A Dynamic Optimization Method and System for Aerodynamic Resistance of Slack Trucks Based on Airflow Simulation

Abstract

This invention discloses a dynamic optimization method and system for the wind resistance of a stake vehicle based on airflow simulation, belonging to the field of automotive aerodynamic optimization. The method involves constructing a three-dimensional structural model of the stake vehicle and extracting the spatial distribution parameters and surface roughness parameters of the fence components; establishing a multi-condition external flow field simulation model to obtain the distribution map of the wind resistance coefficient per unit area; constructing a wind resistance sensitivity matrix to identify highly sensitive components; performing disturbance modeling on the highly sensitive components to generate local deformation configuration samples and establish a simulation model set; performing parallel simulation calculations to extract wind resistance coefficient and aerodynamic stress data; determining the optimal disturbance configuration based on a multi-objective optimization algorithm; and outputting structural adjustment suggestions to guide dynamic design optimization. This method can achieve synergistic optimization of wind resistance reduction and structural safety, improving the overall aerodynamic performance and energy efficiency of the stake vehicle.

Description

Technical Field

[0001] This invention relates to the field of automotive aerodynamic optimization, specifically to a method and system for dynamic optimization of wind resistance in cargo vans based on airflow simulation. Background Technology

[0002] Stake trucks are a type of freight vehicle widely used in logistics transportation. Their structural feature is a fence-like side panel, making them suitable for transporting agricultural products, livestock, building materials, and other goods requiring ventilation or irregular stacking. However, because the fence structure of stake trucks is not completely enclosed, airflow penetration is irregular during operation, resulting in severe turbulence. This leads to a higher drag coefficient than box trucks, higher fuel consumption, poorer driving stability, and nonlinear aerodynamic loads on the vehicle's local structure.

[0003] Currently, research on wind resistance optimization for cargo vans mainly focuses on optimizing the average wind resistance of a fixed vehicle shape. This is insufficient to address the dynamic changes in wind resistance under different airflow conditions during high-speed driving. Especially in mountainous areas with long downhill slopes or crosswinds, turbulent wake vortices can easily form in local areas of the fence structure, leading to vehicle swaying, sudden increases in fuel consumption, local structural fatigue, and even cabin vibration and instability.

[0004] In addition, existing technologies generally lack a dynamic evaluation mechanism for wind resistance that combines real-time airflow field simulation with the linkage of vehicle structural parameters. This makes it impossible to identify and optimize weak points in the aerodynamic performance of the cargo box structure during the design phase, resulting in high design redundancy and difficulty in reducing wind resistance without sacrificing cargo capacity. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic optimization method and system for wind resistance of cargo trucks based on airflow simulation, so as to overcome the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a dynamic optimization method for wind resistance of a cargo box truck based on airflow simulation, comprising:

[0007] Construct a three-dimensional structural model of the target cargo truck and obtain the spatial distribution parameters and surface roughness parameters of its fence components;

[0008] Establish a simulation model of the external flow field of the corresponding vehicle to simulate the airflow state under various typical driving conditions and obtain the distribution map of the wind resistance coefficient per unit area acting on each fence component.

[0009] Based on the obtained distribution map of wind resistance coefficient per unit area, a wind resistance sensitivity matrix of fence components is constructed, and the set of component numbers in the high wind resistance sensitive area is extracted.

[0010] Structural perturbation modeling is performed on each component in the component number set to generate multiple local deformation configuration samples, and a set of perturbed aerodynamic simulation models is constructed.

[0011] Parallel simulation calculations are performed on the aerodynamic simulation model set to obtain the total drag coefficient and local aerodynamic stress distribution of each disturbance sample under each working condition, forming an optimization evaluation function set.

[0012] Multi-objective optimization is performed based on the set of optimization evaluation functions to determine the optimal perturbation configuration. The set of component numbers corresponding to the optimal perturbation configuration satisfies that the decrease in total drag coefficient is greater than a preset threshold and the local aerodynamic stress is within the allowable range.

[0013] Based on the optimal perturbation configuration, structural adjustment suggestions are output to guide the dynamic design optimization of the fence components of the cargo box vehicle.

[0014] Preferably, the three-dimensional structural model of the target cargo box vehicle includes:

[0015] High-precision point cloud data of the vehicle's external structure is obtained by laser point cloud scanning of the actual cargo truck.

[0016] Point cloud reconstruction algorithms are used to denoise, register, and fit surfaces to point cloud data, generating a three-dimensional digital geometric model containing detailed components.

[0017] The fence components in the 3D model are numbered by region, and the spatial position, length, height and thickness information of each component are extracted to form a spatial distribution parameter matrix;

[0018] By acquiring surface texture images and analyzing grayscale texture distribution, combined with a standard roughness modeling library, the surface roughness level of each fence component is estimated and assigned.

[0019] Preferably, establishing the corresponding vehicle external flow field simulation model includes:

[0020] Based on a three-dimensional digital geometric model, a computational fluid dynamics simulation domain containing a vehicle model and an external air domain is constructed, and boundary conditions and inlet velocity parameters are set.

[0021] Various typical driving conditions were selected, including high-speed straight driving, crosswind interference driving, and long downhill coasting operation scenarios, and simulation model groups for the corresponding conditions were constructed respectively.

[0022] Steady-state or transient simulations were performed on each set of working conditions using large eddy simulation or Reynolds time-averaged equation solution methods to obtain the wind resistance distribution per unit area on the surface of the fence components.

[0023] The simulation output is projected onto the surface mesh of each component, the ratio of wind resistance to area is calculated, and a distribution map of wind resistance coefficient per unit area is generated.

[0024] Preferably, the wind resistance sensitivity matrix of the constructed fence component includes:

[0025] The distribution map of wind resistance coefficient per unit area is clustered and integrated according to component number. The average wind resistance coefficient value of each fence component is calculated, and an initial vector of wind resistance coefficient is constructed.

[0026] The component size factor and surface roughness level are used as correction weights to normalize the initial vector of the drag coefficient, generating a standardized drag response vector.

[0027] Based on multi-condition wind resistance response data, the variance and maximum gradient value of wind resistance change of each fence component under different conditions are calculated as wind resistance sensitivity index, and a wind resistance sensitivity matrix is ​​constructed.

[0028] Set a preset sensitivity threshold, and filter the component numbers in the sensitivity matrix whose corresponding values ​​are greater than the sensitivity threshold to form a set of component numbers in the high wind resistance sensitive area.

