Extensible chassis component structure of vehicle and performance optimization design method thereof

Abstract

The invention discloses a performance optimization design method for a vehicle extensible chassis component structure, which comprises the following steps of: based on an extensible chassis strategy, establishing a multi-target topological optimization mathematical model by using a control arm original model, optimizing to obtain a control arm universal force transfer path of an extensible chassis, optimizing to form a control arm reconstruction geometric model, and designing the performance of the extensible chassis component structure. A control arm simulation data set is obtained through simulation, and a stacked integrated learning model constructed and formed by a plurality of base learner models is obtained through training; and outputting an optimal parameter combination of the control arm through multi-objective optimization based on the stacking integration model. According to the method, an extensible chassis strategy, topological optimization, parametric modeling, machine learning and an intelligent optimization algorithm are organically combined, adaptive structure parameter design can be achieved through topological optimization for different chassis components, and multi-target optimization of the model is achieved through a stacked integrated model and a multi-target optimization algorithm; the efficiency of optimizing the structural parameters of the chassis part of the extensible chassis strategy is improved, and the structural performance of the chassis part is remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle structural component design methods, and in particular relates to a vehicle expandable chassis component structure and a performance optimization design method thereof. Background Art

[0002] The personalized needs of the automobile consumer market are becoming increasingly prominent, and the market demand for different sizes of vehicles is becoming more diversified, which places higher demands on the versatility and flexibility of vehicle chassis components. In this context, developing a scalable chassis platform has become the key to improving the competitiveness of automobile companies, and controlling its R&D costs is a core issue related to the survival of the company. Therefore, in the design process of the entire vehicle, how to improve the structural performance and development efficiency of scalable chassis components is very important. In the existing technology, the chassis structure design method mainly relies on experience-driven structural design and uses traditional optimization algorithms to optimize the structure. In terms of structural design, a CAD model is generally constructed with a small number of parameters such as part mass and load conditions. During the structural optimization analysis, the CAD model is repeatedly modified to perform the corresponding finite element analysis. This method relies on experience to construct a CAD model with a small number of parameters. When expanding the wheelbase of chassis of different sizes, the model needs to be repeatedly reconstructed, and the modular adaptation cost is high. In terms of algorithm optimization, traditional optimization algorithms such as the NSGA-II algorithm are generally used to optimize the control arm. The Pareto frontier solution set does not cover the effective design space sufficiently, and the development cycle is lengthened. This method is prone to falling into local optimality in multi-objective optimization, the solution set is unevenly distributed, and the solution speed is slow, which cannot meet the collaborative optimization requirements of the scalable chassis for multiple performance indicators. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the technical problems in the prior art of the scalable design optimization method for vehicle chassis components, which have poor versatility, resulting in low R&D efficiency and difficulty in achieving effective multi-objective optimization, thereby providing a vehicle scalable chassis component structure and its performance optimization design method.

[0004] A performance optimization design method for a vehicle expandable chassis component structure comprises the following steps: determining a design wheelbase of the expandable chassis, determining the size parameters of a control arm original model based on the design wheelbase; establishing a multi-objective topology optimization mathematical model based on the control arm original model of the expandable chassis design strategy, optimizing and obtaining a universal force transmission path of the control arm of the expandable chassis strategy, optimizing and forming a reconstructed geometric model of the control arm, and obtaining a control arm simulation data set through simulation, comprising the following steps: obtaining the control arm original model, defining a design area and a non-design area of the control arm original model; creating an optimization space in the design area by a solid filling method; establishing a multi-objective topology optimization mathematical model, and converting multi-working condition strain energy into a multi-working condition strain energy. Minimization and maximization of natural frequencies are simultaneously set as optimization objectives, and a compromise programming method is used to optimize the optimization space to obtain a universal force transmission path of the control arm; based on the universal force transmission path of the control arm, a reconstructed geometric model of the control arm is optimized through three-dimensional modeling software; based on the reconstructed geometric model of the control arm, multiple key structural dimension parameters are used as design variables, and bending stiffness, first-order modal frequency, and mass are used as output responses, and a control arm simulation data set is obtained through simulation; based on the control arm simulation data set, a stacked ensemble learning model formed by training multiple base learner models is obtained; based on the stacked ensemble model, the optimal parameter combination of the control arm is output through multi-objective optimization.

[0005] Furthermore, when defining the design area and non-design area of the original model of the control arm of the scalable chassis strategy, the ball joint, bushing and arm body contour are defined as the non-design area, and the interior of the arm body is defined as the design area.

