intrusion, or connection reaction force. Finally, the global reward function is weighted and fused with the local reward functions of all agents to construct a hybrid reward function, which serves as a unified optimization objective to guide multi-agent collaborative training.

[0032] Furthermore, through S103, under the coordination of the joint optimization intelligent controller, a multi-agent collaborative training process is initiated. The training adopts a multi-stage reward strategy mechanism: in the first training stage, the weight of the local reward function in the mixed reward function is appropriately increased to guide each agent to prioritize the optimization of the performance of its corresponding components, enabling it to quickly acquire basic optimization capabilities; in the second training stage, the weight of the global reward function is gradually increased to guide the agents to shift their focus from local performance to the overall performance of the vehicle, achieving collaborative optimization and performance balance among multiple components.

[0033] In each training iteration, each agent selects an action from its own design action space based on the currently perceived local state information, updates the structural parameters of the corresponding components, and integrates the updated component solutions into the vehicle finite element model for performance simulation evaluation, calculating global and local reward values. The joint optimization intelligent controller employs a multi-agent reinforcement learning algorithm, updating the policy network parameters of each agent based on reward feedback. Simultaneously, the agents exchange interface state information related to inter-component collaborative optimization through communication channels. After multiple iterations of training, the policy network gradually converges, outputting the trained agent policy network.

[0034] Finally, after training via S104, the converged agent policy network is deployed into the actual design process. When the initial vehicle body design is input, each agent adjusts the design parameters of its corresponding components in parallel based on its learned optimization strategy. The final output is a multi-component joint optimization design result that achieves the best balance between the overall vehicle performance and the performance of each component.

[0035] Optionally, the intelligent agent cluster includes at least a sill beam intelligent agent, an integrated die-cast rear floor intelligent agent, and a B-pillar intelligent agent; The definition of the design action space and performance state space for each agent includes: The design action space of the door sill beam intelligent body is defined by the thickness of the reinforcing plate, and the performance state space of the door sill beam intelligent body is defined by the mass, maximum intrusion, maximum energy absorption value, and contact reaction force at the connection between the door sill beam and the B-column and the floor. The design motion space of the integrated die-cast floor intelligent body is defined by the thickness of the reinforcing ribs, and the performance state space of the integrated die-cast floor intelligent body is defined by its own weight, maximum deformation and stress concentration coefficient at the connection between the integrated die-cast floor and the threshold beam. The action space of the B-pillar intelligent agent is defined by using the segment thickness of the reinforcing plate as a variable, and the performance state space of the B-pillar intelligent agent is defined by the weight of the B-pillar reinforcing plate, the maximum intrusion amount, and the local stiffness of the connection point between the B-pillar and the sill beam.

[0036] Specifically, the component structure optimization intelligent agent cluster in this application embodiment includes: Agent 1: Threshold beam structure optimization agent; Agent 2: Integrated die-cast floor structure optimization agent; Agent 3: B-pillar structure optimization agent; Agent n: Intelligent agent for optimizing the structure of other components.

[0037] Definition of action space and state space for a single component intelligent agent: Definition of intelligent agent for sill beam structure optimization: Action space: The design variable for the sill beam in this application is the thickness of the internal reinforcing plate. , This is the serial number of the internal reinforcement plate. If the thickness of the reinforcing plate is 0, it means that the reinforcing plate does not exist; State space: The state variables of the sill beam involved in this application include the mass of the sill beam itself. Maximum intrusion under side-impact conditions Maximum energy absorption value and the contact reaction force at the connection with the B-pillar and the floor. ; Integrated die-cast floor structure optimization intelligent agent definition: Movement space: The design variable of the one-piece die-cast floor involved in this application is the thickness of the reinforcing ribs on the sides and bottom. , The internal reinforcing ribs are numbered. If the thickness of the reinforcing rib is 0, it means that the reinforcing rib does not exist. State space: The state quantities of the integral die-cast floor in this application include the weight of the integral die-cast part itself. Maximum deformation under bending conditions and the stress concentration factor at the connection with the threshold beam. ; B-pillar structure optimization intelligent agent: The design variables for the B-pillar involved in this application are the thicknesses of the middle, lower, and upper sections of the reinforcing plate. , ; State space: The weight of the B-pillar reinforcement plate involved in this application. Maximum intrusion under side-impact conditions and the local stiffness at the connection point with the threshold beam. ; Optionally, the construction process of the global reward function includes: The torsional stiffness score, bending stiffness score, side impact score, frontal impact score, and rear impact score of the vehicle body are obtained through finite element analysis. Obtain the total mass of the body-in-white and calculate the compatibility penalty values ​​between the components to be optimized; The torsional stiffness score, bending stiffness score, side impact score, frontal impact score, rear impact score, total mass of the white body, and compatibility penalty value are multiplied by their respective weighting coefficients and then weighted and combined to obtain the global reward function value.