[0029] Preferably, the structural perturbation modeling for each component in the component number set includes:

[0030] Based on the spatial geometric information and boundary constraints of each component in the component number set, a set of disturbance parameters is set, including disturbance amplitude, disturbance direction and disturbance type;

[0031] Different combinations of perturbation parameters are applied to each component, and multiple local deformation configuration samples are generated by shape parameter transformation. The deformation types include deflection, torsion and thickness deformation.

[0032] Each local deformation configuration is embedded into the original three-dimensional model of the vehicle to construct the disturbed geometric model of the vehicle, while keeping the structure of the remaining components unchanged;

[0033] For all perturbation configuration models, a set of corresponding computational fluid dynamics simulation models is generated in batches, and the simulation boundary conditions are uniformly set as the data input set for solving the optimization evaluation function.

[0034] Preferably, parallel simulation calculations of the aerodynamic simulation model set include:

[0035] For each disturbance configuration model, a unified set of multi-condition simulation parameters is set, including flow velocity, inflow angle, ground sliding conditions and turbulence model, and a condition configuration file is constructed.

[0036] The perturbation configuration model set is combined with the corresponding operating condition configuration in batches, and parallel simulation is performed using a high-performance computing platform or cloud computing environment.

[0037] After the simulation, the total drag coefficient and aerodynamic load distribution on the surface of the fence components for each disturbance configuration under each working condition are extracted and stored in vector form.

[0038] Based on the changes in drag coefficient and local aerodynamic stress of components under different working conditions, a drag optimization objective function and a stress constraint function are constructed to form a multi-objective optimization evaluation function set.

[0039] Preferably, the multi-objective optimization based on the set of optimization evaluation functions includes:

[0040] Based on the perturbation configuration sample set and the corresponding wind resistance objective function and stress constraint function, the design variable space and objective function space are constructed, and the total wind resistance coefficient reduction threshold T1 and the maximum allowable stress threshold T2 are defined.

[0041] A multi-objective evolutionary algorithm is used to iteratively optimize and search the perturbation configuration samples, resulting in a Pareto optimal solution set that balances wind resistance performance and aerodynamic strength.

[0042] In the Pareto optimal solution set, a solution set is selected where the wind resistance decreases by more than the threshold T1 and the local stress of all disturbing components does not exceed the threshold T2.

[0043] The set of component numbers involved in the optimal disturbance configuration that meets the screening criteria is denoted as M′, and the wind resistance optimization adjustment suggestions and component deformation parameters corresponding to the configuration are output.

[0044] Preferably, the output structure adjustment suggestions based on the optimal perturbation configuration include:

[0045] The perturbation parameters of each component in the optimal perturbation configuration are analyzed, including perturbation amplitude, perturbation direction and perturbation type, and the difference analysis is performed with the original component geometric parameters.

[0046] Based on the difference analysis results, a structural adjustment mapping table is constructed to clarify the deformation operation type and numerical range of each disturbed component;

[0047] Based on the spatial positioning relationship of components in the whole vehicle, evaluate the geometric compatibility of the disturbance configuration with adjacent components and the overall vehicle shape. If the assembly constraints are not met, propose adjustment suggestions.

[0048] This invention also provides a dynamic optimization system for the wind resistance of a cargo box truck based on airflow simulation, comprising:

[0049] The 3D structure modeling module constructs a 3D structural model of the target cargo truck and obtains the spatial distribution parameters and surface roughness parameters of its fence components.

[0050] The working condition simulation module establishes a simulation model of the external flow field of the corresponding vehicle, simulates the airflow state under various typical driving conditions, and obtains the distribution map of the wind resistance coefficient per unit area acting on each fence component.

[0051] The component identification module constructs a wind resistance sensitivity matrix for fence components based on the obtained wind resistance coefficient distribution map per unit area, and extracts the set of component numbers for areas with high wind resistance sensitivity.

[0052] The configuration sample generation module performs structural perturbation modeling on each component in the component number set, generates multiple local deformation configuration samples, and constructs a set of perturbed aerodynamic simulation models.

[0053] The data extraction module performs parallel simulation calculations on the aerodynamic simulation model set to obtain the total drag coefficient and local aerodynamic stress distribution of each disturbance sample under each working condition, and forms an optimization evaluation function set.

[0054] The configuration selection module performs multi-objective optimization based on the optimization evaluation function set to determine the optimal disturbance configuration. The component number set corresponding to the optimal disturbance configuration satisfies that the total drag coefficient reduction is greater than the preset threshold and the local aerodynamic stress is within the allowable range.

[0055] The structural adjustment suggestion output module outputs structural adjustment suggestions based on the optimal disturbance configuration to guide the dynamic design optimization of the fence components of the cargo box vehicle.

[0056] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0057] 1. This invention proposes a systematic, quantifiable, and iterative method for optimizing wind resistance by constructing a three-dimensional structural model of a staked vehicle and a multi-condition aerodynamic simulation environment, combined with wind resistance sensitive area identification, disturbance modeling, and parallel computing. This method can not only accurately identify key component areas affecting the overall vehicle's wind resistance, but also effectively select the optimal disturbance configuration that balances wind resistance reduction and structural safety through a multi-objective optimization algorithm, thereby significantly improving the aerodynamic performance and energy efficiency of the staked vehicle under different driving conditions.

[0058] 2. The dynamic optimization process of this invention supports structural-level perturbation modeling and sensitivity analysis, resulting in higher simulation accuracy and design guidance value. Through modular modeling, perturbation sample generation, parallel simulation calculation, and parameterized output, this invention can achieve an automated closed loop for wind resistance optimization and structural adjustment suggestions, providing an efficient, practical, and scalable engineering solution for lightweight design, energy saving, and wind farm safety of cargo vehicles. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0060] Figure 1 This is a flowchart of the method of the present invention.

[0061] Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example 1, please refer to Figure 1 As shown in this embodiment, the dynamic optimization method for wind resistance of a cargo box vehicle based on airflow simulation includes:

[0064] Construct a three-dimensional structural model of the target cargo truck and obtain the spatial distribution parameters and surface roughness parameters of its fence components;

[0065] Establish a simulation model of the external flow field of the corresponding vehicle to simulate the airflow state under various typical driving conditions and obtain the distribution map of the wind resistance coefficient per unit area acting on each fence component.

[0066] Based on the obtained distribution map of wind resistance coefficient per unit area, a wind resistance sensitivity matrix of fence components is constructed, and the set of component numbers in the high wind resistance sensitive area is extracted.

[0067] Structural perturbation modeling is performed on each component in the component number set to generate multiple local deformation configuration samples, and a set of perturbed aerodynamic simulation models is constructed.

[0068] Parallel simulation calculations are performed on the aerodynamic simulation model set to obtain the total drag coefficient and local aerodynamic stress distribution of each disturbance sample under each working condition, forming an optimization evaluation function set.