[0006] Furthermore, when establishing a multi-objective topology optimization mathematical model, the vertical degrees of freedom of the ball joint connection, the lateral and vertical translational degrees of freedom of the rear bushing connection, and the lateral, vertical and longitudinal translational degrees of freedom of the front bushing connection are taken as model constraints; it also includes the draft constraint and the maximum and minimum member size constraints as constraints.

[0007] Furthermore, the multi-objective topology optimization mathematical model is expressed as: ; in, k Indicates the number of working conditions, ω k Indicates the k The flexibility weighting factor of each working condition; p represents the order of the natural frequency, C i ( x ) indicates the i The initial flexibility of the structure under each working condition is C imax and C imin Indicates the iThe maximum and minimum values of flexibility under each working condition; f p ( x ) indicates the p The natural frequency of the initial structure of the order; f pmax and f pmin The first p The upper and lower limits of the order frequency; ω represents the weight of the flexibility function; The working conditions include two-wheel passing over bumps, two-wheel passing over potholes and two-wheel passing over slopes.

[0008] Furthermore, the method also includes the following steps: establishing a reconstructed geometric model of the control arm based on the scalable chassis strategy, performing mesh deformation parameterization, building a parametric model of the control arm, and performing load condition analysis, including: creating four load steps, corresponding to the two-wheel driving over bumps, potholes, slope driving and modal analysis conditions.

[0009] Furthermore, based on the control arm simulation data set, a stacked ensemble learning model formed by multiple base learner models is trained, including the following method steps: multiple candidate base learner models are selected, and they are trained based on the control arm simulation data set to obtain an optimized candidate base learner model; based on the particle swarm optimization algorithm, the mean square error of the regression evaluation index of the stacked ensemble model is used as the fitness function, and the optimal optimized candidate base learner model combination is obtained through swarm intelligent search; based on the optimal optimized candidate base learner model combination, a two-layer stacked ensemble model is constructed through the k-fold crossover method.

[0010] Furthermore, based on the stacked integrated model, multi-objective optimization is used to output the optimal parameter combination of the control arm, including the following method steps: based on the stacked integrated model, the objective function is multi-objective optimized using the MOMPA multi-objective marine predator algorithm to obtain the Pareto optimal frontier of the control arm output response, the optimal parameter point is selected using the K-Means clustering algorithm, and the optimal parameter combination of the control arm is output.

[0011] Furthermore, it also includes: the parameter combination of the Pareto optimal frontier includes three dimensions: bending stiffness, first-order modal frequency, and mass. The three dimensions are divided into three clusters through the K-Means clustering algorithm. For each cluster, the distance from each point to the cluster center is calculated, and the point with the closest distance is selected as the representative solution, that is, the optimal parameter point.

[0012] A vehicle expandable chassis component structure includes a control arm, and the control arm structure is formed by the above-mentioned performance optimization design method.

[0013] Furthermore, the arm body of the control arm is an integral triangle, and the three corners of the arm body respectively form a front bushing, a rear bushing and a ball joint.

[0014] Beneficial effects: The present invention discloses a performance optimization design method for a vehicle expandable chassis component structure. The method is based on the expansion chassis strategy, establishes a multi-objective topology optimization mathematical model with the original model of the control arm, optimizes and obtains the universal force transmission path of the control arm, optimizes and forms a reconstructed geometric model of the control arm, obtains a control arm simulation data set through simulation, and trains and obtains a stacked ensemble learning model formed by constructing multiple base learner models based on the control arm simulation data set; based on the stacked ensemble model, the optimal parameter combination of the control arm is output through multi-objective optimization. The present invention organically combines topology optimization, parametric modeling, machine learning and intelligent optimization algorithms, and can achieve adaptive structural parameter design for different chassis size component structures through topology optimization, and achieves multi-objective optimization of the model through the stacked ensemble model and the multi-objective optimization algorithm, thereby improving the efficiency of chassis component structural parameter optimization and significantly improving chassis component structural performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0016] Figure 1 This is a flow chart of the performance optimization design method of the present invention; Figure 2 This is a schematic diagram of the original model structure of the control arm of the present invention; Figure 3 This is a schematic diagram of the universal force transmission path structure of the control arm of the present invention; Figure 4 This is a schematic diagram of the geometric model structure of the control arm reconstructed according to the present invention.

[0017] Description of the accompanying drawings: 11. Interior of the arm; 12. Outline of the arm; 2. Front bushing; 3. Rear bushing; 4. Ball joint. DETAILED DESCRIPTION

[0018] To make the above-mentioned objects, features, and advantages of the present application more clearly understood, the specific embodiments of the present application are described in detail below with reference to the accompanying drawings. The following description sets forth many specific details to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar improvements without violating the scope of the present application. Therefore, the present application is not limited to the specific embodiments disclosed below.

