Air spring and vehicle performance collaborative optimization method, device, medium and equipment

Abstract

The application relates to the technical field of air spring optimization design, in particular to a kind of air spring and vehicle performance collaborative optimization method, device, medium and equipment, wherein, the method comprises: the virtual test field technology of whole vehicle, parameterization modeling, adaptive surrogate model and multi-objective intelligent optimization algorithm are deeply fused.The system automatically finds the air spring design parameter or control parameter that makes the overall performance of the whole vehicle optimal through iterative optimization, thereby realizing the collaborative design from parts to system level. Thus, the problems that the existing collaborative optimization method cannot consider model accuracy and simulation efficiency, the "island type" design process can only cause local optimal system performance, the intelligentization and adaptive ability of the optimization method are insufficient, and the real-time and foresight of the control strategy are insufficient are solved.

Description

Technical Field

[0001] This application relates to the field of air spring optimization design technology, and in particular to a method, apparatus, medium and equipment for the synergistic optimization of air spring and vehicle performance. Background Technology

[0002] Air springs are inherently highly nonlinear systems involving geometry, materials, and fluid-structure interaction. While high-precision finite element models can reflect their characteristics well, their computational cost is enormous, making them computationally infeasible for direct application in virtual proving ground (VPG) simulations and iterative optimizations involving the entire vehicle. Empirical formulas or simplified models, though fast, cannot accurately capture the dynamic response of air springs under real random road surface excitations, significantly compromising the reliability of optimization results.

[0003] In the current process, air spring design and vehicle performance development are often two relatively independent stages. This sequential "parts first, then vehicle" approach results in a lack of effective direct correlation and reverse optimization channels between the fine structural parameters of the air spring (such as piston profile and cord angle) and the comprehensive performance indicators of the entire vehicle (such as vehicle acceleration and tire dynamic load). The result is often that the component itself meets performance standards, but fails to contribute optimal performance to the overall vehicle system.

[0004] Faced with complex optimization problems involving multiple operating conditions and objectives, traditional gradient-based optimization methods are prone to getting trapped in local optima. While some intelligent optimization algorithms can perform global searches, they require too many calls to the simulation model, resulting in low efficiency. More importantly, existing methods lack learning and adaptive capabilities, failing to transform the large amounts of virtual test data generated by VPG into intelligent decision-making experience, and making it difficult to generate predictive control strategies that can proactively adapt to unknown future road conditions.

[0005] While structures such as adjustable air chamber volume offer potential for performance optimization, traditional control strategies (such as PID control) lag behind in handling random and transient conditions encountered during vehicle operation, failing to achieve optimal coordination of stiffness and damping across different scenarios. This results in the air spring's potential not being fully realized, leaving significant room for improvement in the vehicle's overall performance. Summary of the Invention

[0006] This application provides a method, apparatus, medium, and equipment for the coordinated optimization of air springs and vehicle performance, in order to solve the problems of existing coordinated optimization methods that are difficult to balance model accuracy and simulation efficiency, "island-style" design processes that can only result in local optima of system performance, insufficient intelligence and adaptive capabilities of optimization methods, and lack of real-time performance and foresight of control strategies.

[0007] The first aspect of this application provides a method for synergistic optimization of air springs and overall vehicle performance, comprising the following steps: The design parameters of the air spring are defined as optimization variables, and a parameterized air spring model is constructed based on the optimization variables. The parameterized air spring model is coupled with a pre-built virtual test field simulation environment to extract vehicle performance indicators and construct a sample database. The pre-built adaptive chaotic multinomial-Kriging model method is trained using the sample database to generate an initial surrogate model; The sample database is updated based on a preset point-addition criterion to obtain an updated sample database, and the updated sample database is used to update the initial proxy model to obtain an updated proxy model. Based on the preset objective function and preset constraints, a multi-objective genetic algorithm is used to perform optimization calculations on the updated surrogate model in order to search for the Pareto optimal solution set; The Pareto optimal solution set is input into a reinforcement learning agent pre-constructed using the virtual test field simulation environment to predict the control commands for the air spring. The control commands for the air spring are then verified. If the performance does not meet the standard, the update, optimization calculation, and prediction process is iteratively executed until the performance meets the standard.

[0008] Optionally, the sample database includes air spring design parameters and vehicle performance indicators, wherein the air spring design parameters include the geometric parameters, structural parameters and state parameters of the air spring, and the vehicle performance indicators include vertical stiffness, damping characteristics and natural frequency.