[0038] Furthermore, the method for obtaining the compatibility penalty value includes: Obtain the maximum equivalent stress on adjacent components at the connection interface of each component and the material yield strength of the adjacent components; Determine the ratio of the maximum equivalent stress to the material yield strength, calculate the stress concentration value of each connection interface based on the difference between the ratio and the preset stress threshold, and sum the stress concentration values ​​of all connection interfaces to obtain the interface stress concentration penalty sub-item; The total mass of a specified local area of ​​the vehicle body and the total mass of the body-in-white are obtained to determine the mass ratio. The mass distribution imbalance penalty item is calculated based on the difference between the mass ratio and a preset mass distribution threshold. The interface stress concentration penalty item and the mass distribution imbalance penalty item are multiplied by their respective weighting coefficients and then weighted and combined to obtain the compatibility penalty value.

[0039] Specifically, this application embodiment constructs a global reward function for intelligent design of vehicle body structure. The reward function The intelligent agent is guided to improve the overall performance of the vehicle while reducing its own weight, specifically calculated using the following formula:

[0040] in, To score torsional stiffness, To score bending stiffness, To score side-impact performance, For head-on collision performance score, Score for rear-end collision performance. The score represents the total mass of the body-in-white. This refers to the compatibility penalty value between various components; , , , , , , These are the weight coefficients for the corresponding items; in, , , , , All results are obtained through finite element analysis. Preferably, if a sufficient amount of such data is accumulated, a machine learning model can be trained to quickly predict the outcome. To obtain directly by measurement, The following formula is used to calculate:

[0041] in, This represents the interface stress concentration penalty, the purpose of which is to prevent one component from being too rigid, causing force to be unable to be effectively transmitted, thus creating a risk of tearing at the connection point with another component. It is calculated according to the following formula:

[0042] in, For the first At each connection point, the maximum equivalent stress on adjacent components This indicates the yield strength of the material of the adjacent component. This is the stress threshold ratio. The number of interfaces considered in a multi-component intelligent agent system; Penalty for Imbalance in Quality Distribution The purpose is to prevent the agent from accumulating all materials in a localized area, leading to an unreasonable mass distribution. The calculation is performed using the following formula:

[0043] in, Let this be the total mass of a certain local area of ​​the vehicle body. The score represents the total mass of the body-in-white. The quality distribution threshold represents the maximum allowable proportion of a certain region to the total quality. Optionally, the construction process of the hybrid reward function includes: Obtain the performance reward and quality penalty corresponding to the target component intelligent agent, wherein the performance reward includes at least the component's stiffness, deformation, intrusion amount, or connection reaction force; The performance reward and quality penalty are multiplied by their respective weight coefficients and then weighted and combined to obtain the local reward function of the target component agent. The global reward function and the sum of all local reward functions are multiplied by their respective weight coefficients and then weighted and summed to obtain the hybrid reward function.

[0044] Specifically, this embodiment establishes a vehicle-component hybrid hierarchical evaluation module to calculate a hybrid reward function for multi-component intelligent agents. Calculate according to the following formula:

[0045] in, A global reward function is designed for intelligent vehicle body structure. For the first The reward function for the intelligent agent that optimizes the structure of individual components. , These are their respective weighting coefficients; in, Calculate according to the following formula:

[0046] in, For the first The performance bonus for each component specifically includes the parameters defined in step S101 of the above embodiment. For the first Quality penalties for individual components , These are their respective weighting coefficients; It should be noted that the weighting coefficients in this embodiment can be determined according to the actual situation.