[0069] Multi-objective optimization is performed based on the set of optimization evaluation functions to determine the optimal perturbation configuration. The set of component numbers corresponding to the optimal perturbation configuration satisfies that the decrease in total drag coefficient is greater than a preset threshold and the local aerodynamic stress is within the allowable range.

[0070] Based on the optimal perturbation configuration, structural adjustment suggestions are output to guide the dynamic design optimization of the fence components of the cargo box vehicle.

[0071] Select a physical prototype of the target cargo van model, ensuring the vehicle surface is clean and unobstructed, and place it in an unobstructed open space or laboratory scanning platform.

[0072] A multi-station laser scanning device (such as a FARO Focus S350 or Leica RTC360 3D laser scanner) is used to set up scanning points around the cargo truck at 360 degrees to complete a multi-view scan of the vehicle body. The scanning device should be set to high-precision mode, with a point spacing accuracy better than 1 mm, and the scanning coverage should include at least the entire vehicle shell, outline, and all fence components.

[0073] The raw point cloud data obtained from each scanning perspective is imported into point cloud processing software (such as Geomagic, CloudCompare, or PolyWorks) for initial stitching and merging.

[0074] The merged point cloud data is subjected to noise point removal, outlier repair and boundary identification. Specifically, the RANSAC (Random Sample Consensus) algorithm can be used to identify noise points and the data can be smoothed by median filtering.

[0075] Global alignment is achieved using point cloud registration technology to ensure consistency between data from different perspectives. This embodiment employs the Iterative Closest Point Registration (ICP) algorithm, whose core idea is to achieve rigid body transformation alignment between point clouds by minimizing the Euclidean distance error function between two sets of point clouds.

[0076] The processed point cloud data is converted into a surface mesh model based on a fitting algorithm. The Poisson surface reconstruction algorithm is preferably used to generate continuous surfaces, forming a 3D digital geometric model containing the complete vehicle body outline and structural details. The generated model retains the spatial information of multi-level components such as fences, posts, and base plates, and is output in STL or OBJ standard formats.

[0077] In the generated 3D digital model, fence components are distinguished from the whole vehicle model through semantic segmentation or manual annotation. Each fence component is identified as an independent geometric object and assigned a unique number.

[0078] For each numbered component, its spatial center coordinates (x, y, z) in the global coordinate system, along with its corresponding length, width, and height dimensions, are extracted. This invention organizes the spatial parameters of all components into a spatial distribution parameter matrix P, where each row represents a component, and the column elements are, in order, the center coordinates and dimension values.

[0079] For example, the i-th row of matrix P is represented as: Pi = [xi, yi, zi, li, wi, hi]; where xi, yi, and zi are the geometric center coordinates of the i-th fence component; li, wi, and hi are the length, width, and height of the component, respectively. This parameter matrix provides the basis for subsequent regional correlation analysis of wind drag coefficients.

[0080] High-resolution surface texture acquisition of the target cargo truck can be achieved by using an industrial-grade structured light camera (such as GOMATOS) combined with a light source to capture surface texture images of each fence component at a fixed angle. The image resolution should be better than 100 pixels / cm.

[0081] Image grayscale analysis was performed on the acquired surface texture images to extract characteristic indicators such as the root mean square error of the grayscale distribution, gradient direction distribution density, and local peak change rate. Based on the ISO 25178 roughness standard, a mapping relationship was established between the above image texture features and roughness levels.

[0082] In this embodiment, the roughness level is defined as five levels (R1~R5), where R1 is the mirror smooth level and R5 is the rough processing level. A KNN (K-nearest neighbor) classification model is constructed using the training set to map the texture features corresponding to each component to the roughness level, and this level information is appended to the component attributes.

[0083] For example: if the mean square error of the grayscale distribution is less than 5 and the local peak density is lower than the threshold of 1, it is judged as R1; if the gradient change rate is higher than 15% and the grayscale fluctuation exceeds 12, it is judged as R5.

[0084] Ultimately, each fence component includes the following parameters: spatial location (x, y, z), dimensional information (length, width, height), and surface roughness grade (R1~R5).

[0085] All component attributes are uniformly stored in the component attribute library, providing accurate input conditions for subsequent airflow simulation load calculation, wind resistance sensitive area extraction, and disturbance modeling.

[0086] Based on the three-dimensional digital geometric model generated in the previous stage, import it into a CFD (Computational Fluid Dynamics) simulation platform (such as ANSYS Fluent, Simcenter STAR-CCM+, or OpenFOAM) to construct a simulation domain that includes the vehicle model and the air region.

[0087] The outer boundary of the air region is a rectangular envelope, with its length (X-axis) being 6 times the vehicle's length, its width (Y-axis) being 4 times the vehicle's width, and its height (Z-axis) being 3 times the vehicle's height. The vehicle model is placed at the front 1 / 3 of the simulation domain, with its bottom in contact with the simulation domain ground to simulate actual road contact conditions.

[0088] The boundary conditions are set as follows:

[0089] The front-end boundary is set as the velocity inlet, and the input wind speed value is set according to the corresponding operating conditions;

[0090] The rear boundary is set as a pressure outlet, and the pressure is set to atmospheric pressure.

[0091] The vehicle surface is set to an impenetrable wall boundary condition;

[0092] The ground is set to a sliding surface to simulate the movement of a vehicle relative to the ground.

[0093] Air is assumed to be an incompressible fluid with a density of 1.225 kg / m³ and a viscosity of [missing value]. Pa·second.

[0094] To ensure the simulation model has universality and scenario relevance, multiple simulation sub-models for different typical driving conditions need to be constructed, mainly including the following three categories:

[0095] High-speed straight-line constant speed driving condition: The input wind speed is set to 100 kilometers per hour to simulate the situation of a vehicle driving at a stable speed on a straight section of a highway.

[0096] Crosswind interference condition: Set the prevailing wind direction to a 30-degree angle with the vehicle's direction of travel and the wind speed to 60 kilometers per hour to simulate coastal areas, mountain passes, and other areas susceptible to crosswinds.

[0097] Long downhill coasting condition: The simulation domain is set with a 5-degree tilt angle, the vehicle is kept in a coasting state without power, and the wind speed is set to 80 kilometers per hour to evaluate the impact of changes in the wake behind the vehicle on the load on the fence.

[0098] A separate simulation sub-model is created for each operating condition, keeping the boundary settings consistent with the air domain size, and only adjusting the flow direction angle and velocity parameters.

