Dynamic performance and bandwidth optimization method for extensible chassis of vehicle

Abstract

The invention discloses a dynamic performance and bandwidth optimization method for an extensible chassis of a vehicle, and the method comprises the steps: building an extensible chassis multi-body dynamic joint simulation model, obtaining a dynamic performance data set under different chassis configurations, building an extensible chassis dynamic performance high-precision prediction agent model based on an integrated learning algorithm, and carrying out the optimization of the dynamic performance and bandwidth of the extensible chassis. And carrying out optimization solution through a multi-objective optimization algorithm so as to obtain an optimized vehicle chassis structure. According to the method, efficient modeling of the vehicle chassis is achieved based on the extensible chassis multi-body dynamics joint simulation model and based on the architecture characteristics and actual working condition requirements of the extensible chassis, meanwhile, through the extensible chassis dynamics performance high-precision prediction agent model, the dynamics performance prediction precision and generalization ability are improved, and the method is suitable for large-scale popularization and application. And the chassis structure is optimized through a multi-objective optimization algorithm, so that the overall dynamic performance of the vehicle is improved, the optimization efficiency is improved, and diversified optimization requirements of various vehicle model architectures can be met.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle chassis structure optimization design, and in particular relates to a method for optimizing the dynamic performance and bandwidth of an expandable vehicle chassis. Background Art

[0002] In automotive R&D, vehicle model expansion based on platform architecture has become a mainstream approach to improve R&D efficiency and reduce production costs. Leveraging platform-based and modular design concepts, automakers are able to derive multiple models from a single platform. Optimizing dynamic performance is crucial in the development of vehicle chassis structures. Current approaches to optimizing vehicle chassis structures often rely on iterative parametric model simulations, which can be computationally complex, inefficient, and require significant computing resources, resulting in lengthy and costly development cycles.

[0003] Therefore, existing technologies utilize intelligent learning and optimization techniques to develop and optimize vehicle chassis structures. However, due to the high dimensionality of the optimization variables and their strong nonlinear relationship with dynamic performance, traditional machine learning methods and multi-objective optimization algorithms struggle to achieve ideal optimization results. Traditional machine learning methods struggle to effectively solve complex, high-dimensional problems, while multi-objective optimization algorithms have difficulty optimizing with strong nonlinear relationships. Furthermore, existing traditional parametric model simulation and iteration methods lack generalizability across different vehicle models and operating conditions, making it difficult to meet diverse market demands. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the defects of low efficiency and poor performance of the performance optimization method of the vehicle chassis structure in the prior art, thereby providing a method for optimizing the dynamic performance and bandwidth of an expandable vehicle chassis.

[0005] A method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis includes the following steps: Establishing a scalable chassis multi-body dynamics co-simulation model, and obtaining dynamics performance data sets under different chassis configurations based on the scalable chassis multi-body dynamics co-simulation model; the dynamics performance data sets include multiple sets of corresponding design variables and corresponding optimization target performance parameters; the design variables include the horizontal positions of shock absorbers, lower control arms, and steering rods in the chassis; and the optimization target performance parameters include the kinematic states of the chassis under various operating conditions; Based on the dynamic performance data set, a high-precision prediction agent model for scalable chassis dynamic performance based on an ensemble learning algorithm is established to predict the optimization target performance parameters corresponding to the design variables for the scalable chassis multi-body dynamics joint simulation model; multiple optimization target performance parameters are predicted by multiple agent models composed of base learners; Through a multi-objective optimization algorithm, the combination of design variables is defined as individuals in a population; in the first iteration, the random control population is guided to move to a more optimal area based on the positions of the better individuals and the best individual, or a random directional disturbance is generated for the population based on the positions of the worst individual and the best individual; in the second iteration, the random control population is guided to move to the positions of the top three best individuals based on the positions of the top three best individuals, or a reference point is generated based on the boundary of the solution space to guide the individuals to diffuse toward the boundary of the solution space; the high-precision prediction agent model of the scalable chassis dynamics performance is optimized and solved.