[0047] Optionally, the agent policy network is trained using a multi-stage training strategy; The multi-stage training strategy includes a first training stage and a second training stage; In the first training phase, the weight of the local reward function in the hybrid reward function is increased to guide each agent to prioritize improving the performance of its corresponding components. In the second training phase, the weight of the global reward function in the hybrid reward function is increased to guide the agent to optimize the overall performance of the vehicle and achieve multi-component collaborative optimization.

[0048] In this embodiment, a multi-stage reward strategy mechanism is established, which includes a stage one and a stage two. Specifically, in the first phase of training, the weights are increased appropriately. This guides intelligent agents to prioritize optimizing their respective components. Phase Two: Gradually increase the weights during the later stages of training. This guides intelligent agents to focus more on overall vehicle performance and achieves collaborative optimization of multiple components; Optionally, the collaborative policy training of the joint optimization intelligent controller adopts a multi-agent reinforcement learning algorithm, and updates the policy network of each agent according to the local reward function and the global reward function value; During the policy network update process, agents communicate with each other through communication channels to exchange interface information in order to achieve collaboration among agents.

[0049] Specifically, in this embodiment, agent 1, agent 2, agent 3, ..., agent n each select an action space based on the current state of the component. The agents modify their respective component structures according to the selected action spaces and integrate them into the vehicle model; based on the updated component structures, analysis and calculations are performed to obtain the component structure optimization agent reward function and the vehicle body structure intelligent design global reward function. Preferably, during computational analysis, a pre-trained performance prediction model is used instead of finite element simulation calculation to improve the training efficiency of the agent.

[0050] The collaborative policy training module utilizes multi-agent algorithms such as the DDPG algorithm model to update the policy network of all agents. At the same time, the agents transmit simplified interface information through communication channels, thereby achieving better collaboration. After multiple iterations of training, a multi-component intelligent agent system for intelligent design of vehicle body structure is output. The intelligent agent system can quickly provide an optimization scheme that coordinates with other parts of the vehicle body based on its own strategy, thereby achieving efficient parallel design.

[0051] Reference Figure 2 , Figure 2 This is a schematic diagram of a multi-component intelligent agent system for intelligent design of vehicle body structure according to an embodiment of this application. Its core process is as follows: the intelligent agent cluster acts under a distributed execution strategy, influencing the high-fidelity model and generating state evaluation. The state is uploaded to the intelligent controller, which performs state fusion and hybrid evaluation, generates a new optimization strategy through collaborative training, and issues the new strategy to guide the intelligent agents to perform the next round of optimization, forming a closed loop of continuous learning and improvement.

[0052] Reference Figure 3 , Figure 3This is a schematic diagram illustrating the design and training process of a multi-component intelligent agent system according to an embodiment of this application. The specific process is as follows: Multi-component intelligent system establishment process: Definition of the action space of a single component intelligent agent; State space definition for a single component intelligent agent; Construct a global reward function for intelligent design of vehicle body structure; Construct a local reward function for intelligent design of vehicle body structure; Establish a hybrid hierarchical evaluation module for complete vehicles and components; Establish a multi-stage reward strategy mechanism; Training process for multi-component intelligent systems: Initialize the component structure state; Multi-component intelligent system; Modify the component structure according to the motion space; Obtain the reward function for the component structure optimization agent; Obtain the global reward function for intelligent design of the vehicle body structure; Update the policy network for all agents; Multiple iterations of training; Output a multi-component intelligent system.