[0009] Optionally, the preset objective function is at least one of minimizing the vehicle body vertical acceleration and minimizing the tire dynamic load, and the preset constraint condition is at least one of the allowable range of suspension dynamic travel parameters and the allowable range of component stress parameters.

[0010] Optionally, the state vector in the reinforcement learning agent is real-time vehicle information, and the action vector is the control command for the air spring.

[0011] A second aspect of this application provides a device for synergistic optimization of air spring and overall vehicle performance, comprising: The system comprises the following modules: a construction module, which defines the design parameters of the air spring as optimization variables and constructs a parameterized air spring model based on these variables; a coupling and extraction module, which couples the parameterized air spring model with a pre-constructed virtual test field simulation environment to extract vehicle performance indicators and construct a sample database; a training module, which uses the sample database to train a pre-constructed adaptive chaotic multinomial-Kriging model method to generate an initial surrogate model; an update module, which updates the sample database based on a preset addition criterion to obtain an updated sample database, and uses the updated sample database to update the initial surrogate model to obtain an updated surrogate model; an optimization calculation module, which uses a multi-objective genetic algorithm based on a preset objective function and preset constraints to perform optimization calculations on the updated surrogate model to search for a Pareto optimal solution set; and a prediction and iteration module, which inputs the Pareto optimal solution set into a reinforcement learning agent pre-constructed using the virtual test field simulation environment to predict the control commands of the air spring and verifies the results of the control commands. If the performance does not meet the standards, the update, optimization calculation, and prediction processes are iteratively executed until the performance meets the standards.

[0012] Optionally, the sample database includes air spring design parameters and vehicle performance indicators, wherein the air spring design parameters include the geometric parameters, structural parameters and state parameters of the air spring, and the vehicle performance indicators include vertical stiffness, damping characteristics and natural frequency.

[0013] Optionally, the preset objective function is at least one of minimizing the vehicle body vertical acceleration and minimizing the tire dynamic load, and the preset constraint condition is at least one of the allowable range of suspension dynamic travel parameters and the allowable range of component stress parameters.

[0014] Optionally, the state vector in the reinforcement learning agent is real-time vehicle information, and the action vector is the control command for the air spring.

[0015] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for synergistic optimization of air spring and vehicle performance as described in the above embodiments.

[0016] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for co-optimizing air spring and overall vehicle performance.

[0017] The proposed methods, apparatus, media, and equipment for the coordinated optimization of air springs and vehicle performance in this application establish a direct mapping from the fine parameters of the air springs to the overall vehicle performance. This allows the multi-objective optimization algorithm to find the optimal solution within the entire design space, avoiding local optima caused by traditional serial design, and thus finding the Pareto optimal solution at the system level. The introduced adaptive surrogate model replaces the time-consuming VPG simulation with mathematical approximation, enabling massive and rapid optimization iterations. This shortens the "design-prototype-test" cycle, which originally required weeks or even months, to within a few days, greatly reducing R&D costs and time. By using the massive and diverse virtual environment data generated by VPG to train the reinforcement learning agent, the resulting control strategy possesses predictive and adaptive capabilities to cope with complex and unknown road conditions, which is impossible with traditional rule-based control methods, significantly improving the intelligent driving experience of the vehicle. A closed-loop verification system is constructed to ensure the effectiveness and reliability of the virtual optimization results in the real world, forming a complete technology chain from virtual to physical, and improving the success rate of technology transfer.

[0018] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0019] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating a method for synergistic optimization of air spring and overall vehicle performance according to an embodiment of this application. Figure 2 This is a schematic diagram illustrating the specific implementation of a method for synergistic optimization of air spring and vehicle performance according to an embodiment of this application; Figure 3 This is a schematic diagram illustrating the specific implementation of a point-addition criterion provided in an embodiment of this application; Figure 4 This is a block diagram of a device for synergistic optimization of air spring and vehicle performance according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application.

[0020] Explanation of reference numerals in the attached figures: 40-Coordinated optimization device for air spring and vehicle performance, 401-Construction module, 402-Coupling and extraction module, 403-Training module, 404-Update module, 405-Optimization calculation module, 406-Prediction and iteration module, 501-Memory, 502-Processor, 503-Communication interface. Detailed Implementation