[0099] To obtain more detailed data on local airflow variations, this embodiment selects Large Eddy Simulation (LES) as the primary solution method to capture transient turbulent behavior between complex components. When computational resources are limited, the Reynolds-Averaged Navier-Stokes (RANS) model can be used for steady-state conditions, with the SST k-ω model being preferred, balancing computational efficiency and accuracy.

[0100] The solution process is as follows:

[0101] Mesh generation: A structured and refined mesh is used, with the mesh size refined to within 2 mm on the surface of the fence components to ensure accurate simulation of the turbulent boundary layer;

[0102] Initial conditions: The initial air temperature is set at 20 degrees Celsius, and the initial turbulence intensity is 5%.

[0103] Time step setting: The LES method uses a time step of 0.001 seconds and a total simulation time of 2 seconds;

[0104] Convergence criterion: Residuals less than As a convergence criterion;

[0105] The solver is set to a pressure-based solver, and the SIMPLE algorithm is used for pressure-velocity coupling.

[0106] After the simulation calculations are completed, the force data for each surface element of the fence component is exported, including the aerodynamic force vector per unit area. The wind resistance coefficient per unit area is defined as follows:

[0107] The wind resistance coefficient per unit area = local wind resistance / component unit area; where local wind resistance is a composite of wall pressure and shear force output by the simulation platform, and its direction of action is consistent with the prevailing wind direction. The surface of each fence component is divided into multiple grid units, and after calculating the wind resistance coefficient of each unit, a wind resistance coefficient distribution map of the component is formed.

[0108] Taking component number Mi as an example, its surface is composed of n grid cells, and the wind resistance coefficient corresponding to cell j is denoted as Cij. Then the wind resistance coefficient spectrum per unit area of ​​component Mi is the set {Ci1, Ci2, ..., Cin}.

[0109] This atlas can be viewed as a two-dimensional matrix data structure and supports mapping onto a three-dimensional model surface for visualization analysis. Simulation results are exported as CSV or VTK files for subsequent use in wind resistance sensitivity matrix construction and disturbance optimization algorithms.

[0110] The wind drag coefficient distribution map per unit area was clustered and integrated according to the component number. The wind drag coefficient distribution map is the result output by the CFD simulation platform under various typical working conditions, which is represented as the local wind drag (per unit area) value on the surface grid cell of each fence component, in Newtons per square meter.

[0111] The average drag coefficient of a component is obtained by averaging the drag coefficient values ​​of all grid cells under the same component number. Let the component number be Mi, and the drag coefficients of its corresponding n grid cells be C1, C2, ..., Cn. Then the average drag coefficient Ci_avg of component Mi can be expressed as: Ci_avg = (C1 + C2 + ... + Cn) / n; this average value reflects the aerodynamic drag level of the component under the corresponding operating conditions.

[0112] After performing the above processing on all components, an initial wind resistance coefficient vector R0 is constructed, where: R0 = [C1_avg, C2_avg, ..., Cm_avg]; m is the total number of fence components.

[0113] To improve the objectivity and comparability of wind resistance response, this invention introduces two component physical characteristics as weighting coefficients to normalize the initial wind resistance vector R0:

[0114] Component size factor Si: This represents the influence of component size on wind resistance. Its value is the ratio of the component's projected area to the average projected area, i.e.: Si = Ai / A_avg; where Ai is the projected area of ​​component Mi facing the airflow direction, and A_avg is the average projected area of ​​all components.

[0115] Surface roughness grade factor Qi: In this embodiment, roughness grades R1~R5 are converted into numerical grades 1~5. The larger the value, the rougher the surface and the greater the impact on wind resistance. After normalization, the value of Qi is between [0.2, 1.0].

[0116] The formula for calculating the standardized wind resistance response value Ei of the final component is: Ei = Ci_avg × Si × Qi; Ei is normalized for all components to generate a standard wind resistance response vector R*, whose element range is [0, 1], which is used to construct subsequent matrices with multi-condition data.

[0117] To characterize the stability and sensitivity of a component's wind resistance response under different operating conditions, it is necessary to integrate wind resistance response data from multiple operating conditions and calculate the volatility and response gradient of each component.

[0118] Suppose n typical operating conditions are established (such as high-speed straight road, crosswind, downhill, etc.), and the standardized wind resistance response value of each component under each operating condition is denoted as Ei1, Ei2, ..., Ein. Then the variance of the wind resistance response Vi of the component is defined as: ; where μi is the mean from Ei1 to Ein.

[0119] In addition, the maximum variation gradient Gi is introduced to measure the degree of abrupt change in the component's drag response under different operating conditions. Its calculation formula is as follows: , where j ≠ k, and j, k ∈ [1, n]; that is, take the maximum difference between the response values ​​under any two working conditions.

[0120] Finally, Vi and Gi are linearly weighted and combined to form the sensitivity index Si: ; where α and β are weighting parameters, set according to the simulation scenario, generally taken as α = 0.5, β = 0.5. Alternatively, α = 0.4, β = 0.6 can be set according to the specific scenario to emphasize the impact of abrupt changes. The Si values ​​of all components are combined to form the wind resistance sensitivity matrix S: S = [S1, S2, ..., Sm].

[0121] Each element in the matrix represents the overall response sensitivity of the corresponding component to changes in wind resistance under different operating conditions. The higher the value, the more significant its impact on the aerodynamic performance of the whole vehicle.

[0122] To identify aerodynamically weak areas in a structure, a selected threshold T is set in the sensitivity matrix S. If Si ≥ T for a certain component, then the component is identified as a highly wind-resistant component.

[0123] In the experiment, the present invention sets T to 0.8 (normalized value), that is, only components within the highest sensitivity of 20% are extracted.

[0124] Construct a set of component numbers M, which contains all component numbers that satisfy Si ≥ T: M = {Mi | Si ≥ T, i∈ [1, m]}; set M is the target region for subsequent disturbance modeling and structural optimization, representing the group of components with the strongest wind resistance response and the greatest impact on the aerodynamic performance of the whole vehicle.

[0125] After determining the set M of highly wind-sensitive components, structural disturbance modeling is prepared for each component in the set. First, a set of disturbance parameters D needs to be established. This parameter set describes the control dimensions of the component during structural deformation, and mainly includes:

[0126] Disturbance amplitude Δ: Represents the intensity of change in the local geometric features of a component, set as a percentage range relative to the original size. For example, the value range of disturbance amplitude Δ can be set to ±5%, ±10%, or ±15%.