[0019] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0020] In this application, unless otherwise specified or limited, the terms "installed," "connected," "connect," "fixed," etc. should be understood in a broad sense. For example, they can refer to fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; internal communication between two elements or interaction between two elements, unless otherwise specified. Those skilled in the art will understand the specific meanings of the above terms in this application based on specific circumstances.

[0021] Example 1: Reference Figure 1 As shown, this embodiment provides a performance optimization design method for a vehicle expandable chassis component structure, comprising the following steps: Step S1: Based on the scalable chassis strategy, a multi-objective topology optimization mathematical model is established with the original control arm model to optimize and obtain a universal force transmission path of the control arm, optimize and form a reconstructed geometric model of the control arm, and obtain a control arm simulation data set through simulation; Step S2: Based on the control arm simulation data set, training a stacked ensemble learning model constructed by multiple base learner models; Step S3: Based on the stacked integrated model, the optimal parameter combination of the control arm is output through multi-objective optimization.

[0022] In this embodiment, the method uses topology optimization to find the optimal force transmission configuration of the key component structure of the scalable chassis strategy, and completes parametric modeling based on this configuration, adapting to the needs of chassis platform wheel and wheelbase expansion through a parameter-adjustable architecture; at the same time, a performance prediction model is used to establish a parameter-performance mapping of the chassis component structure, and the MIAOA adaptive iterative marine predator optimization algorithm is used in conjunction to solve its Pareto frontier, thereby achieving performance optimization of the chassis components under a multi-parameter adjustable scalable architecture.

[0023] This approach, based on a scalable chassis design strategy, integrates a collaborative framework involving topology optimization of key chassis components, parametric modeling, performance prediction models, and intelligent optimization algorithms. This versatile and scalable approach can be applied to the optimization of various chassis architecture platforms and key chassis components. For example, for control arms (such as A-arms, L-arms, or wishbones) of varying chassis architecture platforms and configurations, topology optimization is used to determine material distribution to obtain a universal chassis architecture's optimal force transmission path, adapting to varying suspension hardpoint constraints. Control arm geometric features (such as arm length, cross-sectional shape, and connection point location) are parameterized to support rapid adjustments to changes in wheelbase or load. For other chassis components, such as subframes, which must accommodate the varying mounting points of multiple powertrains (fuel / electric) while also meeting NVH and crash safety requirements, topology optimization is used to determine the force transmission path (transferring battery pack crash loads to the longitudinal rails). Subframe mounting point locations and cavity thickness are parameterized to accommodate varying motor sizes. The Pareto solution is optimized based on modal frequency, weight, and crash energy absorption.

[0024] As one embodiment of the present invention, the optimized control arm model utilizes a triangular control arm for a McPherson strut suspension structure. Its main body is an aluminum alloy casting, equipped with dual bushings (differentiated by assembly position, front and rear bushings) and a single ball joint. In this embodiment, the control arm model comprises a triangular control arm body, with the front bushing, rear bushing, and ball joint secured to the three corners of the control arm body.

[0025] Specifically, in this embodiment, step S1: establishing a multi-objective topology optimization mathematical model based on the original model of the control arm, optimizing to obtain a universal force transmission path of the control arm, optimizing to form a reconstructed geometric model of the control arm, and obtaining a control arm simulation data set through simulation; includes the following method steps: Step S1.1: Based on the scalable chassis strategy, obtain the original model design space of the control arm, refer to Figure 2 As shown in the figure, the design area and non-design area of the original model of the control arm are defined; an optimization space is created in the design area by using the solid filling method; a multi-objective topology optimization mathematical model is established, and the minimization of multi-condition strain energy and the maximization of natural frequency are simultaneously set as optimization objectives. The compromise programming method is used to optimize the optimization space to obtain a universal force transmission path of the control arm; In this embodiment, the scalable chassis strategy is used to determine the design wheelbase of the scalable chassis and, based on the design wheelbase, to determine the dimensional parameters of the control arm prototype model. After determining the vehicle's design wheelbase, variable parameters are set in the Y-axis length of the control arm body and the length from the control arm ball joint to the arm body in the Y-axis to adjust the wheelbase. By determining these variable parameters, the relative position data of the ball joint and bushing of the control arm structure can be determined. The ball joint and bushing are connected through the control arm body, thereby determining the arm structure dimensional parameters of the control arm prototype model.