[0006] Furthermore, an extensible chassis multi-body dynamics joint simulation model is established, and dynamic performance data sets under different chassis configurations are obtained based on the extensible chassis multi-body dynamics joint simulation model; the parameters of the extensible chassis multi-body dynamics joint simulation model include wheelbase, front track, rear track, curb weight, front and rear axle weights, and front and rear stiffness; chassis configurations of different sizes are selected, and simulations are performed on four working conditions: straight-line acceleration, braking, steady-state turning, and serpentine driving, to obtain dynamic performance data sets.

[0007] Furthermore, an extensible chassis multi-body dynamics joint simulation model is established, and a dynamics performance data set under different chassis configurations is obtained based on the extensible chassis multi-body dynamics joint simulation model; the design variables of the dynamics performance data set include the upper horizontal coordinates of the shock absorbers at the front and rear of the chassis, the outer horizontal coordinates of the lower control arms at the front and rear of the chassis, and the outer horizontal coordinates of the steering rods at the front and rear of the chassis; the optimized target performance parameters include the pitch angle gradient during straight-line acceleration and braking, the maximum lateral acceleration during steady-state rotation, the roll angle gradient and steering angle gradient at a lateral acceleration of 0.2g, the average roll angle during serpentine driving, the average steering angle of the steering wheel, and the average yaw angular velocity.

[0008] Furthermore, based on the dynamic performance dataset, a scalable chassis dynamic performance high-precision prediction agent model based on an ensemble learning algorithm is established; six learners, namely random forest, extreme random tree, gradient boosting tree algorithm, XGBoost, LightGBM, and Catboost, are selected as base learners, and a linear regression algorithm is selected as a second-layer classifier; it also includes: training and predicting the base learners based on the dynamic performance dataset, obtaining predicted values ​​and calculating evaluation indicators of each base learner, and screening the top three base learners according to the determination coefficient; for the top three base learners, their predicted values ​​for the dynamic performance dataset are used as input, and the true values ​​of the dynamic performance dataset are used as output, and the linear regression algorithm is used as a meta-model for training to obtain an integrated model, which can be used to expand the chassis dynamic performance high-precision prediction agent model.

[0009] Furthermore, the high-precision prediction agent model of the scalable chassis dynamic performance is optimized and solved through a multi-objective optimization algorithm, including the following method steps: initializing the population by generating a chaotic sequence with high spatial uniformity; before the current number of iterations reaches half of the total number of iterations, the algorithm performs an extensive search in the solution space; after the current number of iterations reaches half of the total number of iterations, the algorithm performs a detailed search in the solution space; and solving through random population iterations.

[0010] Furthermore, the population is initialized by generating a chaotic sequence with high spatial uniformity, including: generating a chaotic sequence; mapping the generated chaotic sequence to the initialized population to complete the population initialization; the individuals of the population represent the combination of the design variables; The chaotic sequence is generated as follows: ; in, Indicates the j+1 A chaotic sequence, Indicates the j A chaotic sequence; Mapping the generated chaotic sequence to the initialized population is expressed as: ; Among them, i represents the population number, N represents the number of populations, j represents the population dimension number, and D represents the dimension of the population; Transform the population according to its upper and lower bounds to complete the population initialization: ; in, represents the i-th population, is the j-th dimension variable value of the i-th population, and L They represent the maximum and minimum values ​​of the population on the jth dimension respectively.

[0011] Furthermore, before the current number of iterations reaches half of the total number of iterations, the algorithm performs an extensive search in the solution space, which can be expressed as: If JF <0.5: ; Else: ; ; Among them, JF represents a random number between 0 and 1, represents the current position of the population, r, and Represent random numbers from 0 to 1 respectively; Represents the average position information of the top 50% of the fitness value group; represents a set of random numbers based on Brownian motion; represents the optimal individual; Indicates the worst individual; represents the adaptive control factor; t represents the current number of iterations, and T represents the maximum number of iterations.