[0127] Disturbance direction Θ: Defines the directionality of the disturbance, mainly including: along the prevailing wind direction (X-axis); the normal direction perpendicular to the prevailing wind direction (Y-axis); and local torsional direction (rotation around the Z-axis). Disturbance type τ: Indicates the deformation mode adopted by the structural disturbance, including but not limited to: linear stretching (increase in member length); local contraction (decrease in member height); central depression (depression within the member's central area); edge upturning or downturning (deflection at the member's ends); and torsional deformation (rotation of the cross-section). The parameter set D can be formalized as a triple: D = (Δ, Θ, τ).

[0128] Each component can select multiple combination schemes from the parameter set D according to the preset disturbance strategy, and generate corresponding local deformation configurations respectively.

[0129] For each component Mi, based on the perturbation parameter set D, multiple local perturbation deformation configuration samples are constructed using the shape parameter-driven modeling method. The specific operation steps are as follows:

[0130] Import the original geometry of the component into a CAD environment (such as Siemens NX, SolidWorks, or Rhinoceros) and build a parametric modeling sketch.

[0131] By using each combination of disturbance parameter set D as input variables, the control points, sketch constraints, or curve fitting relationships are adjusted to achieve geometric deformation.

[0132] The shape-driven modeling module generates multiple sets of different deformation configuration samples. For example, for component Mi, deformation samples Mi1, Mi2, Mi3...Mik are generated, and each sample corresponds to a set of perturbation parameters.

[0133] Ensure that each deformed sample does not overlap or break at geometric continuity and connection surface boundaries to maintain its embedding compatibility in the whole vehicle model.

[0134] A typical example of a perturbation deformation sample is shown below:

[0135] Component number: M17;

[0136] Perturbation parameter combination: Δ = +10%, Θ = Y-axis, τ = central depression;

[0137] Deformation sample number generated: M17_3;

[0138] Description: The component sinks by 10% in the central region in the vertical wind direction to simulate the buckling state under wind pressure.

[0139] After constructing the deformable configuration sample, it needs to be embedded into the full vehicle 3D structural model. The operation steps are as follows:

[0140] Load the original 3D model of the vehicle and locate the geometric position, connection points, and boundary relationships of component Mi.

[0141] Delete the original component Mi and replace it with the perturbation sample Mi_j, ensuring interface alignment and keeping adjacent components unchanged.

[0142] Verify the integrity of the model to ensure that the replacement of disturbed components does not affect the overall shape closure of the vehicle body.

[0143] Each disturbance configuration replaces only one component, while the remaining components remain unchanged, in order to analyze the impact of a single component disturbance on the overall wind resistance.

[0144] Finally, for each sensitive component Mi, k disturbance configuration samples are used to generate k vehicle-level 3D models, forming a set of vehicle models after disturbance.

[0145] Import all perturbation configuration models into the CFD simulation platform to automatically generate corresponding simulation calculation models, including:

[0146] Geometric import and flow field construction: Each disturbance model is imported into the simulation software, maintaining the same boundary dimensions, inlet wind speed, outlet pressure, and other conditions as the original simulation field.

[0147] Mesh generation and refinement: Local mesh refinement is performed in the disturbed component area to ensure simulation accuracy; the original simulation model settings are used in the remaining areas.

[0148] Physics field settings and solution configuration:

[0149] Use the same solver settings as the original model (such as RANS or LES).

[0150] Set consistent parameters such as time step and convergence criteria;

[0151] Maintain consistent wind direction, temperature, and turbulence intensity to ensure comparability of comparison results.

[0152] Model naming and organization: Each disturbance simulation model is named with the component number and disturbance combination number, such as M17_3_FlowModel, forming a structured model library.

[0153] Finally, a multidimensional disturbance simulation sample set is obtained, denoted as: F = {F1, F2, ..., Fk}; where each Fi represents an aerodynamic simulation model corresponding to a specific disturbance configuration.

[0154] First, a unified set of simulation parameters needs to be established for each disturbance configuration. To ensure comparability, this invention standardizes and models multiple typical driving conditions, specifically including:

[0155] High-speed straight-line driving condition: Simulated vehicle speed of 100 km / h, wind direction is the same as vehicle direction;

[0156] Crosswind interference condition: Set the wind direction to be at a 30-degree angle to the vehicle direction, and the wind speed to 60 kilometers per hour;

[0157] Downhill skiing conditions: Set a slope of 5 degrees and a wind speed of 80 kilometers per hour.

[0158] The above operating conditions correspond to different boundary condition settings in the simulation platform, including:

[0159] Velocity inlet boundary conditions: specify wind speed and flow direction;

[0160] Sliding ground boundary conditions: Simulating relative driving conditions;

[0161] Turbulence model selection: This invention preferably uses the SST k-ω model in the RANS (Reynolds-Averaged Navier–Stokes) model, which balances simulation efficiency and accuracy;

[0162] Other parameters: air density is 1.225 kg / m³, dynamic viscosity is... Pa·second.

[0163] The above parameters are uniformly encapsulated into a working condition configuration file (e.g., JSON or XML format) for subsequent simulation calls and automated deployment.

[0164] After completing the correspondence between configuration samples and operating parameters, the task is uploaded to a high-performance computing platform or cloud computing environment (such as ANSYS HPC cluster, AWS EC2 CFD computing service, etc.), and a parallel simulation scheduling mechanism is adopted to significantly improve computing efficiency.

[0165] The specific steps of parallel scheduling are as follows:

[0166] Task generation: Each perturbation configuration model is combined with all preset operating conditions to form a perturbation operating condition task pair. For example, the simulation task for configuration Mi_j under operating condition Ck is T(i,j,k).

[0167] Resource allocation strategy: Use task pool scheduling algorithms (such as static block allocation or dynamic load balancing mechanism) to distribute tasks to multiple computing nodes for parallel execution.

[0168] Automated solution process: The solution process for each task is controlled by scripts, including mesh generation, initialization, iteration control, and convergence determination (residual less than 1 / 3). ) and export the results.

[0169] Automatic result archiving: After each simulation task is completed, the output results are automatically compressed and archived to the server database by naming the configuration number + operating condition number.

[0170] This parallel mechanism can complete hundreds of simulation tasks within 48 hours, significantly improving the efficiency of system evaluation.

[0171] After the simulation is completed, key evaluation indicators need to be extracted from each set of perturbation models, mainly including:

[0172] Total drag coefficient (Cd): Automatically calculated by CFD software, it is defined as the ratio of aerodynamic drag to air dynamic pressure acting on the vehicle, expressed as: Where: Fd is the total wind resistance in Newtons; ρ is the air density; V is the relative wind speed; and A is the windward area. This index is used as the main optimization objective, and the smaller the expected value, the better.