[0021] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0022] The following description, with reference to the accompanying drawings, outlines a method, apparatus, medium, and device for the coordinated optimization of air springs and vehicle performance according to embodiments of this application. Regarding the traditional methods mentioned in the background section, there is a dilemma in air spring modeling: while using empirical formulas or simplified mathematical models (such as 1 / 4 or 1 / 2 vehicle models) offers fast computation, it cannot accurately reflect the true nonlinear dynamic stiffness and fluid-structure interaction characteristics of the air spring, leading to significant deviations between the optimization results and actual performance. On the other hand, while using high-fidelity finite element models can improve accuracy, their computational cost is enormous and time-consuming, making them unsuitable for vehicle-level optimization processes requiring massive iterations. Traditional design processes typically follow a sequential "components first, then vehicle" approach, where air springs... The air springs are first designed and optimized independently, and then verified in a vehicle model. This process isolates the fine structural parameters of the air springs (such as piston profile and effective area curve) from the comprehensive performance indicators of the entire vehicle (such as vehicle acceleration and tire dynamic load), lacking a closed-loop mechanism that can optimize the key design parameters of the air springs in reverse from the perspective of the overall system performance goals. Existing optimization methods (such as traditional particle swarm optimization) often require a large number of simulation models when facing complex optimization problems with multiple operating conditions and multiple objectives, resulting in low efficiency and a tendency to get trapped in local optima. More importantly, these... The current method lacks learning capabilities, failing to transform the large amount of virtual test data generated by the VPG into intelligent decision-making experience, and making it difficult to generate predictive control strategies that can proactively adapt to unknown future road conditions. For structurally complex air springs (such as multi-chamber air springs), their potential largely depends on the control strategy. Existing control methods (such as PID control or control based on preset lookup tables) are mostly passive responses, adjusting according to the current vehicle body state, which has a lag and makes it difficult to maintain the optimal balance between comfort and handling under complex and ever-changing real road conditions. This application provides a collaborative optimization method for air springs and overall vehicle performance. In this method, the vehicle virtual test track (VPG) is transformed from a traditional, late-stage "performance verification tool" into a "performance exploration and data engine" for drive design. The VPG simulates thousands of operating conditions to generate a high-quality dataset for training AI models. Then, an adaptive surrogate model is used as a bridge to transform the high-fidelity but time-consuming physical simulation into a "digital sandbox" that can be explored billions of times quickly, making it possible to use complex multi-objective optimization algorithms and reinforcement learning to find the global optimum. In addition, a clear and reproducible technical path is provided: high-precision parametric modeling: key design parameters such as the piston profile and additional air chamber volume of the air spring are parametrically defined, making them variables that can be directly optimized by the algorithm; construction and application of adaptive surrogate model: methods such as adaptive chaotic polynomial-Kriging model are adopted.This model not only approximates the input-output relationship but also intelligently guides the next simulation calculation position through criteria such as expected improvement (EI), maximizing the model's accuracy in key areas. This is the technological hub for achieving efficient optimization. Multi-objective optimization for overall vehicle performance: The optimization objective is directly set as a comprehensive performance index at the vehicle level (such as ride comfort and handling), and a multi-objective genetic algorithm (such as NSGA-II) is used to find the optimal solution on the surrogate model, ultimately obtaining a series of Pareto optimal solutions, providing designers with multiple performance trade-offs. Data-driven predictive control: Furthermore, diverse driving scenario data generated by VPG is used to train a reinforcement learning agent, enabling it to learn a predictive control strategy. This strategy can predictively adjust the air spring height or stiffness based on real-time vehicle status (such as vehicle speed and acceleration), achieving a leap from "passive response" to "active adaptation."

[0023] Specifically, Figure 1 This is a flowchart illustrating a method for synergistic optimization of air spring and overall vehicle performance provided in an embodiment of this application.

[0024] like Figure 1 As shown, the method for synergistic optimization of air spring and overall vehicle performance includes the following steps: In step S101, the design parameters of the air spring are defined as optimization variables, and a parameterized air spring model is constructed based on the optimization variables.

[0025] In step S102, the parameterized air spring model is coupled with a pre-built virtual test field simulation environment to extract vehicle performance indicators and build a sample database.

[0026] In practice, this step forms the basis for all subsequent optimization efforts, with the goal of creating a precise virtual prototype that can automatically adjust its parameters.

[0027] The process of establishing a parametric air spring model is as follows: Input parameter determination: The key design parameters of the air spring are defined as optimization variables. The optimization variables mainly include geometric parameters (such as the piston profile curve and the volume of the additional air chamber), structural parameters (such as the area of ​​the chamber connecting valve port for a multi-chamber air spring), and material parameters (such as the layup angle of the airbag cord).

[0028] High-fidelity modeling: In finite element software (such as Abaqus or ANSYS), establish a finite element model that can accurately reflect the nonlinear stiffness and fluid-structure interaction characteristics of the air spring. This requires precise definition of the hyperelastic constitutive model of the rubber material, the reinforcement model of the cord layer, and the equation of state of the gas inside the airbag.