[0012] Furthermore, after the current number of iterations reaches half of the total number of iterations, the algorithm performs a fine search in the solution space, which is expressed as: If JF <0.5: ; Else: ; in, , and represent the best, second best and third best individuals respectively, Represents a set of random numbers based on the Lévy motion.

[0013] Furthermore, the solution is obtained through random population iteration, which is expressed as: ; Among them, RF represents A random number within and Represent a random population respectively.

[0014] A vehicle expandable chassis is formed by optimizing the above-mentioned dynamic performance and bandwidth optimization method.

[0015] Beneficial effects: The present invention discloses a method for optimizing the dynamic performance and bandwidth of a vehicle's expandable chassis, establishes a multi-body dynamics joint simulation model for an expandable chassis, obtains dynamic performance data sets under different chassis configurations, establishes a high-precision prediction agent model for the dynamic performance of an expandable chassis based on an ensemble learning algorithm, and optimizes and solves the problem through a multi-objective bionic archerfish optimization algorithm, thereby obtaining an optimized vehicle chassis structure. The present invention is based on a multi-body dynamics joint simulation model for an expandable chassis, and based on the architectural characteristics and actual working conditions of the expandable chassis, achieves efficient modeling of the vehicle chassis. At the same time, through a high-precision prediction agent model for the dynamic performance of an expandable chassis, the accuracy and generalization ability of the dynamic performance prediction are improved, and through a multi-objective bionic archerfish optimization algorithm, the chassis structure is optimized, significantly improving the overall dynamic performance of the vehicle, effectively improving the optimization efficiency, and being able to adapt to the diverse optimization needs of various vehicle architectures. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] 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.

[0017] Figure 1 Schematic diagram of the main process of the present invention; Figure 2 This is a schematic diagram of the expandable chassis structure of a vehicle according to the present invention.

[0018] Description of the drawings: 1. Front shock absorber; 2. Rear shock absorber; 3. Front lower control arm; 4. Rear lower control arm; 5. Front steering rod; 6. Rear steering rod. DETAILED DESCRIPTION

[0019] 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.

[0020] 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.

[0021] 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.

[0022] It should be noted that when an element is referred to as being "fixed to" or "disposed on" another element, it may be directly on the other element or there may be an intermediate element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be an intermediate element. The terms "vertical," "horizontal," "upper," "lower," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only implementation methods.

[0023] Example 1: Reference Figure 1 As shown, this embodiment provides a method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis, including the following method steps: Step S1: establishing a scalable chassis multi-body dynamics co-simulation model, and obtaining dynamics performance data sets under different chassis configurations based on the scalable chassis multi-body dynamics co-simulation model; the dynamics performance data sets include multiple sets of corresponding design variables and corresponding optimization target performance parameters; the design variables include the horizontal positions of the shock absorber, the lower control arm, and the steering rod in the chassis, and the optimization target performance parameters include the kinematic states of the chassis under various working conditions; Step S2: Based on the dynamic performance data set, predicting the optimization target performance parameters corresponding to the design variables for the scalable chassis multi-body dynamics joint simulation model, and establishing a high-precision prediction agent model for scalable chassis dynamic performance based on an ensemble learning algorithm; Step S3: The combination of design variables is defined as individuals of a population through a multi-objective bionic archerfish optimization algorithm; in the first iteration, the random control population is guided to move to a more optimal area based on the positions of the better individuals and the best individuals, or a random directional disturbance is generated for the population based on the positions of the worst individual and the best individual; in the second iteration, the random control population is guided to move to the positions of the top three best individuals based on the positions of the top three best individuals, or a reference point is generated based on the boundary of the solution space to guide the individuals to diffuse to the boundary of the solution space; the high-precision prediction agent model of the scalable chassis dynamics performance is optimized and solved.