[0026] Specifically, when using the variable density method for control arm topology optimization in finite element software, the solid infill method sets the control arm's main structure as the design domain, and uses the relative density of the elements as the design variable. This variable ranges from 0 to 1: a density of 0 indicates complete material removal (empty region), a density of 1 indicates complete material retention (solid region), and intermediate values between 0 and 1 correspond to hypothetical transitional density material states.

[0027] In this embodiment, the non-design area is the ball joint, bushing and arm body contour, and the design area is the interior of the arm body; When establishing the multi-objective topology optimization mathematical model, the vertical degrees of freedom of the ball joint connection, the lateral and vertical translational degrees of freedom of the rear bushing connection, and the lateral, vertical, and longitudinal translational degrees of freedom of the front bushing connection are taken as model constraints. To ensure the manufacturability of the optimization results, draft constraints and maximum and minimum member size constraints are also taken as constraints.

[0028] In this embodiment, considering that single-condition optimization can easily lead to design flaws, a multi-condition optimization strategy is adopted, with three extreme conditions set as optimization boundary conditions. This ensures a reasonable force transmission path for the optimized structure, reduces the risk of structural failure, and achieves more reliable optimization results. The three extreme conditions include two wheels passing over a bump, two wheels passing over a pothole, and two wheels passing over a slope. The weight coefficients for these conditions are set to 0.5, 0.3, and 0.2, respectively, based on their importance.

[0029] In this embodiment, a flexibility topology optimization model under multiple working conditions is constructed: ; ; Wherein, ρ represents the material density; m represents the total number of load conditions, in this embodiment, m=3; represents the weight coefficient of the kth working condition; q represents the penalty factor and q≥2, in this embodiment, q=2; represents the total strain energy function of the element for the kth load case; represents the maximum value of the total strain energy function of the structure under the kth load case; represents the minimum value of the total strain energy function of the structure under the kth load case; represents the effective volume of the optimized structure; represents the original volume of the structure; f represents the percentage of volume constraint, in this embodiment f=0.3; represents the maximum stress value of the kth load case; Indicates the allowable stress of the material.

[0030] In the multi-condition flexibility topology optimization, a single-condition optimization strategy was employed to establish independent models for three typical conditions: dual-wheel traversal of bumps, pits, and slopes. Compliance was used as the single objective function, and the same volume fraction constraint (f = 0.3) was applied. The improvement in structural stiffness achieved by topology optimization was evaluated by comparing the maximum flexibility of the original design model with the minimum flexibility after single-condition optimization.

[0031] Build a natural frequency topology optimization model: ; in, represents the total volume of the design domain, β represents the maximum allowed volume fraction, which is 0.3 in this embodiment. Represents the structural material density function obtained based on the variable density interpolation model (SIMP).

[0032] In the natural frequency topology optimization, the first three natural frequencies of the control arm are optimized and calculated. By comparing the minimum natural frequency calculated from the original model with the maximum natural frequency after optimization, the impact of topology optimization on the dynamic performance of the structure is evaluated.

[0033] Considering the complex coupling relationship between flexibility and natural frequency, the traditional single-objective or simple linear weighted method has limitations. In this embodiment, based on the results of flexibility topology optimization and natural frequency topology optimization, a compromise planning method is adopted to set the minimization of multi-condition strain energy and the maximization of natural frequency as the goals at the same time, and a multi-objective topology optimization mathematical model is established. After the optimization is completed, the post-processing software is analyzed and the optimized control arm conceptual design model is lightweight in the middle area. The key force transmission path retains the reinforced structure, which improves the structural stiffness and material utilization. Specifically, after finite element post-processing software analysis, based on the optimization results of the final iterative step, the unit density threshold is set to 0.45 for material filtering, and the universal force transmission path of the control arm with a clear geometric boundary is extracted. Figure 3 As shown: ; in, k Indicates the number of working conditions, ω k Indicates the k The flexibility weighting factor of each working condition; p represents the order of the natural frequency, C i ( x ) indicates the i The initial flexibility of the structure under each working condition is C imax and C imin Indicates the i The maximum and minimum values of flexibility under each working condition; fp ( x ) indicates the p The natural frequency of the initial structure of the order; f pmax and f pmin The first p The upper and lower limits of the order frequency; ω Represents the weight of the flexibility function.

[0034] Step S1.2: Based on the universal force transmission path of the control arm, a reconstructed geometric model of the control arm is optimized by 3D modeling software; specifically, in the CAD modeling software, the conceptual model is reconstructed, large holes are retained, small holes are removed, the position of thin-walled reinforcement ribs is adjusted to form an I-beam cross-section, the rib distribution is optimized, and draft angles are added, and the fillet radius and wall thickness distribution are optimized. Finally, the reconstructed geometric model of the control arm is obtained, and reference is made to the Figure 4 shown.