[0029] Parametric implementation: Secondary development is performed using the scripting capabilities of finite element software (such as Abaqus / Python) to correlate the input parameters with the model geometry, generating a parametric model. Modifying the input parameters automatically generates a new finite element model and submits it for calculation.

[0030] Furthermore, the parameterized vehicle virtual test track (VPG) simulation environment is integrated, and the specific process is as follows: Model integration: Integrate the parameterized air spring model into the multibody dynamics model of the whole vehicle (such as building it in ADAMS / Car). The whole vehicle model should include subsystems such as body, suspension guiding mechanism, shock absorber, and tires.

[0031] VPG Environment Setup: In VPG software, a high-precision 3D digital road surface (such as Belgian roads or washboard roads) is constructed, and an F-tire tire model capable of accurately simulating high-frequency dynamic characteristics is used. This is crucial for accurate load extraction.

[0032] Furthermore, through automated batch simulation, a massive sample library is built for training the agent model. The specific process is as follows: Experimental Design: The Latin hypercube experimental design method is adopted. Within the air spring design variable space determined in step S101, N groups (e.g., 500-1000 groups) of sample points with good spatial homogeneity and representativeness are strategically selected. This method can explore the design space to the maximum extent with a small sample size.

[0033] Automated batch simulation: Write an automated script (such as using Python or MATLAB) to couple the parameterized model in step S101 with the VPG simulation environment (such as ADAMS). The script can automatically read the parameters of the sample points, update the model, submit the calculation, and extract the results.

[0034] For each set of sample points (i.e. a specific combination of air spring parameters), the VPG vehicle model is automatically driven to perform simulations under various typical working conditions (such as different road surfaces and different vehicle speeds).

[0035] Performance data extraction and storage: After each simulation, key vehicle performance indicators are automatically extracted, mainly including the root mean square value of vertical acceleration at the vehicle's center of gravity (measuring ride comfort), the root mean square value of tire dynamic load (measuring handling safety), and suspension dynamic travel. All sample points (inputs) and their corresponding performance indicator results (outputs) are stored in a structured sample database (such as CSV or HDF5 format).

[0036] In step S103, the pre-built adaptive chaotic multinomial-Kriging model method is trained using a sample database to generate an initial surrogate model.

[0037] In actual implementation, such as Figure 2 As shown, to address the high computational cost of VPG simulation, step S103 constructs a computationally efficient surrogate model to replace the original simulation. This is crucial for achieving rapid optimization. The specific construction process is as follows: Initial surrogate model training: The adaptive chaotic multinomial-Kriging model method is trained based on the sample database generated by S102 to construct an initial surrogate model, which can approximate the complex nonlinear relationship between air spring design parameters and vehicle performance indicators through mathematical expressions.

[0038] The sample database includes air spring design parameters (i.e., input values) and vehicle performance indicators (i.e., output values). The air spring design parameters include geometric parameters, structural parameters, and state parameters, while the vehicle performance indicators include vertical stiffness, damping characteristics, and natural frequency.

[0039] It should be noted that the geometric parameters, including the piston profile and the additional air chamber volume, determine the effective pressure-bearing area and the total air chamber volume of the air spring, directly affecting the stiffness and natural frequency.

[0040] Structural parameters, including cord angle, number of cord layers, and cord spacing, constitute the "skeleton" of the airbag and greatly affect its stiffness and strength. For example, changes in cord angle can significantly affect vertical stiffness.

[0041] State parameters, including initial inflation pressure and operating height, determine the initial operating point of the air spring, affecting the vehicle's attitude and basic performance.

[0042] Vertical stiffness is a quantification of the air spring's ability to resist compression or extension, and is directly related to the vehicle's ride comfort (smoothness).

[0043] Damping characteristics work in conjunction with shock absorbers to dissipate vibration energy, affecting vehicle stability and handling.

[0044] The natural frequency is the inherent vibration characteristic of the suspension system and is a core indicator for evaluating ride comfort; the goal is usually to pursue a low frequency.

[0045] In step S104, the sample database is updated based on the preset addition criteria to obtain the updated sample database, and the initial proxy model is updated using the updated sample database to obtain the updated proxy model.

[0046] It should be noted that the addition criteria (such as the expected improvement criterion) are used to determine at which new sample points the simulation can most effectively improve the accuracy of the surrogate model, especially in the region of possible optimal solutions.