[0024] In this application, bandwidth optimization refers to the rational adjustment and optimization of the chassis' performance range and adaptability to better meet the needs of different vehicle models and application scenarios. This embodiment optimizes dynamic performance by modifying the basic model to establish a scalable chassis multibody dynamics co-simulation model. This can meet the adaptive optimization of different performance ranges and adaptability under different chassis configurations, thereby achieving bandwidth optimization.

[0025] Specifically, step S1: establishing an extensible chassis multi-body dynamics co-simulation model, and acquiring dynamics performance data sets under different chassis configurations based on the extensible chassis multi-body dynamics co-simulation model; The scalable chassis multibody dynamics co-simulation model is obtained by modifying parameters of a base model, which is a standard template provided by multibody dynamics analysis software. The parameters of the scalable chassis multibody dynamics co-simulation model include wheelbase, front track, rear track, curb weight, front and rear axle weights, and front and rear stiffness. Different chassis configurations are selected and simulated for four operating conditions: straight-line acceleration, braking, steady-state turning, and serpentine driving, to obtain a dynamics performance data set. As a preferred embodiment of this embodiment, batch simulation and data processing are performed using the ISIGHT platform to integrate multibody dynamics analysis software and MATLAB to obtain a dynamics performance data set.

[0026] Reference Figure 2 As shown, in this embodiment, the front suspension selects the McPherson suspension and the rear suspension selects the double wishbone independent suspension. After completing the modeling of each subsystem, the multi-body dynamics model of the whole vehicle is obtained.

[0027] In this embodiment, as shown in Table 1, the design variables of the dynamic performance data set include the upper horizontal coordinates of the shock absorber at the front of the chassis (f_top_mount_x, f_top_mount_y), the upper horizontal coordinates of the shock absorber at the rear of the chassis (r_top_mount_x, r_top_mount_y), the outer horizontal coordinates of the lower control arm at the front of the chassis (f_lca_outer_x, f_lca_outer_y), the outer horizontal coordinates of the lower control arm at the rear of the chassis (r_lca_outer_x, r_lca_outer_y), the outer horizontal coordinates of the steering rod at the front of the chassis (f_tierod_outer_x, f_tierod_outer_y), and the outer horizontal coordinates of the steering rod at the rear of the chassis (r_tierod_outer_x, r_tierod_outer_y).

[0028] In this embodiment, the upper horizontal coordinate of the shock absorber at the front of the chassis represents the top coordinate of the right front shock absorber, the upper horizontal coordinate of the shock absorber at the rear of the chassis represents the top coordinate of the right rear shock absorber, the outer horizontal coordinate of the lower control arm at the front of the chassis represents the rightmost front coordinate of the right front lower control arm, the outer horizontal coordinate of the lower control arm at the rear of the chassis represents the rightmost front coordinate of the right rear lower control arm, the outer horizontal coordinate of the steering rod at the front of the chassis represents the rightmost point coordinate of the right front steering rod, and the outer horizontal coordinate of the steering rod at the rear of the chassis represents the rightmost point coordinate of the right rear steering rod. As a preferred embodiment of this embodiment, the positions of the shock absorber, lower control arm and steering rod on the left side of the chassis are obtained by symmetrical calculation of the right coordinates. In this embodiment, the front, back, up, down, left and right azimuth descriptions representing the orientation are based on the orientation of the vehicle body itself.

[0029] Table 1 Coordinate parameter names of each design variable

[0030] As shown in Table 2, the optimization target performance parameters include the pitch angle gradient acc_pitch_slope during linear acceleration and the pitch angle gradient bra_pitch_slope during braking, the maximum lateral acceleration max_ay during steady-state turning, the roll angle gradient roll_angle_slope and the steering angle gradient steer_slope at a lateral acceleration of 0.2g, the average roll angle roll_angle_mean, the average steering wheel steering angle steer_mean, and the average yaw angular velocity yaw_mean during serpentine driving.