[0053] Furthermore, this application also provides specific embodiments of the aforementioned intelligent agent: S01. Initial scenario of vehicle body component structure: The initial state of the vehicle body to be optimized is disassembled to obtain the initial structural state of the door sill beam, the integrated die-cast rear floor, and the B-pillar. The initial structural state is the initial version of the solution output in the concept design stage. The mechanical properties and lightweight level of the components in this state do not meet the standards. S02. Agent Optimization and Conflicts: Input the initial structural states of the door sill beam, the one-piece die-cast floor, and the B-pillar into the agent system to obtain the optimized door sill beam, the one-piece die-cast floor, and the B-pillar structure; Specifically, when optimizing the door sill beam separately, Agent 1 tends to increase the thickness of the reinforced version and the use of ultra-high strength steel to maximize side impact performance; When optimizing the floor separately, Agent 2 may reduce the wall thickness in the area connecting to the threshold beam in order to reduce weight; The above optimization process will result in a huge stress concentration at the connection between the rigid sill beam and the relatively weak rear floor during a side impact, which may cause the rear floor to tear, disrupt the overall load path, and reduce the safety of the vehicle. S03. Multi-agent joint optimization process: Agent 1: Significantly increases the stiffness of the door sill beam; Agent 2: Reduces the thickness of the rear floor connection area; Test results: During the side impact, the rear floor connection fails prematurely, and the overall intrusion amount deteriorates; Global reward: Due to compatibility penalty, the score is very low; Agent learning: Both agents receive negative feedback. Agent 1: Targeted reinforcement of specific areas connected to the B-pillar and rear floor, while reducing weight in non-critical areas. Agent 2: Adding local reinforcing ribs or appropriate wall thickness at the connection with the door sill beam to increase the strength of the connection area. Test results: Reasonable force transmission, excellent side impact performance, high level of lightweighting. Overall reward: High score. S04. Solution Output: After tens of thousands of joint training iterations, the optimal joint design solution is formed.

[0054] Reference Figure 4 This application also provides a vehicle body structure collaborative optimization system based on multi-component intelligent agents, including: The system construction module 410 is used to construct a multi-component intelligent agent system and define the design action space and performance state space of each intelligent agent; the multi-component intelligent agent system includes a finite element model of the body-in-white, intelligent agent clusters corresponding to different body components, and a joint optimization intelligent controller; The function construction module 420 is used to construct a global reward function and a local reward function for intelligent design of the vehicle body structure, and to construct a hybrid reward function based on the global reward function and the local reward function; Training module 430 is used to perform iterative multi-stage collaborative training under the coordination of the joint optimization intelligent controller, with the hybrid reward function as the optimization objective, and output the trained agent policy network. The collaborative optimization module 440 is used to perform parallel collaborative optimization of the initial vehicle body design scheme using the trained intelligent agent policy network, so as to obtain the coordinated multi-component optimized design result.

[0055] Optionally, the intelligent agent cluster includes at least a sill beam intelligent agent, an integrated die-cast rear floor intelligent agent, and a B-pillar intelligent agent; The definition of the design action space and performance state space for each agent includes: The design action space of the door sill beam intelligent body is defined by the thickness of the reinforcing plate, and the performance state space of the door sill beam intelligent body is defined by the mass, maximum intrusion, maximum energy absorption value, and contact reaction force at the connection between the door sill beam and the B-column and the floor. The design motion space of the integrated die-cast floor intelligent body is defined by the thickness of the reinforcing ribs, and the performance state space of the integrated die-cast floor intelligent body is defined by its own weight, maximum deformation and stress concentration coefficient at the connection between the integrated die-cast floor and the threshold beam. The action space of the B-pillar intelligent agent is defined by using the segment thickness of the reinforcing plate as a variable, and the performance state space of the B-pillar intelligent agent is defined by the weight of the B-pillar reinforcing plate, the maximum intrusion amount, and the local stiffness of the connection point between the B-pillar and the sill beam.

[0056] Optionally, the construction process of the global reward function includes: The torsional stiffness score, bending stiffness score, side impact score, frontal impact score, and rear impact score of the vehicle body are obtained through finite element analysis. Obtain the total mass of the body-in-white and calculate the compatibility penalty values ​​between the components to be optimized; The torsional stiffness score, bending stiffness score, side impact score, frontal impact score, rear impact score, total mass of the white body, and compatibility penalty value are multiplied by their respective weighting coefficients and then weighted and combined to obtain the global reward function value.

[0057] Optionally, the method for obtaining the compatibility penalty value includes: Obtain the maximum equivalent stress on adjacent components at the connection interface of each component and the material yield strength of the adjacent components; Determine the ratio of the maximum equivalent stress to the material yield strength, calculate the stress concentration value of each connection interface based on the difference between the ratio and the preset stress threshold, and sum the stress concentration values ​​of all connection interfaces to obtain the interface stress concentration penalty sub-item; The total mass of a specified local area of ​​the vehicle body and the total mass of the body-in-white are obtained to determine the mass ratio. The mass distribution imbalance penalty item is calculated based on the difference between the mass ratio and a preset mass distribution threshold. The interface stress concentration penalty item and the mass distribution imbalance penalty item are multiplied by their respective weighting coefficients and then weighted and combined to obtain the compatibility penalty value.