[0035] In this embodiment, the arm body of the control arm reconstructs the geometric model: the edge connecting the front bushing and the ball joint is defined as the first edge, the edge connecting the ball joint and the rear bushing is defined as the second edge, and the edge connecting the front bushing and the rear bushing is defined as the third edge; along the front bushing to the ball joint, the arm body is hollowed out to form the first hollow, the second hollow and the third hollow in sequence; along the front bushing to the rear bushing, the arm body is hollowed out to form the fourth hollow, the fifth hollow and the sixth hollow in sequence; the arm body area between the second hollow and the third hollow and the first edge is defined as the ball joint front longitudinal beam, the arm body structure between the first hollow and the first edge is defined as the front bushing front longitudinal beam, and the fourth hollow and the The arm area between the first sides is defined as the rear longitudinal beam of the front bushing, the arm area between the third hollow and the third side is defined as the rear outer longitudinal beam of the rear bushing, the arm area between the first hollow and the fourth hollow is defined as the front inner crossbeam of the front bushing, the arm area between the fifth hollow and the second side is defined as the rear inner crossbeam of the rear bushing, the arm area between the second hollow and the third hollow is defined as the ball joint front crossbeam, the arm area between the first hollow and the second hollow is defined as the ball joint rear crossbeam, the arm area between the fifth hollow and the sixth hollow is defined as the rear bushing front crossbeam, and the arm area between the fourth hollow and the fifth hollow is defined as the rear crossbeam of the rear bushing.

[0036] Step S1.3: reconstructing a geometric model based on the control arm, performing mesh deformation parameterization, building a control arm parameterized model, and performing load condition analysis; In this example, the reconstructed model was parameterized using mesh deformation based on a finite element model. 77 mesh deformation spaces were divided and 200 control points were assigned to construct a parametric control arm model. Four load steps were created, corresponding to the two-wheel bump crossing, pothole, slope driving, and modal analysis conditions. In this example, the control arm parametric model was constructed using a finite element software module.

[0037] Step S1.4: Based on the reconstructed geometric model of the control arm, multiple key structural dimension parameters are used as design variables, and bending stiffness, first-order modal frequency, and mass are used as output responses. A control arm simulation data set is obtained through simulation; according to the wheelbase expansion architecture requirements of the chassis platform, multiple key structural dimension parameters of the control arm are set as design variables.

[0038] Specifically, in this embodiment, finite element and optimization software are used to jointly simulate and obtain data. In terms of the characteristic design variables and output responses of the control arm, a number of key structural dimension parameters are determined as design variables, and their adjustment can change the shape of the control arm and the wheelbase expansion capability, optimize the load transfer path, and the mesh quality still meets the requirements under extreme deformation. Among them, the adjustment of parameter SH1 can achieve the change of the length of the control arm body in the Y direction, which directly determines the wheelbase expansion capability of the chassis platform; the remaining parameters are used to adjust the size and thickness of the horizontal and longitudinal beam structure inside the control arm, etc., to optimize the load transfer path. As a preferred embodiment of this embodiment, simulation data is obtained by joint simulation of optimization software and finite element software.

[0039] As a preferred embodiment of this invention, 11 key structural dimension parameters are selected, and SH1 to SH11 are respectively set as the length of the front longitudinal beam of the spherical joint, the width of the front longitudinal beam of the front bushing, the width of the rear longitudinal beam of the front bushing, the width of the rear outer longitudinal beam of the rear bushing, the width of the inner crossbeam of the front bushing, the width of the inner crossbeam of the rear bushing, the width of the front crossbeam of the spherical joint, the width of the rear crossbeam of the spherical joint, the width of the front crossbeam of the rear bushing, the width of the rear crossbeam of the rear bushing, and the width of the front longitudinal beam of the spherical joint, as shown in Table 1: Table 1 Parametric design variables and variable control ranges

[0040] The output response selected is the hard point bending stiffness K_bend, the first-order modal frequency Mode_1, and the control arm mass Mass, taking into account the structural performance, dynamic characteristics, and lightweight requirements, as shown in Table 2: Table 2 Control arm output response

[0041] After determining key structural dimensional parameters and output responses, this embodiment uses the Optimal Latin Hypercube Sampling (OLHS) method to sample simulation data and obtain a control arm simulation dataset. An optimization algorithm is then used to further improve the uniformity of sample point distribution, reducing sampling error and improving sampling efficiency. In this embodiment, the control arm simulation dataset contains 1,000 complete sets of samples, each corresponding to a complete simulation analysis process. The Optimal Latin Hypercube Sampling (OLHS) method ensures uniform spatial distribution of sample points, avoiding sample aggregation or sparseness in local areas, and achieving full coverage of the design space.