[0031] Table 2 Performance indicators corresponding to each optimization goal

[0032] Specifically, step S2: based on the dynamic performance data set, establishing a scalable chassis dynamic performance high-precision prediction agent model based on an ensemble learning algorithm; Six learners, namely random forest, extreme random tree, gradient boosting tree algorithm, XGBoost, LightGBM and Catboost, are selected as base learners, and a linear regression algorithm is selected as a second-layer classifier; the method also includes: training and predicting the base learners based on the dynamic performance data set, obtaining predicted values ​​and calculating evaluation indicators of each base learner, and screening the top three base learners according to the determination coefficient R²; for the top three base learners, the predicted values ​​of the dynamic performance data set are used as input, and the true values ​​of the dynamic performance data set are used as output, and the linear regression algorithm is used as a meta-model for training to obtain an integrated model, thereby expanding the high-precision prediction agent model of chassis dynamic performance.

[0033] Ensemble machine learning algorithms such as random forests (RF), extreme random trees (ET), gradient boosted tree algorithm (GBDT), XGBoost, LightGBM, and Catboost offer different advantages. RF, for example, constructs multiple decision trees using bootstrap sampling and averages the predicted values ​​during regression. This effectively reduces the risk of overfitting in individual trees and is robust when dealing with nonlinear, high-dimensional data and complex feature interactions. ET has a similar structure to RF, but its feature partitioning points are completely random, resulting in higher computational efficiency and strong generalization, making it suitable for high-dimensional, nonlinear data and large-scale training tasks. GBDT employs an additive model, using decision trees as base learners and gradually optimizing the residuals via gradient descent. This effectively captures complex feature interactions and excels at processing high-dimensional, sparse data. XGBoost builds on GBDT by introducing L1 and L2 regularization terms to control model complexity and prevent overfitting. It also supports parallel computing and automatically handles missing values ​​and outliers. LightGBM, based on leaf node splitting, offers fast training speed and strong processing capabilities for categorical features, without requiring additional coding preprocessing. Catboost has unique advantages in processing categorical features, can automatically perform target encoding, and is highly robust to missing values ​​and data noise.

[0034] In this example, based on a dynamic performance dataset covering four operating conditions for an scalable chassis, the constructed adaptive stacking ensemble model was used to establish dynamic performance response proxy models. The dynamic performance responses of the test set samples were then predicted. As shown in Table 3, most of the dynamic performance response proxy models included the CatBoost (CBR) algorithm in their base model combinations, with the exception of the proxy model for the average steering angle in the serpentine condition. This demonstrates the strong generalization capability of the CBR algorithm. XGBoost (XGBR) and LightGBM (LGBR) algorithm models also appeared frequently, indicating that gradient-based algorithms are more suitable for establishing proxy models for high-dimensional problems. RF, ET, and GBDT also appeared in the base model combination, demonstrating that the adaptive stacking ensemble algorithm effectively integrates the differences among the base models and selects the base model that is appropriate for the current dataset.

[0035] Table 3 Base model filtering results

[0036] Table 4 shows the accuracy prediction results for each surrogate model. The adaptive stacking ensemble algorithm achieves high accuracy for predicting various dynamic performance characteristics. Predicted values ​​for linear acceleration and braking conditions are close to the true values. While the accuracy for steady-state turning and serpentine driving conditions is slightly lower, it remains high. For conditions with strong nonlinearity, the prediction accuracy reaches 0.86 or higher. Furthermore, the response time for predicting dynamic performance using surrogate models is short, significantly improving efficiency without sacrificing computational accuracy compared to finite element simulation software.