[0058] Optionally, the construction process of the hybrid reward function includes: Obtain the performance reward and quality penalty corresponding to the target component intelligent agent, wherein the performance reward includes at least the component's stiffness, deformation, intrusion amount, or connection reaction force; The performance reward and quality penalty are multiplied by their respective weight coefficients and then weighted and combined to obtain the local reward function of the target component agent. The global reward function and the sum of all local reward functions are multiplied by their respective weight coefficients and then weighted and summed to obtain the hybrid reward function.

[0059] Optionally, the agent policy network is trained using a multi-stage training strategy; The multi-stage training strategy includes a first training stage and a second training stage; In the first training phase, the weight of the local reward function in the hybrid reward function is increased to guide each agent to prioritize improving the performance of its corresponding components. In the second training phase, the weight of the global reward function in the hybrid reward function is increased to guide the agent to optimize the overall performance of the vehicle and achieve multi-component collaborative optimization.

[0060] Optionally, the collaborative policy training of the joint optimization intelligent controller adopts a multi-agent reinforcement learning algorithm, and updates the policy network of each agent according to the local reward function and the global reward function value; During the policy network update process, agents communicate with each other through communication channels to exchange interface information in order to achieve collaboration among agents.

[0061] Reference Figure 5 Based on the methods in the above embodiments, this application provides an electronic device that may include: a processor 510, a communications interface 520, a memory 530, and a communication bus 540. The processor 510, communications interface 520, and memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions stored in the memory 530 to execute the methods in the above embodiments.

[0062] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, 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 a portion 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.

[0063] Based on the methods in the above embodiments, this application provides a computer-readable storage medium storing a computer program that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0064] Based on the methods in the above embodiments, this application provides a computer program product that, when run on a processor, causes the processor to execute the methods in the above embodiments.

[0065] It is understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. A general-purpose processor can be a microprocessor or any conventional processor.

[0066] The method steps in this application embodiment can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.

[0067] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0068] It is understood that the various numerical designations used in the embodiments of this application are merely for the convenience of description and are not intended to limit the scope of the embodiments of this application.

[0069] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for collaborative optimization of vehicle body structure based on multi-component intelligent agents, characterized in that, include: Construct a multi-component intelligent agent system and define the design action space and performance state space for each intelligent agent; The multi-component intelligent agent system includes a finite element model of the body-in-white, intelligent agent clusters corresponding to different body components, and a joint optimization intelligent controller. Construct a global reward function and a local reward function for intelligent design of vehicle body structure, and construct a hybrid reward function based on the global reward function and the local reward function; Under the coordination of the joint optimization intelligent controller, iterative multi-stage collaborative training is performed with the hybrid reward function as the optimization objective, and the trained agent policy network is output. The trained agent policy network is used to perform parallel collaborative optimization of the initial vehicle body design scheme, resulting in a coordinated multi-component optimized design.

2. The vehicle body structure collaborative optimization method based on multi-component intelligent agents according to claim 1, characterized in that, The intelligent agent cluster includes at least a sill beam intelligent agent, an integrated die-cast rear floor intelligent agent, and a B-pillar intelligent agent. The definition of the design action space and performance state space for each agent includes: The design action space of the door sill beam intelligent body is defined by the thickness of the reinforcing plate, and the performance state space of the door sill beam intelligent body is defined by the mass, maximum intrusion, maximum energy absorption value, and contact reaction force at the connection between the door sill beam and the B-column and the floor. The design motion space of the integrated die-cast floor intelligent body is defined by the thickness of the reinforcing ribs, and the performance state space of the integrated die-cast floor intelligent body is defined by its own weight, maximum deformation and stress concentration coefficient at the connection between the integrated die-cast floor and the threshold beam. The action space of the B-pillar intelligent agent is defined by using the segment thickness of the reinforcing plate as a variable, and the performance state space of the B-pillar intelligent agent is defined by the weight of the B-pillar reinforcing plate, the maximum intrusion amount, and the local stiffness of the connection point between the B-pillar and the sill beam.