[0173] Local aerodynamic stress distribution: Extract the pressure and shear force per unit area on the surface of the disturbed component, calculate the maximum stress σmax and its location, and use it to determine the structural safety and fatigue risk of the component.

[0174] Data vector format organization: The results of each disturbance configuration under multiple operating conditions are organized into vector groups. For example:

[0175] The drag vector of configuration Mi_j is: [Cd1, Cd2, Cd3];

[0176] The maximum stress vector of configuration Mi_j is: [σmax1, σmax2, σmax3].

[0177] All perturbation configuration data are organized in matrix form and input into the optimization function module.

[0178] Based on the data extracted from the simulation, a set of multi-objective optimization evaluation functions is established, including at least one main objective function and one constraint function:

[0179] Objective function F1: Minimize the weighted average of the total drag coefficients: F1(Mi_j) = (Cd1 + Cd2 + Cd3) / 3; where Cd1~Cd3 are the total drag coefficients of configuration Mi_j under three operating conditions. The objective is to find the disturbance configuration that minimizes F1.

[0180] Constraint function F2: Maximum stress constraint: F2(Mi_j) = max(σmax1, σmax2, σmax3) ≤ σ_allow; where σ_allow is the safe stress threshold of the material, which is set to the range of 120 MPa to 250 MPa according to the safety standards of commonly used structural components of aluminum alloy or steel, depending on the material of the component.

[0181] The final optimized evaluation function set is expressed as follows:

[0182] MinimizeF1(Mi_j);

[0183] Subject toF2(Mi_j) ≤ σ_allow;

[0184] This function set supports subsequent optimization of perturbation configurations using genetic algorithms (such as NSGA-II) or multi-objective particle swarm optimization (MOPSO).

[0185] We construct a perturbation optimization design space and objective function space, which serve as the input parameter set for a multi-objective evolutionary algorithm.

[0186] The disturbance design variables are denoted as a triple: D = (Δ, Θ, τ); where: Δ is the disturbance amplitude, defined as the percentage change in the geometric dimensions of the component, with a value range of ±5%, ±10%, ±15%; Θ is the disturbance direction, including along the wind direction (X-axis), perpendicular to the wind direction (Y-axis), and torsion around the Z-axis; τ is the disturbance type, including deformation modes such as stretching, compression, indentation, torsion, and edge warping.

[0187] Each set of perturbation variables defines a unique perturbation configuration sample, corresponding to the simulation model and its output data.

[0188] The objective function has a multi-objective structure, including:

[0189] The objective function for wind resistance, F1, is to minimize the weighted average of the overall drag coefficient Cd of the vehicle under various operating conditions. Let the drag coefficients under n typical operating conditions be Cd1 to Cdn, then: F1 = (Cd1 + Cd2 + … + Cdn) ÷ n;

[0190] The stress constraint function F2 constrains the maximum aerodynamic stress σ_max of each disturbed component to not exceed the allowable stress σ_allow of the component material. Specifically, it is defined as: F2 = max(σ1, σ2, ..., σm) ≤ σ_allow; where σi represents the maximum aerodynamic stress response of the i-th disturbed component under multiple operating conditions, and m is the number of disturbed components.

[0191] Drag reduction threshold T1: The percentage decrease in drag coefficient relative to the original undisturbed model, preferably set to 8%;

[0192] Stress tolerance threshold T2: Determined based on the mechanical properties of the material. In this embodiment, 150 MPa is used as σ_allow for the aluminum alloy fence component.

[0193] After defining the variable space and objective function, the Non-Dominated Sorting Genetic Algorithm II (NSGA-II) is used as the optimization solution tool in this invention.

[0194] The core process of the NSGA-II algorithm is as follows:

[0195] Population initialization: Randomly select perturbation combinations from the perturbation configuration sample set as the initial population. The population size is set to 100 individuals, and each individual corresponds to a perturbation configuration.

[0196] Fitness assessment: For each individual, calculate its value under the objective function F1 and the constraint function F2, and classify it according to the non-dominated ranking principle.

[0197] Selection and crossover operations: Elite individuals are selected by crowding distance sorting to participate in crossover and mutation, generating the next generation of perturbation combinations.

[0198] Iterative optimization: Repeat the evolutionary process of crossover, mutation, and fitness evaluation, with the number of iterations set to 50 rounds, or until the objective function changes converge.

[0199] Output Pareto optimal solution set: All non-dominated solutions constitute the Pareto front, representing the optimal trade-off between wind resistance performance and structural safety.

[0200] This algorithm has global optimization capabilities and is suitable for complex nonlinear multi-objective problems.

[0201] The Pareto optimal solution set obtained from the optimization algorithm is further screened to ensure that the final optimization result is feasible and safe in practical engineering.

[0202] The filtering criteria are as follows:

[0203] The wind resistance objective function value F1 in the solution set must be greater than the wind resistance reduction threshold T1 (e.g., 8%) relative to the unperturbed model.

[0204] The maximum aerodynamic stress σ_max of all disturbed components under multiple operating conditions must be less than the stress allowable threshold T2 (e.g., 150 MPa).

[0205] The solutions that satisfy the above two conditions constitute the final set of optimized feasible solutions, from which the configuration with the best performance can be selected for implementation.

[0206] For the finally determined optimal perturbation configuration, the component numbers involved are extracted, an optimized component number set M′ is generated, and structural adjustment suggestions are generated simultaneously.

[0207] The specific steps are as follows:

[0208] Traverse the structural model of the optimal perturbation configuration, extract the component numbers of all perturbations, such as M17, M22, M38, etc., and form a set M′;

[0209] For each disturbed component, record its disturbance parameters (Δ, Θ, τ) as a structural deformation suggestion. For example:

[0210] M17: +10% stretch, Y-axis direction;

[0211] M22: -5% thickness, Z-axis direction;

[0212] M38: +15% torsion, 12 degrees angle.

[0213] The above set of numbers and parameter suggestions can be output as an engineering adjustment manual, which can be directly connected to CAD modeling or production line manufacturing instructions.

[0214] For all disturbed components in the final determined optimal disturbance configuration, disturbance parameters are analyzed, and their deformation data, including disturbance amplitude, disturbance direction and disturbance type, are extracted.

[0215] The disturbance parameter triplet is denoted as: D(i) = (Δi, Θi, τi); where: Δi represents the disturbance amplitude of component i, defined as the percentage change relative to the original size, in % %.

[0216] Θi represents the direction of disturbance, with values ​​including prevailing wind direction (X-axis), vertical wind direction (Y-axis), and vertical or torsional direction (Z-axis); τi represents the type of disturbance, supporting five basic deformations: linear stretching, local contraction, central depression, edge warping, and cross-sectional torsion.