[0037] Table 4 Prediction accuracy of each proxy model

[0038] Specifically, step S3: Optimizing and solving the high-precision prediction agent model for the scalable chassis dynamics performance using a multi-objective bionic archerfish optimization algorithm includes the following method steps: initializing the population by generating a chaotic sequence with high spatial uniformity; before the current iteration number reaches half of the total number of iterations, the algorithm performs an extensive search in the solution space; after the current iteration number reaches half of the total number of iterations, the algorithm performs a detailed search in the solution space; during the development phase, defining an escape mechanism to simulate the random swimming behavior of an archerfish after being startled by a bird. In this embodiment, the following method steps are included: Step S3.1: Initializing the population by generating a chaotic sequence with high spatial uniformity, including: generating a chaotic sequence; mapping the generated chaotic sequence to the initialized population to complete the population initialization; The chaotic sequence is generated as follows: ; in, Indicates the j A chaotic sequence; Mapping the generated chaotic sequence to the initialized population is expressed as: ; Among them, i represents the population number, N represents the number of populations, j represents the population dimension number, and D represents the dimension of the population; Transform the population according to its upper and lower bounds to complete the population initialization: ; in, represents the i-th population, is the j-th dimension variable value of the i-th population, and L They represent the maximum and minimum values ​​of the population on the jth dimension respectively.

[0039] S3.2: Before the current iteration number reaches half of the total number of iterations, the algorithm performs an extensive search in the solution space, expressed as: If JF <0.5: ; Else:

[0040]

[0041] Among them, JF is a random number between 0 and 1, represents the current position of the population, r, and Both represent a random number between 0 and 1. It represents the average position information of the top 50% of the fitness value group. Represents a collection of random numbers based on Brownian motion. represents the optimal individual. Indicates the worst individual. is the adaptive control factor used to control the step size of reverse learning. t is the current iteration number, and T is the maximum iteration number.

[0042] After the current number of iterations reaches half of the total number of iterations, the algorithm performs a fine search in the solution space, which is expressed as: If JF <0.5: ; Else:

[0043] in, , and represent the best, second best and third best individuals respectively, Represents a set of random numbers based on the Lévy motion.

[0044] S3.3: During the development phase, archerfish will randomly dart around in the water after being startled by birds. Based on this, this paper constructs an escape mechanism, which is expressed as:

[0045] Among them, RF represents A random number within, and Represent a random population respectively.

[0046] In this embodiment, the quality of the Pareto front is further optimized by fast non-dominated sorting and crowding distance calculation. Fast non-dominated sorting is a commonly used sorting technique in multi-objective optimization algorithms. Its principle is to hierarchically divide the individuals in the population according to the Pareto dominance relationship. This embodiment achieves efficient sorting through a dual statistical mechanism: first, the dominating solution set and the dominated frequency of each individual are calculated, and then a complete dominance relationship network is constructed; then, hierarchical processing is implemented based on the network to divide the solution set into multiple non-dominated frontier layers. This hierarchical strategy can efficiently identify the Pareto optimal solution set in the population, thereby effectively improving the convergence efficiency and solution set quality of the multi-objective optimization process. The specific sorting steps are as follows: Step 1: Set two parameters for all individuals p in the population to control the number of solutions to p and the set of solutions dominated by p ; Step 2: Let i=1, The liberation of the collection ; Step 3: For the , and traverse each of the solutions , and Each solution in minus one; Step 4: Let i=i+1, and change the current The liberation of ; Step 5: Repeat Step 3 and Step 4 until all solutions have their own level sets.

[0047] After completing the non-dominated hierarchical sorting of individuals in the population, it is necessary to further evaluate the distribution density of individuals within the same non-dominated layer. Crowding distance is a key indicator for measuring the distribution of the solution set. The larger its value, the better the distribution sparsity of the solution in the target space. Based on this, the optimal solution archiving strategy adopts a double screening mechanism: individuals in lower non-dominated layers are given priority; for individuals in the same layer, secondary screening is performed based on the crowding distance, and sparsely distributed solutions are retained first. This archiving process continues until the preset archiving capacity limit is reached, thereby ensuring that the final solution set achieves a balance between convergence and distribution. Its calculation steps are expressed as: Step 1: For each objective function, sort the solutions in the same level from small to large according to the objective function value; Step 2: For the first and last solutions in each level, record their objective function values ​​as min and max respectively, and set their crowding distance to infinity to ensure that the boundary solutions are not excluded; Step 3: For each sorted objective function, calculate the distance between each solution and its adjacent solutions, that is, the difference between the i-th solution and the i-1th and i+1th solutions on the objective function, and normalize the difference using max and min; Step 4: For each solution, sum its distances on all objective functions to obtain the crowding distance of the solution; Step 5: Calculate the crowding distance of the next level.