[0047] In actual implementation, it is first necessary to determine the number of initial sample points in order to strike a balance between model accuracy and computational cost. This is to avoid the situation where too few sample points may fail to capture the complex nonlinear relationship between air spring parameters and vehicle performance, while too many sample points will cause a sharp increase in VPG simulation costs.

[0048] A common starting point is to use 10d to 15d (where d is the number of design variables) as the initial sample size. For example, if the optimization variables include five key parameters such as piston profile control points (3 points) and additional air chamber volume (1 point), then the initial sample points can be selected between 50 and 75.

[0049] Regardless of the number, it is essential to ensure that these sample points are evenly distributed within the design space to maximize the exploration of the parameter space. Therefore, this application employs Latin hypercube sampling (LHS), a classic and efficient method for achieving this goal. It guarantees that there is only one sample point in each stratification interval for each design variable, thereby avoiding sample clustering.

[0050] Furthermore, after the initial proxy model is built, its accuracy may still be unsatisfactory globally. Therefore, the core idea of ​​the adaptive point addition criterion in this application is to intelligently and purposefully select new sample points that can most effectively improve the model accuracy for VPG simulation, and add the new data to the training set to update the model. A commonly used and efficient algorithm is the Expected Improvement (EI) criterion.

[0051] It's important to explain how the EI criterion works: the EI criterion balances "exploration" and "exploitation." It tends to select points that are either located in regions of high current prediction uncertainty (exploring unknown territory) or have the potential to further optimize the objective function (utilizing information near the current optimal solution).

[0052] The formula for calculating the EI value can be expressed as: EI(x) = (f_min - μ(x)) * Φ(Z) + σ(x) * φ(Z) Where f_min is the optimal objective function value in the current sample set, μ(x) is the prediction mean of the surrogate model at point x, σ(x) is the prediction standard deviation (i.e. uncertainty) of the surrogate model at point x, Φ(Z) and φ(Z) represent the cumulative distribution function and probability density function of the standard normal distribution, respectively, and Z = (f_min - μ(x)) / σ(x).

[0053] In regions where the surrogate model predicts high uncertainty or is close to the optimal solution, new sample points are automatically added, and the VPG simulation in step S101 is invoked to obtain the actual performance data of that point.

[0054] New sample points are added to the database, and the surrogate model is retrained and updated. This process is repeated until the prediction accuracy of the surrogate model (e.g., coefficient of determination R² > 0.95) meets the preset requirements. This method can obtain a high-accuracy surrogate model with the fewest simulations.

[0055] Specifically, such as Figure 3 As shown, it is determined whether the initial surrogate model meets the model accuracy requirements. If it does not, the EI value of all candidate points is calculated according to the EI value calculation formula, and the point with the largest EI value is selected as the new sample point. VPG simulation is run to obtain the true response of the point, thereby updating the original sample database to obtain the updated sample database. The updated sample database is then used to update the initial surrogate model to obtain the updated surrogate model. The accuracy of the updated surrogate model is then judged. If it still does not meet the accuracy requirements, the aforementioned calculation, running, and updating process is iterated until the accuracy requirements are met, and the final surrogate model is obtained.

[0056] In step S105, based on the preset objective function and preset constraints, a multi-objective genetic algorithm is used to perform optimization calculations on the updated surrogate model in order to search for the Pareto optimal solution set.

[0057] In some embodiments, the preset objective function is at least one of minimizing the vehicle body vertical acceleration and minimizing the tire dynamic load, and the preset constraint is at least one of the allowable range of suspension dynamic travel parameters and the allowable range of component stress parameters.

[0058] In actual implementation, this embodiment of the application performs a global optimization search based on a highly efficient and accurate proxy model. The specific execution process is as follows: Multi-objective optimization problem formulation: Design variables: namely, the air spring parameters defined in step S101 (such as piston profile control point coordinates, additional air chamber volume, etc.).

[0059] Objective function: Typically, it includes multiple objectives such as minimizing vehicle vertical acceleration (comfort) and minimizing tire dynamic load (handling). These objectives are often in conflict with each other.

[0060] Constraints: Set the allowable range for parameters such as suspension dynamic travel and component stress.

[0061] Intelligent optimization algorithm execution: A multi-objective genetic algorithm (such as NSGA-II) is used to perform optimization calculations on the surrogate model. Because the surrogate model is computationally extremely fast, the algorithm can perform massive evaluations and iterations (e.g., tens of thousands of times), thereby efficiently searching for a Pareto optimal solution set within the design space. This solution set provides multiple trade-offs between comfort and operability for designers to decide.

