CN-121980690-A - Modelica-based automobile single-quality system modeling and self-adaptive parameter optimization method

CN121980690ACN 121980690 ACN121980690 ACN 121980690ACN-121980690-A

Abstract

The invention relates to the field of automobile dynamics simulation, and particularly discloses a model-based automobile single-mass system modeling and self-adaptive parameter optimization method, which comprises the steps of determining a simplified module and a component connection mode of an automobile single-mass system according to demand analysis, and constructing a basic simulation model based on model; the method comprises the steps of carrying out initial simulation on a basic simulation model to obtain a simulation output result under initial parameters, collecting actual measurement data of a real vehicle, constructing a comprehensive error function, carrying out iterative optimization on key physical parameters of the basic simulation model by adopting a parameter optimization engine until convergence conditions are met to obtain optimized parameters, and feeding back the optimized parameters to the basic simulation model. The invention integrates the advantages of Modelica multi-field physical modeling and a data-driven intelligent optimization algorithm, realizes the span from artificial experience parameter adjustment to automatic intelligent optimization, obviously reduces the dependence on accurate initial parameters, and effectively improves the model simulation precision and engineering application efficiency.

Inventors

LI JUNJIE
LI YANZHEN
WU MANTING
DONG CHAO
QIU JIBIN
Yan Jiashuo
ZHOU SHENGXIN

Assignees

烟台大学

Dates

Publication Date: 20260505
Application Date: 20260408

Claims (10)