[0042] Specifically, step S2: based on the control arm simulation data set, training a stacked ensemble learning model formed by multiple base learner models; including the following method steps: Step S2.1: Select multiple candidate base learner models, train them based on the control arm simulation data set, and obtain an optimized candidate base learner model; In this embodiment, the candidate base learner models include random forest, XGBoost, LightGBM and GBR, and the hyperparameter optimization of the candidate base learner models is achieved through Bayesian optimization.

[0043] Step S2.2: Based on the particle swarm optimization algorithm, the mean square error of the regression evaluation index of the stacked ensemble model is used as the fitness function, and the optimal optimization candidate base learner model combination is obtained through swarm intelligent search; Step S2.3: Based on the optimally optimized candidate base learner model combination, a two-layer stacked ensemble model is constructed using a k-fold crossover method. Specifically, in the first layer, the original dataset is partitioned using a 5-fold crossover method, and the optimal base learner combination is used to train and predict the partitioned data, generating a corresponding set of prediction results. In the second layer, the prediction results of the weighted concatenated base learners based on the PSO optimization weights are used as a new feature set. This new feature set is used as the input of the meta-learner, and the true labels are used as the supervision signal to train the LRM meta-learner, ultimately completing the construction of the stacked ensemble model BO-LRSM.

[0044] In this embodiment, for three learning tree models, Random Forest (RF), XGBoost, and LightGBM, four key hyperparameters are set for each model for optimization. The hyperparameter settings and their value ranges are shown in Table 3 below: Table 3 Hyperparameter settings of the base learner model

[0045] In this example, hyperparameter optimization of the three integrated models described above was performed on the regression task of the control arm ball joint lateral bending stiffness (Kbend). During the optimization process, a maximum number of iterations was set to 30 to ensure a thorough exploration of the hyperparameter space within a reasonable computational cost. The optimal parameter combinations obtained after optimization are shown in Table 4 below.

[0046] Table 4 Hyperparameter optimized values

[0047] It can be seen that Bayesian optimization significantly improves the prediction performance of RF, XGBoost, and LightGBM, and all three models show extremely high goodness of fit (R²>0.97).

[0048] In the PSO algorithm, the weights of each pair of candidate base learners are optimized. The goal is to minimize the mean squared error (MSE) of the combined model, constraining the weights to be within the range [0, 1] and summing to 1. Table 5 below shows the optimal base learner combinations selected by the PSO algorithm and their corresponding weights.

[0049] Table 5 PSO optimized base learner weight combination

[0050] The optimization results show that the combination of GBR and LGBM has the smallest MSE. The two complement each other in feature selection and data sampling, and the data distribution modeling is consistent, with good synergy. Therefore, in this example, GBR and LGBM are selected to construct the base learner combination of BO-LRSM.

[0051] A two-layer stacked ensemble model is constructed based on the k-fold crossover method. For the optimal base learner combination GBR and LGBM, the PSO optimization weight w = [0.6366, 0.3634], and the original predictions of the base learners M1 (GBR) and M2 (LGBM) can be expressed as: ; Integrating the optimal weight of PSO optimization into the weighted features can be expressed as: ; The prediction of the linear regression meta-learner LRM is expressed as: ; Among them, β1 and β2 represent the coefficients learned by the meta-learner LRM.

[0052] The weight optimization process for the weighted feature inputs of the stacked model is divided into two steps. The first step is to independently train the base learners and generate out-of-focus predictions using a k-fold crossover. The second step is to use the PSO-weighted concatenated predictions as a new feature set for training the LRM meta-learner. This approach enhances model robustness through weight prior optimization while maintaining the flexibility of the stacking framework. A 5-fold crossover control arm simulation dataset is used. The prediction results of the LGBM and GBR base learners on the corresponding data folds are used as the prediction set. The PSO-weighted concatenated prediction set is used as a new feature dataset for training the LRM meta-learner, thereby completing the construction of the control arm stacking prediction model BO-LRSM. A comparison of the three regression evaluation metrics of the final trained BO-LRSM model and the four candidate base learners under the three optimization objectives is shown in Table 6 below. The hyperparameters of the XGB, LGBM, and RF base learners have been Bayesian optimized.