[0062] In step S106, the Pareto optimal solution set is input into the reinforcement learning agent pre-constructed in a virtual test field simulation environment to predict the control command of the air spring, and the control command of the air spring is verified. If the performance does not meet the standard, the update, optimization calculation and prediction process is iteratively executed until the performance meets the standard.

[0063] In some embodiments, the state vector in the reinforcement learning agent is real-time vehicle information, and the action vector is the control command for the air spring.

[0064] In actual implementation, for active / semi-active systems such as multi-cavity air springs, this application further optimizes their control strategy, the specific process of which is as follows: First, construct a reinforcement learning environment: encapsulate the simulation environment based on the VPG model into an interactive environment for a reinforcement learning agent (such as using Deep Q-Network or PPO algorithm).

[0065] Training reinforcement learning agents: The state vector is defined to contain real-time vehicle information, such as vehicle speed, vehicle acceleration, and steering wheel angle.

[0066] The motion vector is the control command for the air spring, such as switching chambers or adjusting valve ports.

[0067] The reward function is designed based on vehicle performance indicators (such as negative vehicle acceleration) to encourage the agent to learn behaviors that improve overall performance.

[0068] Policy learning and validation involves the agent learning a predictive control policy through interaction with the environment. Therefore, by inputting the Pareto optimal solution set into a reinforcement learning agent pre-constructed in a virtual testbed simulation environment, the policy can predict the optimal action in the near future based on the current state, issuing control commands for the air spring, thus achieving forward-looking control and further improving performance.

[0069] Finally, the embodiments of this application also rigorously verify the optimization results to form a closed loop, and the specific process is as follows: Virtual verification: The optimized air spring design parameters (or control strategy) are substituted into the original high-fidelity VPG model for simulation. The vehicle performance curves before and after optimization are compared to quantitatively evaluate the performance improvement.

[0070] Physical verification: The parameters determined by virtual optimization are applied to the manufacture of physical prototypes. On a hydraulic servo test bench, the load spectrum extracted from VPG is input to conduct durability tests on the air spring assembly; or it is installed in a vehicle for actual road testing to ultimately verify the effectiveness and superiority of this patented method.

[0071] In summary, the method for synergistic optimization of air springs and overall vehicle performance proposed in the embodiments of this application has the following beneficial effects: (1) This application sets the optimization target directly as the comprehensive performance index of the whole vehicle (such as smoothness and handling stability), and automatically optimizes the key parameters of the air spring through intelligent algorithms. This ensures that the optimization result is the global optimal solution from the perspective of the whole vehicle system, rather than the local optimal solution that compromises the performance of the parts and the whole vehicle, thereby significantly improving the overall driving quality of the vehicle.

[0072] (2) By combining high-precision VPG simulation with efficient intelligent optimization algorithms (such as agent model-based optimization), this application can complete a large number of "test-optimization" iterations in a virtual environment before physical prototype manufacturing. This greatly shortens the R&D cycle, reduces the reliance on expensive and time-consuming physical prototypes and road tests, and thus significantly reduces development costs.

[0073] (3) The method of this application is particularly suitable for optimizing advanced structures such as dual-chamber, triple-chamber or volume-adjustable air springs. It can accurately optimize their stiffness and damping characteristics under various complex working conditions and generate intelligent predictive control strategies, thereby giving full play to the performance potential of advanced air spring structures and providing core technical support for high-end vehicle suspension systems.

[0074] (4) By using the extreme working condition data generated by VPG to train reinforcement learning and other AI agents, this application can generate a predictive control strategy. This strategy can predict the real-time road conditions or driving intentions and adjust the state of the air springs in advance to achieve seamless, smooth and active switching of performance modes, fundamentally improving the vehicle's adaptability to different environments and driving needs, and bringing users a more comfortable and safer intelligent driving experience.

[0075] (5) This application constructs a complete digital closed-loop process from “virtual design-virtual experiment-intelligent optimization-decision support”, which not only reduces the excessive reliance of R&D on personal experience, but also makes the R&D process more standardized, traceable and intelligent, laying a solid foundation for the future realization of fully autonomous vehicle performance development.

[0076] (6) It can accurately optimize the combination of complex parameters such as the connection strategy of different chambers and piston profile, so as to fully explore the performance potential of these advanced structures. It is suitable for optimizing advanced structures with multiple controllable degrees of freedom, such as dual-chamber and triple-chamber air springs.

[0077] Next, referring to the accompanying drawings, a device for synergistic optimization of air spring and vehicle performance according to an embodiment of this application is described.

[0078] Figure 4 This is a block diagram illustrating a device for synergistic optimization of air spring and overall vehicle performance, as provided in an embodiment of this application.