1. The model-based automobile single-quality system modeling and self-adaptive parameter optimization method is characterized by comprising the following steps of: according to the requirement analysis, a simplified module and a component connection mode of an automobile single-mass system are determined, and a basic simulation model is built based on Modelica; Performing initial simulation on the basic simulation model to obtain a simulation output result under initial parameters; Collecting actual measurement data of a real vehicle, preprocessing, constructing a comprehensive error function to evaluate the difference between the simulation output result and the actual measurement data, and carrying out iterative optimization on key physical parameters of the basic simulation model by adopting a parameter optimization engine until convergence conditions are met to obtain optimized parameters; and feeding the optimized parameters back to a basic simulation model to obtain a calibrated high-precision automobile single-mass system model.
2. The model-based vehicle single-mass system modeling and adaptive parameter optimization method according to claim 1, wherein the simplified module and component connection mode of the vehicle single-mass system are determined according to demand analysis, and a basic simulation model is built based on the model, and specifically comprises the following steps: simplifying a vehicle system according to a vehicle smoothness analysis target and a frequency decoupling principle, wherein the core simplifying module at least comprises a sprung mass module representing a vehicle body, a spring damping component representing a suspension system and a tire model representing tire rigidity; Analyzing the physical relationship of each element in the simplified module, determining that an input signal of the system is vertical displacement of a road surface, and an output signal at least comprises vertical acceleration of a vehicle body, and determining that the topological connection mode of each component is that a spring is connected with a damper in parallel, one end of the spring is connected with a sprung mass, the other end of the spring is connected with a tire model, and the tire model is connected with a road surface excitation source; Selecting a corresponding component model based on a Modelica standard library, wherein the corresponding component model at least comprises a translational Mass, a linear Spring, a linear Damper and a displacement source Position; when the standard library does not contain components with required characteristics, a component model is established in a custom modeling mode, constants are defined in a parameter definition area, and a nonlinear physical equation is input in an equation definition area; And performing physical interface connection on each component model subjected to parameterization setting through a connect statement of Modelica according to the topological connection mode to generate a complete automobile single-quality system basic simulation model.
3. The modeling and adaptive parameter optimization method of vehicle single quality system based on Modelica according to claim 2, wherein when the standard library does not contain the components with required characteristics, the component model is built by a custom modeling mode, constants are defined in a parameter definition area, and nonlinear physical equations are input in an equation definition area, which specifically comprises: Determining the physical characteristics of a required custom assembly, and analyzing the nonlinear behavior of the custom assembly, wherein the nonlinear behavior at least comprises a piecewise linear relation between nonlinear spring force and relative displacement, and a nonlinear relation between nonlinear damping force and relative speed; Creating a new component class in a Modelica modeling environment, and inheriting a part of connection interface classes in a standard library; the method comprises the steps of declaring custom parameters in a parameter definition area of the new component class, wherein the parameters at least comprise stiffness coefficients, damping coefficients of piecewise linear intervals or input and output data pairs required by table lookup; In an equation definition area of the new component class, writing an equation describing the physical behavior of the component by using Modelica language, wherein the equation at least comprises a conditional equation for calculating force output according to relative displacement or relative speed segmentation, or calling a predefined data table through an interpolation function to realize nonlinear mapping; and storing the custom assembly model to a local model library for calling in the construction of a subsequent model.
4. The model-based vehicle single-quality system modeling and adaptive parameter optimization method according to claim 3, wherein the initial simulation is performed on the basic simulation model, and a simulation output result under initial parameters is obtained, and the method specifically comprises the following steps: setting simulation time parameters, including simulation start time and simulation end time, and selecting a numerical integration algorithm and a step control mode according to analysis requirements, wherein a solver selects DASSL suitable for a rigid system or CVODE suitable for high-precision requirements, and step control adopts fixed step or self-adaptive step; Configuring initial parameters of a basic simulation model, wherein the initial parameters at least comprise sprung mass parameters, suspension stiffness coefficients, damping coefficients and road surface excitation signal types and amplitudes, and the road surface excitation adopts sine waves to simulate periodic jolts or adopts band-limited white noise to simulate random road surfaces; running a simulation program, solving and calculating the basic simulation model, and obtaining dynamic response time domain data of the system under initial parameters; And extracting a simulation output result from the dynamic response time domain data, wherein the simulation output result at least comprises vehicle body vertical acceleration and suspension dynamic travel, and is respectively used for subsequent smoothness evaluation and suspension travel check.
5. The model-based vehicle single-quality system modeling and adaptive parameter optimization method according to claim 4, wherein the steps of collecting actual measurement data of a vehicle and preprocessing the actual measurement data, constructing a comprehensive error function to evaluate the difference between the simulation output result and the actual measurement data, and performing iterative optimization on key physical parameters of the basic simulation model by using a parameter optimization engine until convergence conditions are met, so as to obtain optimized parameters, and specifically comprise: Carrying out real lane testing under corresponding working conditions, collecting a vehicle body vertical acceleration signal as actual measurement data, and preprocessing the actual measurement data to obtain a standardized actual measurement data set; Constructing a comprehensive error function for quantitatively evaluating the difference between the simulation output result and the standardized measured data set; And carrying out iterative optimization on the key physical parameters by adopting a two-stage hybrid optimization strategy, judging whether the comprehensive error function value after iterative optimization meets a preset convergence condition, if so, terminating optimization and outputting optimized parameters, and if not, continuing iteration until convergence.
6. The model-based vehicle single-quality system modeling and adaptive parameter optimization method according to claim 5, wherein the real lane testing is performed under the corresponding working conditions, the vehicle vertical acceleration signal is collected as actual measurement data, and the actual measurement data is preprocessed to obtain a standardized actual measurement data set, and the method specifically comprises the following steps: Selecting an actual road or a test site consistent with the simulation working condition, arranging an acceleration sensor at the mass center position of the vehicle body, performing real lane testing, and collecting a vehicle body vertical acceleration time domain signal as original actual measurement data; Detecting and eliminating abnormal values of the original measured data, and identifying and removing abnormal data points caused by sensor interference or road surface burst impact by adopting a statistical criterion; performing low-pass filtering treatment on the measured data with the abnormal values removed, removing high-frequency noise interference by adopting a fourth-order Butterworth low-pass filter, and retaining effective frequency components related to vertical vibration of a vehicle body; and resampling the filtered measured data, and reducing the original sampling frequency to the sampling frequency consistent with the simulation output to obtain a standardized measured data set for subsequent comparison and evaluation with the simulation output result.
7. The model-based vehicle single-quality system modeling and adaptive parameter optimization method according to claim 6, wherein the constructing the integrated error function is used for quantitatively evaluating the difference between the simulation output result and the standardized measured data set, and specifically comprises: Determining an error component type of the integrated error function, the error component including at least a root mean square error RMSE, a frequency error FreqError, a phase error PhaseError, and a peak error PeakError; Respectively calculating the numerical value of each error component, and distributing a weight coefficient for each error component, wherein the weight coefficient is determined according to the key requirement of vehicle performance analysis; Construction of the Complex error function RMSE FreqError PhaseError PeakError, wherein To the point of For the corresponding weight coefficient, output the comprehensive error function value As an objective function value in the subsequent parameter optimization iteration process.
8. The model-based vehicle single-quality system modeling and adaptive parameter optimization method according to claim 7, wherein the iterative optimization of the key physical parameters by adopting a two-stage hybrid optimization strategy specifically comprises: Initializing a two-stage hybrid optimization strategy, and setting a feasible region range of key physical parameters, wherein the key physical parameters at least comprise a suspension stiffness coefficient and a damping coefficient, and the feasible region range is determined according to a vehicle type and a suspension design experience value; Performing a first-stage global search, performing global exploration in the feasible region by adopting an improved genetic algorithm, and performing iterative evolution with the minimum comprehensive error function value as a target to obtain a global optimal parameter initial value; Performing local fine tuning in the second stage, and performing local fine searching by adopting a self-adaptive gradient descent algorithm by taking the initial value of the optimized parameter as a starting point, wherein the self-adaptive gradient descent algorithm dynamically adjusts the learning rate step length according to the error change rate in the iterative process; And in the local fine tuning process, re-calculating the comprehensive error function value after each iteration, judging whether the variation of the comprehensive error function value of two adjacent iterations is smaller than a preset threshold, judging convergence if the variation is smaller than the preset threshold, and outputting the final optimized parameters meeting the convergence condition.
9. The model-based automobile single-mass system modeling and self-adaptive parameter optimization method based on Modelica according to claim 8, wherein the optimized parameters are fed back to a basic simulation model to obtain a calibrated high-precision automobile single-mass system model, and specifically comprising the following steps: Acquiring optimized key physical parameters, wherein the key physical parameters at least comprise a suspension stiffness coefficient and a damping coefficient; Writing the optimized key physical parameters into a parameter configuration file of a basic simulation model, replacing original initial parameters in the model, and completing updating and calibration of the model parameters; Running simulation again based on the updated parameters, obtaining dynamic response data of the calibrated high-precision automobile single-mass system model under the same working condition, and verifying the optimization effect; And predicting and analyzing the vehicle performance based on the calibrated high-precision automobile single mass system model.
10. The model-based vehicle single-mass system modeling and adaptive parameter optimization method according to claim 9, wherein the prediction and analysis of vehicle performance are performed based on the calibrated high-precision vehicle single-mass system model, and specifically comprises the following steps: carrying out frequency domain analysis on the vehicle body vertical acceleration time domain signal in the dynamic response data, converting the vehicle body vertical acceleration time domain signal into a frequency domain signal by adopting fast Fourier transform, extracting an amplitude-frequency characteristic curve, and identifying the main resonance frequency of the system; Comparing and verifying the identified main resonance frequency with a theoretical calculated value or an actual measurement frequency spectrum, and judging that the frequency domain characteristic of the model meets the requirement when the resonance frequency error is controlled within 2%; performing time domain analysis on the time domain signal of the vertical acceleration of the vehicle body in the dynamic response data, performing frequency weighting processing on the acceleration signal, and calculating a weighted acceleration root mean square value as a quantitative evaluation index of the smoothness of the vehicle; And outputting a vehicle performance analysis report according to the results of the frequency domain analysis and the time domain analysis, and guiding the suspension system parameter optimization design.