[0053] Table 6 Comparison of BO-LRSM fitting evaluation indicators for control arm output response

[0054] From the comparative analysis of model fitting evaluation indicators, it can be seen that the BO-LRSM model after stacking has greatly improved the fitting accuracy of the three optimization objectives compared with any candidate basis learner. The R2 coefficient of the fitting accuracy for the first-order modal performance (Mode_1) with strong nonlinearity is above 0.9, and the R2 coefficient of the fitting accuracy for the control arm mass (Mass) and the spherical joint lateral stiffness (Kbend) is above 0.98. Therefore, the prediction effect of the stacked model meets the accuracy requirements for constructing a control arm prediction model.

[0055] Specifically, step S3: outputting the optimal parameter combination of the control arm through multi-objective optimization based on the stacked integrated model; including the following method steps: Step S3: Based on the stacked integrated model, the objective function is optimized by the MOMPA multi-objective marine predator algorithm to obtain the Pareto optimal frontier of the control arm output response, the optimal parameter point is selected by the K-Means clustering algorithm, and the optimal parameter combination of the control arm is output.

[0056] In this embodiment, the objective function of MIAOA optimization is expressed as: ; in, are the mass of the control arm structure, the lateral bending stiffness of the spherical joint and the first-order modal frequency of the control arm structure, respectively. m t 、 S b and f tare the initial mass of the control arm structure, the initial lateral bending stiffness of the spherical joint and the first-order initial modal frequency of the control arm structure; x i is the design parameter value, x 1, x 2,…, x 11 They represent the 11 selected control arm structural dimension design parameters SH1 to SH11, and the values of each parameter must meet the preset boundary constraints.

[0057] Specifically, in this embodiment, the MOMPA algorithm was used to perform multi-objective optimization of the control arm stack prediction model, contributing to the determination of the Pareto optimal frontier of structural parameters. The corresponding parameters of the MOMPA algorithm were adjusted, with the parameter value population size in the algorithm set to 500 and the maximum number of iterations set to 100.

[0058] This example uses the K-Means clustering algorithm to select the optimal parameter points from the Pareto optimal frontier obtained above. Data points are assigned to clusters so that the distance between each data point and the center (centroid) of its cluster is minimized, that is, the intra-cluster squared error is minimized: ; in, k Indicates the number of clusters; C i Indicates the i clusters; μ i Indicates the i The centroid of the cluster; x − μ i ∥ 2 Represents data points x and centroid μ i The square of the Euclidean distance between them.

[0059] The K-Means clustering algorithm can be used to divide the solutions on the Pareto front into several groups, and then a representative solution is selected from each group. Because the Pareto parameter combination includes the three dimensions of mass, stiffness, and mode, the cluster types are divided into three groups, namely three clusters. For each cluster, the distance from each point to the cluster center is calculated, and the point with the closest distance is selected as the representative solution. To maximize the lateral bending stiffness at the spherical joint, the point with the largest bending stiffness value among the three representative solutions is selected as the final optimal parameter point. Finally, the optimal parameter combination is determined from the Pareto front.

[0060] This embodiment provides a performance optimization design method for a vehicle's expandable chassis component structure. A multi-objective topology optimization mathematical model is established based on the expandable chassis strategy and the original control arm model. The universal force transmission path of the control arm is optimized, and a reconstructed geometric model of the control arm is optimized. A control arm simulation data set is obtained through simulation. Based on the control arm simulation data set, a stacked ensemble learning model constructed by multiple base learner models is trained. Based on the stacked ensemble model, the optimal control arm parameter combination is output through multi-objective optimization. The present invention organically combines topology optimization, parametric modeling, machine learning, and intelligent optimization algorithms. Topology optimization can achieve adaptive structural parameter design for different chassis component sizes. Multi-objective optimization of the model is achieved through the stacked ensemble model and a multi-objective optimization algorithm, thereby improving the efficiency of chassis component structural parameter optimization and significantly enhancing chassis component structural performance.

[0061] Example 2: This embodiment provides an expandable chassis component structure for a vehicle, wherein the chassis component structure includes a control arm, and the control arm structure is formed by the performance optimization design method described in the first embodiment.

[0062] In this embodiment, the arm body of the control arm is an integral triangle, and the three corners of the arm body respectively form a front bushing, a rear bushing and a ball joint.