[0217] The original geometric parameters of component i are denoted as: G0(i) = (Li, Hi, Ti, αi); where: Li is the length, Hi is the height, Ti is the thickness, and αi is the original curvature or angle parameter.

[0218] For component i in the optimal configuration, the deformed parameter G1(i) is extracted, and the difference ΔG(i) is calculated: ΔG(i) = G1(i) − G0(i); this difference is used to quantitatively describe the deformation intensity. For example, if ΔG(i).Li = +12 mm, it means that the length of the component increases by 12 mm after the disturbance; if αi changes from 0° to 15°, it means that the component experiences 15 degrees of edge warping. After unifying the units of all difference data, they are recorded in an intermediate data structure for subsequent suggestion generation.

[0219] Based on the above difference analysis results, this invention constructs a structural adjustment mapping table for engineering implementation, which clearly indicates the deformation type, numerical range and operation direction to be implemented for each disturbed component.

[0220] Table 1 Structural Adjustment Mapping Table

[0221]

[0222] As shown in Table 1, the component number is M17; the disturbance type τ = central depression; the disturbance direction Θ = Y-axis (perpendicular to the wind direction); the original height Hi = 800 mm, and the deformed height H'i = 720 mm; therefore, Δi = −10%.

[0223] Based on the disturbance type and the magnitude of the difference quantization, and combined with the standard processing methods and process specifications in the engineering database, adjustment suggestions are automatically generated. For example:

[0224] Tensile disturbance → Increase the length of the rod, which can be achieved by welding or module extension;

[0225] Torsional disturbance → Adjust the angle of the connecting piece or use an eccentric support;

[0226] Thickness disturbance → Use thicker plates or surface overlay welding.

[0227] Considering that although the disturbance configuration is optimal in terms of wind resistance performance, it may cause assembly interference, local interference or geometric overrun problems in the actual vehicle structure, a space compatibility assessment is required to ensure that the optimal design is feasible for assembly.

[0228] Extract the 3D positioning data of each component in the vehicle model: P(i) = (Xi, Yi, Zi); where Xi, Yi, and Zi represent the center position of component i in the vehicle coordinate system.

[0229] The AABB (Axis-Aligned Bounding Box) algorithm is used to quickly detect collisions between the deformed geometric boundary of the disturbed component and the boundaries of its neighboring components to determine whether there are assembly conflicts or overlaps.

[0230] If a spatial conflict is detected, the system will automatically adjust the disturbance amplitude Δi or direction Θi based on the degree of overlap of the conflict boundaries, generating a "feasible deformation alternative." For example, the original suggestion was +15% torsion, but the conflict area was too large; after fine-tuning, the suggestion is +10% torsion, eliminating the conflict. This adjustment process ensures that design optimization will not disrupt the original assembly logic and structural stability.

[0231] Example 2, please refer to Figure 2 As shown in this embodiment, the dynamic optimization system for wind resistance of a cargo box vehicle based on airflow simulation includes:

[0232] The 3D structure modeling module constructs a 3D structural model of the target cargo truck and obtains the spatial distribution parameters and surface roughness parameters of its fence components.

[0233] The working condition simulation module establishes a simulation model of the external flow field of the corresponding vehicle, simulates the airflow state under various typical driving conditions, and obtains the distribution map of the wind resistance coefficient per unit area acting on each fence component.

[0234] The component identification module constructs a wind resistance sensitivity matrix for fence components based on the obtained wind resistance coefficient distribution map per unit area, and extracts the set of component numbers for areas with high wind resistance sensitivity.

[0235] The configuration sample generation module performs structural perturbation modeling on each component in the component number set, generates multiple local deformation configuration samples, and constructs a set of perturbed aerodynamic simulation models.

[0236] The data extraction module performs parallel simulation calculations on the aerodynamic simulation model set to obtain the total drag coefficient and local aerodynamic stress distribution of each disturbance sample under each working condition, and forms an optimization evaluation function set.

[0237] The configuration selection module performs multi-objective optimization based on the optimization evaluation function set to determine the optimal disturbance configuration. The component number set corresponding to the optimal disturbance configuration satisfies that the total drag coefficient reduction is greater than the preset threshold and the local aerodynamic stress is within the allowable range.

[0238] The structural adjustment suggestion output module outputs structural adjustment suggestions based on the optimal disturbance configuration to guide the dynamic design optimization of the fence components of the cargo box vehicle.

[0239] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A dynamic optimization method for wind resistance of a cargo box vehicle based on airflow simulation, characterized in that: include: Construct a three-dimensional structural model of the target cargo truck and obtain the spatial distribution parameters and surface roughness parameters of its fence components; Establish a simulation model of the external flow field of the corresponding vehicle to simulate the airflow state under various typical driving conditions and obtain the distribution map of the wind resistance coefficient per unit area acting on each fence component. Based on the obtained distribution map of wind resistance coefficient per unit area, a wind resistance sensitivity matrix of fence components is constructed, and the set of component numbers in the high wind resistance sensitive area is extracted. Structural perturbation modeling is performed on each component in the component number set to generate multiple local deformation configuration samples, and a set of perturbed aerodynamic simulation models is constructed. Parallel simulation calculations are performed on the aerodynamic simulation model set to obtain the total drag coefficient and local aerodynamic stress distribution of each disturbance sample under each working condition, forming an optimization evaluation function set. Multi-objective optimization is performed based on the set of optimization evaluation functions to determine the optimal disturbance configuration. The set of component numbers corresponding to the optimal disturbance configuration satisfies that the decrease in total drag coefficient is greater than a preset threshold and the local aerodynamic stress is within the allowable range. Based on the optimal perturbation configuration, structural adjustment suggestions are output to guide the dynamic design optimization of the fence components of the cargo box vehicle.

2. The method for dynamic optimization of wind resistance of a cargo box vehicle based on airflow simulation according to claim 1, characterized in that: The construction of the three-dimensional structural model of the target cargo box vehicle includes: High-precision point cloud data of the vehicle's external structure is obtained by laser point cloud scanning of the actual cargo truck. Point cloud reconstruction algorithms are used to denoise, register, and fit surfaces to point cloud data, generating a three-dimensional digital geometric model containing detailed components. The fence components in the 3D model are numbered by region, and the spatial position, length, height and thickness information of each component are extracted to form a spatial distribution parameter matrix; By acquiring surface texture images and analyzing grayscale texture distribution, combined with a standard roughness modeling library, the surface roughness level of each fence component is estimated and assigned.