[0048] The calculation method is expressed as: .

[0049] In this example, the population is divided into several Pareto Fronts. Each individual is assigned a non-dominated rank R(x) based on its objective function value f(x) = [f1(x), f2(x), ..., fm(x)]. A smaller value indicates better fitness. If the ranks are the same, the individual with the higher crowding distance is selected as the final solution for the design variable. Fitness = non-dominated rank + crowding distance, expressed as: Fitness(x)=(R(x),-CD(x)).

[0050] This embodiment provides a method for optimizing the dynamic performance and bandwidth of a vehicle's expandable chassis. This method establishes a multi-body dynamics co-simulation model for the expandable chassis, obtains dynamic performance data sets under different chassis configurations, establishes a high-precision prediction proxy model for the expandable chassis' dynamic performance based on an ensemble learning algorithm, and performs optimization and solution using a multi-objective bionic archerfish optimization algorithm to obtain an optimized vehicle chassis structure. The present invention is based on the expandable chassis multi-body dynamics co-simulation model, the architectural characteristics of the expandable chassis, and actual operating requirements to achieve efficient modeling of the vehicle chassis. The high-precision prediction proxy model for the expandable chassis' dynamic performance improves the accuracy and generalization of dynamic performance predictions, and the multi-objective bionic archerfish optimization algorithm optimizes the chassis structure, significantly improving the vehicle's overall dynamic performance and effectively increasing optimization efficiency. This method can adapt to the diverse optimization requirements of various vehicle architectures.

[0051] Example 2: This embodiment provides an expandable vehicle chassis, which is optimized by the dynamic performance and bandwidth optimization method described in the first embodiment.

[0052] 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.

[0053] 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 method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis, characterized in that: The method comprises the following steps: establishing a scalable chassis multi-body dynamics co-simulation model, and obtaining a dynamics performance data set under different chassis configurations based on the scalable chassis multi-body dynamics co-simulation model; the dynamics performance data set includes multiple sets of corresponding design variables and corresponding optimization target performance parameters; the design variables include the horizontal positions of the shock absorber, the lower control arm, and the steering rod in the chassis, and the optimization target performance parameters include the motion state of the chassis under various working conditions; based on the dynamics performance data set, establishing a high-precision prediction agent model for scalable chassis dynamics performance based on an ensemble learning algorithm, and realizing prediction of the optimization target performance parameters corresponding to the design variables for the scalable chassis multi-body dynamics co-simulation model; Through a multi-objective optimization algorithm, the combination of design variables is defined as individuals in a population; in the first iteration, the random control population is guided to move to a more optimal area based on the positions of the better individuals and the best individual, or a random directional disturbance is generated for the population based on the positions of the worst individual and the best individual; in the second iteration, the random control population is guided to move to the positions of the top three best individuals based on the positions of the top three best individuals, or a reference point is generated based on the boundary of the solution space to guide the individuals to diffuse toward the boundary of the solution space; the high-precision prediction agent model of the scalable chassis dynamics performance is optimized and solved.

2. The method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis according to claim 1, characterized in that: A scalable chassis multi-body dynamics co-simulation model is established, and dynamic performance data sets under different chassis configurations are obtained based on the scalable chassis multi-body dynamics co-simulation model; the parameters of the scalable chassis multi-body dynamics co-simulation model include wheelbase, front track, rear track, curb weight, front and rear axle weights, and front and rear stiffness; chassis configurations of different sizes are selected, and simulations are performed on four working conditions: straight-line acceleration, braking, steady-state turning, and serpentine driving to obtain dynamic performance data sets.