3. The vehicle body structure collaborative optimization method based on multi-component intelligent agents according to claim 1, characterized in that, The construction process of the global reward function includes: The torsional stiffness score, bending stiffness score, side impact score, frontal impact score, and rear impact score of the vehicle body are obtained through finite element analysis. Obtain the total mass of the body-in-white and calculate the compatibility penalty values ​​between the components to be optimized; The torsional stiffness score, bending stiffness score, side impact score, frontal impact score, rear impact score, total mass of the white body, and compatibility penalty value are multiplied by their respective weighting coefficients and then weighted and combined to obtain the global reward function value.

4. The vehicle body structure collaborative optimization method based on multi-component intelligent agents according to claim 3, characterized in that, The method for obtaining the compatibility penalty value includes: Obtain the maximum equivalent stress on adjacent components at the connection interface of each component and the material yield strength of the adjacent components; Determine the ratio of the maximum equivalent stress to the material yield strength, calculate the stress concentration value of each connection interface based on the difference between the ratio and the preset stress threshold, and sum the stress concentration values ​​of all connection interfaces to obtain the interface stress concentration penalty sub-item; The total mass of a specified local area of ​​the vehicle body and the total mass of the body-in-white are obtained to determine the mass ratio. The mass distribution imbalance penalty item is calculated based on the difference between the mass ratio and a preset mass distribution threshold. The interface stress concentration penalty item and the mass distribution imbalance penalty item are multiplied by their respective weighting coefficients and then weighted and combined to obtain the compatibility penalty value.

5. The vehicle body structure collaborative optimization method based on multi-component intelligent agents according to claim 1, characterized in that, The construction process of the hybrid reward function includes: Obtain the performance reward and quality penalty corresponding to the target component intelligent agent, wherein the performance reward includes at least the component's stiffness, deformation, intrusion amount, or connection reaction force; The performance reward and quality penalty are multiplied by their respective weight coefficients and then weighted and combined to obtain the local reward function of the target component agent. The global reward function and the sum of all local reward functions are multiplied by their respective weight coefficients and then weighted and summed to obtain the hybrid reward function.

6. The vehicle body structure collaborative optimization method based on multi-component intelligent agents according to claim 1, characterized in that, The agent policy network is trained using a multi-stage training strategy; The multi-stage training strategy includes a first training stage and a second training stage; In the first training phase, the weight of the local reward function in the hybrid reward function is increased to guide each agent to prioritize improving the performance of its corresponding components. In the second training phase, the weight of the global reward function in the hybrid reward function is increased to guide the agent to optimize the overall performance of the vehicle and achieve multi-component collaborative optimization.

7. The vehicle body structure collaborative optimization method based on multi-component intelligent agents according to claim 1, characterized in that, The collaborative policy training of the joint optimization intelligent controller adopts a multi-agent reinforcement learning algorithm, which updates the policy network of each agent according to the local reward function and the global reward function value. During the policy network update process, agents communicate with each other through communication channels to exchange interface information in order to achieve collaboration among agents.

8. A vehicle body structure collaborative optimization system based on multi-component intelligent agents, characterized in that, include: The system construction module is used to build a multi-component intelligent agent system and define the design action space and performance state space of each intelligent agent; the multi-component intelligent agent system includes a finite element model of the body-in-white, intelligent agent clusters corresponding to different body components, and a joint optimization intelligent controller; The function construction module is used to construct the global reward function and the local reward function for intelligent design of the vehicle body structure, and to construct a hybrid reward function based on the global reward function and the local reward function; The training module is used to perform iterative multi-stage collaborative training under the coordination of the joint optimization intelligent controller, with the hybrid reward function as the optimization objective, and output the trained agent policy network. The collaborative optimization module is used to perform parallel collaborative optimization of the initial vehicle body design scheme using the trained agent policy network, so as to obtain the coordinated multi-component optimized design result.

9. An electronic device, characterized in that, include: At least one memory for storing computer programs; At least one processor is configured to execute a program stored in the memory, wherein when the program stored in the memory is executed, the processor is configured to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is run on the processor, it causes the processor to perform the method as described in any one of claims 1-7.

Images