[0079] like Figure 4 As shown, the air spring and vehicle performance co-optimization device 40 includes: a construction module 401, a coupling and extraction module 402, a training module 403, an update module 404, an optimization calculation module 405, and a prediction and iteration module 406.

[0080] The system comprises the following modules: Module 401 defines the design parameters of the air spring as optimization variables and constructs a parameterized air spring model based on these variables. Module 402 couples the parameterized air spring model with a pre-constructed virtual test track simulation environment to extract vehicle performance indicators and build a sample database. Module 403 trains a pre-constructed adaptive chaotic multinomial-Kriging model using the sample database to generate an initial surrogate model. Module 404 updates the sample database based on a preset addition criterion to obtain an updated sample database, and then uses this updated sample database to update the initial surrogate model, resulting in an updated surrogate model. Module 405 performs optimization calculations on the updated surrogate model using a multi-objective genetic algorithm based on a preset objective function and preset constraints to search for a Pareto optimal solution set. The prediction and iteration module 406 is used to input the Pareto optimal solution set into the reinforcement learning agent that is pre-built in a virtual test field simulation environment to predict the control command of the air spring and verify the result of the control command of the air spring. If the performance does not meet the standard, the update, optimization calculation and prediction process is iteratively executed until the performance meets the standard.

[0081] In some embodiments, the sample database includes air spring design parameters and vehicle performance indicators, wherein the air spring design parameters include the geometric parameters, structural parameters and state parameters of the air spring, and the vehicle performance indicators include vertical stiffness, damping characteristics and natural frequency.

[0082] In some embodiments, the preset objective function is at least one of minimizing the vehicle body vertical acceleration and minimizing the tire dynamic load, and the preset constraint is at least one of the allowable range of suspension dynamic travel parameters and the allowable range of component stress parameters.

[0083] In some embodiments, the state vector in the reinforcement learning agent is real-time vehicle information, and the action vector is the control command for the air spring.

[0084] It should be noted that the explanation of the aforementioned embodiment of the method for synergistic optimization of air spring and vehicle performance also applies to the synergistic optimization device for air spring and vehicle performance in this embodiment, and will not be repeated here.

[0085] The air spring and vehicle performance synergistic optimization device proposed according to the embodiments of this application has the following beneficial effects: (1) This application sets the optimization target directly as the comprehensive performance index of the whole vehicle (such as smoothness and handling stability), and automatically optimizes the key parameters of the air spring through intelligent algorithms. This ensures that the optimization result is the global optimal solution from the perspective of the whole vehicle system, rather than the local optimal solution that compromises the performance of the parts and the whole vehicle, thereby significantly improving the overall driving quality of the vehicle.

[0086] (2) By combining high-precision VPG simulation with efficient intelligent optimization algorithms (such as agent model-based optimization), this application can complete a large number of "test-optimization" iterations in a virtual environment before physical prototype manufacturing. This greatly shortens the R&D cycle, reduces the reliance on expensive and time-consuming physical prototypes and road tests, and thus significantly reduces development costs.

[0087] (3) The method of this application is particularly suitable for optimizing advanced structures such as dual-chamber, triple-chamber or volume-adjustable air springs. It can accurately optimize their stiffness and damping characteristics under various complex working conditions and generate intelligent predictive control strategies, thereby giving full play to the performance potential of advanced air spring structures and providing core technical support for high-end vehicle suspension systems.

[0088] (4) By using the extreme working condition data generated by VPG to train reinforcement learning and other AI agents, this application can generate a predictive control strategy. This strategy can predict the real-time road conditions or driving intentions and adjust the state of the air springs in advance to achieve seamless, smooth and active switching of performance modes, fundamentally improving the vehicle's adaptability to different environments and driving needs, and bringing users a more comfortable and safer intelligent driving experience.

[0089] (5) This application constructs a complete digital closed-loop process from “virtual design-virtual experiment-intelligent optimization-decision support”, which not only reduces the excessive reliance of R&D on personal experience, but also makes the R&D process more standardized, traceable and intelligent, laying a solid foundation for the future realization of fully autonomous vehicle performance development.

[0090] (6) It can accurately optimize the combination of complex parameters such as the connection strategy of different chambers and piston profile, so as to fully explore the performance potential of these advanced structures. It is suitable for optimizing advanced structures with multiple controllable degrees of freedom, such as dual-chamber and triple-chamber air springs.

[0091] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0092] The electronic device may include: a memory 501, a processor 502, and a computer program stored on the memory 501 and capable of running on the processor 502.