3. The method for dynamic optimization of wind resistance of a cargo box vehicle based on airflow simulation according to claim 2, characterized in that: The establishment of the corresponding vehicle external flow field simulation model includes: Based on a three-dimensional digital geometric model, a computational fluid dynamics simulation domain containing a vehicle model and an external air domain is constructed, and boundary conditions and inlet velocity parameters are set. Various typical driving conditions were selected, including high-speed straight driving, crosswind interference driving, and long downhill coasting operation scenarios, and simulation model groups for the corresponding conditions were constructed respectively. Steady-state or transient simulations were performed on each set of working conditions using large eddy simulation or Reynolds time-averaged equation solution methods to obtain the wind resistance distribution per unit area on the surface of the fence components. The simulation output is projected onto the surface mesh of each component, the ratio of wind resistance to area is calculated, and a distribution map of wind resistance coefficient per unit area is generated.

4. The method for dynamic optimization of wind resistance of a cargo box vehicle based on airflow simulation according to claim 3, characterized in that: The wind resistance sensitivity matrix of the constructed fence components includes: The distribution map of wind resistance coefficient per unit area is clustered and integrated according to component number. The average wind resistance coefficient value of each fence component is calculated, and an initial vector of wind resistance coefficient is constructed. The component size factor and surface roughness level are used as correction weights to normalize the initial vector of the drag coefficient, generating a standardized drag response vector. Based on multi-condition wind resistance response data, the variance and maximum gradient value of wind resistance change of each fence component under different conditions are calculated as wind resistance sensitivity index, and a wind resistance sensitivity matrix is ​​constructed. Set a preset sensitivity threshold, and filter the component numbers in the sensitivity matrix whose corresponding values ​​are greater than the sensitivity threshold to form a set of component numbers in the high wind resistance sensitive area.

5. The method for dynamic optimization of wind resistance of a cargo box vehicle based on airflow simulation according to claim 4, characterized in that: The structural perturbation modeling for each component in the component number set includes: Based on the spatial geometric information and boundary constraints of each component in the component number set, a set of disturbance parameters is set, including disturbance amplitude, disturbance direction and disturbance type; Different combinations of perturbation parameters are applied to each component, and multiple local deformation configuration samples are generated by shape parameter transformation. The deformation types include deflection, torsion and thickness deformation. Each local deformation configuration is embedded into the original three-dimensional model of the vehicle to construct the disturbed geometric model of the vehicle, while keeping the structure of the remaining components unchanged; For all perturbation configuration models, a set of corresponding computational fluid dynamics simulation models is generated in batches, and the simulation boundary conditions are uniformly set as the data input set for solving the optimization evaluation function.

6. The method for dynamic optimization of wind resistance of a cargo box vehicle based on airflow simulation according to claim 5, characterized in that: Parallel simulation calculations of the aerodynamic simulation model set include: For each disturbance configuration model, a unified set of multi-condition simulation parameters is set, including flow velocity, inflow angle, ground sliding conditions and turbulence model, and a condition configuration file is constructed. The perturbation configuration model set is combined with the corresponding operating condition configuration in batches, and parallel simulation is performed using a high-performance computing platform or cloud computing environment. After the simulation, the total drag coefficient and aerodynamic load distribution on the surface of the fence components for each disturbance configuration under each working condition are extracted and stored in vector form. Based on the changes in drag coefficient and local aerodynamic stress of components under different working conditions, a drag optimization objective function and a stress constraint function are constructed to form a multi-objective optimization evaluation function set.

7. The method for dynamic optimization of wind resistance of a cargo box vehicle based on airflow simulation according to claim 6, characterized in that: The multi-objective optimization based on the set of optimization evaluation functions includes: Based on the sample set of disturbance configurations and the corresponding wind resistance objective function and stress constraint function, the design variable space and objective function space are constructed, and the total wind resistance coefficient reduction threshold T1 and the maximum allowable stress threshold T2 are defined. A multi-objective evolutionary algorithm is used to iteratively optimize and search the perturbation configuration samples, resulting in a Pareto optimal solution set that balances wind resistance performance and aerodynamic strength. In the Pareto optimal solution set, a solution set is selected where the wind resistance decreases by more than the threshold T1 and the local stress of all disturbing components does not exceed the threshold T2. The set of component numbers involved in the optimal disturbance configuration that meets the screening criteria is denoted as M′, and the wind resistance optimization adjustment suggestions and component deformation parameters corresponding to the configuration are output.

8. The method for dynamic optimization of wind resistance of a cargo box vehicle based on airflow simulation according to claim 7, characterized in that: The output structure adjustment suggestions based on the optimal perturbation configuration include: The perturbation parameters of each component in the optimal perturbation configuration are analyzed, including perturbation amplitude, perturbation direction and perturbation type, and the difference analysis is performed with the original component geometric parameters. Based on the difference analysis results, a structural adjustment mapping table is constructed to clarify the deformation operation type and numerical range of each disturbed component; Based on the spatial positioning relationship of components in the whole vehicle, evaluate the geometric compatibility of the disturbance configuration with adjacent components and the overall vehicle shape. If the assembly constraints are not met, propose adjustment suggestions.

9. A dynamic optimization system for the wind resistance of a cargo box vehicle based on airflow simulation, used to implement the dynamic optimization method for the wind resistance of a cargo box vehicle based on airflow simulation as described in any one of claims 1-8, characterized in that: include: The 3D structure modeling module constructs a 3D structural model of the target cargo truck and obtains the spatial distribution parameters and surface roughness parameters of its fence components. The working condition simulation module establishes a simulation model of the external flow field of the corresponding vehicle, simulates the airflow state under various typical driving conditions, and obtains the distribution map of the wind resistance coefficient per unit area acting on each fence component. The component identification module constructs a wind resistance sensitivity matrix for fence components based on the obtained wind resistance coefficient distribution map per unit area, and extracts the set of component numbers for areas with high wind resistance sensitivity. The configuration sample generation module performs structural perturbation modeling on each component in the component number set, generates multiple local deformation configuration samples, and constructs a set of perturbed aerodynamic simulation models. The data extraction module performs parallel simulation calculations on the aerodynamic simulation model set to obtain the total drag coefficient and local aerodynamic stress distribution of each disturbance sample under each working condition, and forms an optimization evaluation function set. The configuration selection module performs multi-objective optimization based on the optimization evaluation function set to determine the optimal disturbance configuration. The component number set corresponding to the optimal disturbance configuration satisfies that the total drag coefficient reduction is greater than the preset threshold and the local aerodynamic stress is within the allowable range. The structural adjustment suggestion output module outputs structural adjustment suggestions based on the optimal disturbance configuration to guide the dynamic design optimization of the fence components of the cargo box vehicle.

Images