3. The method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis according to claim 1, characterized in that: An extensible chassis multi-body dynamics co-simulation model is established, and dynamics performance data sets under different chassis configurations are obtained based on the extensible chassis multi-body dynamics co-simulation model; the design variables of the dynamics performance data sets include the upper horizontal coordinates of the shock absorbers at the front and rear of the chassis, the outer horizontal coordinates of the lower control arms at the front and rear of the chassis, and the outer horizontal coordinates of the steering rods at the front and rear of the chassis; the optimized target performance parameters include the pitch angle gradient during straight-line acceleration and braking, the maximum lateral acceleration during steady-state rotation, the roll angle gradient and steering angle gradient at a lateral acceleration of 0.2g, the average roll angle during serpentine driving, the average steering angle of the steering wheel, and the average yaw angular velocity.

4. The method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis according to claim 1, characterized in that: Based on the dynamic performance data set, an extensible chassis dynamic performance high-precision prediction agent model based on an ensemble learning algorithm is established; six learners, including random forest, extreme random tree, gradient boosting tree algorithm, XGBoost, LightGBM, and Catboost, are selected as base learners, and a linear regression algorithm is selected as a second-layer classifier; it also includes: training and predicting the base learners based on the dynamic performance data set, obtaining predicted values ​​and calculating evaluation indicators of each base learner, and screening the top three base learners according to the determination coefficient; for the top three base learners, their predicted values ​​for the dynamic performance data set are used as input, and the true values ​​of the dynamic performance data set are used as output, and a linear regression algorithm is used as a meta-model for training to obtain an integrated model, thereby expanding the chassis dynamic performance high-precision prediction agent model.

5. The method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis according to claim 1, characterized in that: The high-precision prediction agent model for the scalable chassis dynamics performance is optimized and solved through a multi-objective optimization algorithm, including the following method steps: initializing the population by generating a chaotic sequence with high spatial uniformity; before the current number of iterations reaches half of the total number of iterations, the algorithm performs an extensive search in the solution space; after the current number of iterations reaches half of the total number of iterations, the algorithm performs a detailed search in the solution space; and solving through random population iterations.

6. The method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis according to claim 5, characterized in that: Initializing a population by generating a chaotic sequence with high spatial uniformity, including: generating a chaotic sequence; mapping the generated chaotic sequence to an initialized population to complete the population initialization; individuals in the population represent combinations of design variables; The chaotic sequence is generated as follows: ； in, Indicates the j+1 A chaotic sequence, Indicates the j A chaotic sequence; Mapping the generated chaotic sequence to the initialized population is expressed as: ； Among them, i represents the population number, N represents the number of populations, j represents the population dimension number, and D represents the dimension of the population; Transform the population according to its upper and lower bounds to complete the population initialization: ； in, represents the i-th population, is the j-th dimension variable value of the i-th population, and L They represent the maximum and minimum values ​​of the population on the jth dimension respectively.

7. The method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis according to claim 6, characterized in that: Before the current number of iterations reaches half of the total number of iterations, the algorithm performs an extensive search in the solution space, which can be expressed as: If JF <0.5: ； Else: ； ； Among them, JF represents a random number between 0 and 1, represents the current position of the population, r, and Represent random numbers from 0 to 1 respectively; Represents the average position information of the top 50% of the fitness value group; represents a set of random numbers based on Brownian motion; represents the optimal individual; Indicates the worst individual; represents the adaptive control factor; t represents the current number of iterations, and T represents the maximum number of iterations.

8. The method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis according to claim 7, characterized in that: After the current number of iterations reaches half of the total number of iterations, the algorithm performs a fine search in the solution space, which is expressed as: If JF <0.5: ； Else: ； in, , and represent the best, second best and third best individuals respectively, Represents a set of random numbers based on the Lévy motion.

9. The method for optimizing the dynamic performance and bandwidth of a vehicle expandable chassis according to claim 8, characterized in that: Solved by random population iteration, expressed as: ； Among them, RF represents A random number within and Represent a random population respectively.

10. An expandable chassis for a vehicle, characterized in that: It is formed by optimizing the dynamic performance and bandwidth optimization method according to any one of claims 1 to 9.

Images