[0093] When the processor 502 executes the program, it implements the method for co-optimizing air spring and vehicle performance provided in the above embodiments.

[0094] Furthermore, electronic devices also include: Communication interface 503 is used for communication between memory 501 and processor 502.

[0095] The memory 501 is used to store computer programs that can run on the processor 502.

[0096] Memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0097] If the memory 501, processor 502, and communication interface 503 are implemented independently, then the communication interface 503, memory 501, and processor 502 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0098] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0099] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0100] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for co-optimizing air spring and overall vehicle performance.

[0101] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0102] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0103] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0104] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0105] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0106] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0107] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0108] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for synergistic optimization of air spring and overall vehicle performance, characterized in that, Includes the following steps: The design parameters of the air spring are defined as optimization variables, and a parameterized air spring model is constructed based on the optimization variables. The parameterized air spring model is coupled with a pre-built virtual test field simulation environment to extract vehicle performance indicators and construct a sample database. The pre-built adaptive chaotic multinomial-Kriging model method is trained using the sample database to generate an initial surrogate model; The sample database is updated based on a preset point-addition criterion to obtain an updated sample database, and the updated sample database is used to update the initial proxy model to obtain an updated proxy model. Based on the preset objective function and preset constraints, a multi-objective genetic algorithm is used to perform optimization calculations on the updated surrogate model in order to search for the Pareto optimal solution set; The Pareto optimal solution set is input into a reinforcement learning agent pre-constructed using the virtual test field simulation environment to predict the control commands for the air spring. The control commands for the air spring are then verified. If the performance does not meet the standard, the update, optimization calculation, and prediction process is iteratively executed until the performance meets the standard.

2. The method for synergistic optimization of air spring and vehicle performance according to claim 1, characterized in that, The sample database includes air spring design parameters and vehicle performance indicators. The air spring design parameters include the geometric parameters, structural parameters, and state parameters of the air spring. The vehicle performance indicators include vertical stiffness, damping characteristics, and natural frequency.

3. The method for synergistic optimization of air spring and overall vehicle performance according to claim 1, characterized in that, The preset objective function is at least one of minimizing the vehicle body vertical acceleration and minimizing the tire dynamic load, and the preset constraint condition is at least one of the allowable range of suspension dynamic travel parameters and the allowable range of component stress parameters.

4. The method for synergistic optimization of air spring and overall vehicle performance according to claim 1, characterized in that, The state vector in the reinforcement learning agent is the real-time information of the vehicle, and the action vector is the control command for the air spring.

5. A device for synergistic optimization of air spring and overall vehicle performance, characterized in that, include: A construction module is used to define the design parameters of the air spring as optimization variables and to construct a parameterized air spring model based on the optimization variables; The coupling and extraction module is used to couple the parameterized air spring model with the pre-built virtual test field simulation environment to extract vehicle performance indicators and build a sample database. The training module is used to train a pre-built adaptive chaotic multinomial-Kriging model method using the sample database to generate an initial surrogate model. The update module is used to update the sample database based on a preset addition criterion to obtain an updated sample database, and to use the updated sample database to update the initial proxy model to obtain an updated proxy model. The optimization calculation module is used to perform optimization calculations on the updated surrogate model based on a preset objective function and preset constraints, using a multi-objective genetic algorithm to search for a Pareto optimal solution set. The prediction and iteration module is used to input the Pareto optimal solution set into a reinforcement learning agent pre-constructed in the virtual test field simulation environment to predict the control command of the air spring and verify the result of the control command of the air spring. If the performance does not meet the standard, the update, optimization calculation and prediction process is iteratively executed until the performance meets the standard.

6. The air spring and vehicle performance synergistic optimization device according to claim 5, characterized in that, The sample database includes air spring design parameters and vehicle performance indicators. The air spring design parameters include the geometric parameters, structural parameters, and state parameters of the air spring. The vehicle performance indicators include vertical stiffness, damping characteristics, and natural frequency.

7. The air spring and vehicle performance synergistic optimization device according to claim 5, characterized in that, The preset objective function is at least one of minimizing the vehicle body vertical acceleration and minimizing the tire dynamic load, and the preset constraint condition is at least one of the allowable range of suspension dynamic travel parameters and the allowable range of component stress parameters.

8. The air spring and vehicle performance synergistic optimization device according to claim 5, characterized in that, The state vector in the reinforcement learning agent is the real-time information of the vehicle, and the action vector is the control command for the air spring.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the method for synergistic optimization of air springs and vehicle performance as described in any one of claims 1-4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the method for co-optimizing air springs and overall vehicle performance as described in any one of claims 1-4.

Images