[0063] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0064] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A performance optimization design method for a vehicle expandable chassis component structure, characterized in that: The following steps are involved: Determine the design wheelbase of the scalable chassis and determine the dimensional parameters of the original control arm model based on the design wheelbase; A multi-objective topology optimization mathematical model is established based on the original model of the control arm, the universal force transmission path of the control arm is optimized, the reconstructed geometric model of the control arm is optimized, and a simulation data set of the control arm is obtained through simulation, including: obtaining the original model of the control arm, defining the design area and non-design area of the original model of the control arm; creating an optimization space in the design area by a solid filling method; establishing a multi-objective topology optimization mathematical model, setting the minimization of multi-working condition strain energy and the maximization of natural frequency as optimization goals at the same time, optimizing the optimization space by a compromise programming method, and obtaining the universal force transmission path of the control arm; based on the universal force transmission path of the control arm, a reconstructed geometric model of the control arm is optimized by three-dimensional modeling software; based on the reconstructed geometric model of the control arm, multiple key structural size parameters are used as design variables, and bending stiffness, first-order modal frequency, and mass are used as output responses, and a simulation data set of the control arm is obtained through simulation; based on the simulation data set of the control arm, a stacked ensemble learning model formed by multiple base learner models is trained; based on the stacked ensemble model, the optimal parameter combination of the control arm is output through multi-objective optimization.

2. The performance optimization design method for a vehicle expandable chassis component structure according to claim 1, characterized in that: When defining the design area and the non-design area of the original model of the control arm, the ball joint, the bushing and the arm body contour are defined as the non-design area, and the interior of the arm body is defined as the design area.

3. The performance optimization design method for a vehicle expandable chassis component structure according to claim 1, characterized in that: When establishing a multi-objective topology optimization mathematical model, the vertical degrees of freedom of the ball joint connection, the lateral and vertical translational degrees of freedom of the rear bushing connection, and the lateral, vertical, and longitudinal translational degrees of freedom of the front bushing connection are taken as model constraints; the draft constraint and the maximum and minimum member size constraints are also included as constraints.

4. The performance optimization design method for a vehicle expandable chassis component structure according to claim 1, characterized in that: The multi-objective topology optimization mathematical model is expressed as: ； in, k Indicates the number of working conditions, ω k Indicates the k The flexibility weighting factor of each working condition; p represents the order of the natural frequency, C i ( x ) indicates the i The initial flexibility of the structure under each working condition is C imax and C imin Indicates the i The maximum and minimum values of flexibility under each working condition; f p ( x ) indicates the p The natural frequency of the initial structure of the order; f pmax and f pmin The first p The upper and lower limits of the order frequency; ω Represents the weight of the flexibility function; the working conditions include the two-wheel over bump condition, the two-wheel over pothole condition and the two-wheel over slope condition.

5. The performance optimization design method for a vehicle expandable chassis component structure according to claim 1, characterized in that: The method also includes the following steps: reconstructing a geometric model based on the control arm, performing mesh deformation parameterization, building a parametric model of the control arm, and performing load condition analysis, including: creating four load steps corresponding to two-wheel driving over bumps, potholes, slope driving, and modal analysis conditions.

6. The performance optimization design method for a vehicle expandable chassis component structure according to claim 1, characterized in that: Based on the control arm simulation data set, a stacked ensemble learning model formed by multiple base learner models is obtained through training, including the following method steps: multiple candidate base learner models are selected, and the optimized candidate base learner model is obtained through training based on the control arm simulation data set; based on the particle swarm optimization algorithm, the mean square error of the regression evaluation index of the stacked ensemble model is used as the fitness function, and the optimal optimized candidate base learner model combination is obtained through swarm intelligent search; based on the optimal optimized candidate base learner model combination, a two-layer stacked ensemble model is constructed through the k-fold crossover method.

7. The performance optimization design method for a vehicle expandable chassis component structure according to claim 1, characterized in that: Based on the stacked integrated model, the optimal parameter combination of the control arm is output through multi-objective optimization, including the following method steps: based on the stacked integrated model, the objective function is multi-objective optimized through the MOMPA multi-objective marine predator algorithm to obtain the Pareto optimal frontier of the control arm output response, the optimal parameter point is selected through the K-Means clustering algorithm, and the optimal parameter combination of the control arm is output.

8. The performance optimization design method for a vehicle expandable chassis component structure according to claim 1, characterized in that: Also includes: The parameter combination of the pareto optimal frontier includes three dimensions: bending stiffness, first-order modal frequency, and mass. The three dimensions are divided into three clusters using the K-Means clustering algorithm. For each cluster, the distance from each point to the cluster center is calculated, and the point with the closest distance is selected as the representative solution, that is, the optimal parameter point.

9. A vehicle expandable chassis component structure, characterized in that: The chassis component structure includes a control arm, and the control arm structure is formed by the performance optimization design method according to any one of claims 1 to 8.

10. The vehicle expandable chassis component structure according to claim 9, characterized in that: The arm body of the control arm is an integral triangle, and the three corners of the arm body respectively form a front bushing, a rear bushing and a ball joint.

Images