Description

Modelica-based automobile single-quality system modeling and self-adaptive parameter optimization method Technical Field The invention relates to the technical field of automobile dynamics simulation, in particular to a model-based automobile single-mass system modeling and self-adaptive parameter optimization method. Background The automobile is actually a multi-degree-of-freedom complex vibration system consisting of an automobile body, a suspension, wheels, a road surface and the like. In the stage of engineering preliminary design or theoretical research, in order to rapidly focus on the basic characteristics of vehicle vertical vibration, a simple mass system model is often adopted for simplified analysis. The model ignores secondary factors such as wheel mass, a suspension specific guide mechanism and the like, and regards a vehicle body as a concentrated mass block which is connected with a road surface or a wheel through an equivalent spring and a damper. The simplification can help researchers to quickly master core problems such as vibration frequency, damping effect, driving smoothness and the like, refine the influence rule of key parameters and provide a theoretical basis for the follow-up construction of complex multi-body dynamics models. At present, modeling and analyzing methods for a single mass system of an automobile mainly comprise the following steps: 1. A simplified modeling method based on multi-body dynamics software is characterized in that a single-mass or 1/4 vehicle model is generally built in commercial software such as ADAMS/CAR, and sprung mass, suspension stiffness and damping coefficient are defined in a parameterized mode. Although the method is simpler and more convenient than the method for establishing the whole vehicle multi-body model, the model accuracy is highly dependent on accurate input parameters such as accurate centroid position, moment of inertia and the like. In practical engineering applications, if the host plant does not provide such data, the engineer will often make an estimate by a simplified geometry of the CAD model, which introduces an estimation error of 10% to 15%. In addition, although the model is simplified, a trade-off between complexity and computational efficiency is required, and there is a certain requirement for computational resources. 2. The modeling method based on the mathematical equation directly derives and establishes a differential equation describing the motion of the system according to classical mechanics theory such as Newton's second law, dalambert theory and the like. The advantage is that the physical concept is clear. However, in order to make the equations solvable, this approach will generally make a more simplified assumption for the actual situation, for example, neglecting nonlinear characteristics (e.g., dry friction, nonlinear damping forces) in the suspension system, resulting in an inherent deviation of the model from the actual system. Meanwhile, parameter uncertainty is also a key factor influencing the accuracy of the method, and key parameters such as suspension spring stiffness, damper damping coefficient and the like often need to be estimated, and the accuracy directly influences the reliability of the output result of the model. For complex road surface mechanisms or variable working conditions, a purely mathematical model often has difficulty in comprehensively and accurately describing the dynamic behavior of the system. 3. The modeling method based on the physical experiment is characterized in that the method directly obtains the input and output data of the system by building a physical bench (such as a vibrating table) or performing a physical road test, so as to identify the characteristics of the system or build an experience model. The method has the defects of high experiment cost, purchase and maintenance of equipment such as a vibrating table, a high-precision sensor and the like, and a large number of renaturation experiments for covering various working conditions. Temperature, humidity variations and noise disturbances in the experimental environment can also affect the accuracy of the measured data. In addition, the model built through a specific experiment is poor in universality, and when system parameters or external environments change (such as different vehicle types and different loads), the original model often fails, and experimental modeling needs to be performed again. In summary, the existing modeling method for the automobile single-quality system generally has the following technical bottlenecks: Geometric simplification and parameter sensitivity, namely the prior method or excessively simplifying the detail of a part, or relying on accurate mass center position, moment of inertia and other data, wherein the data are often estimated in actual engineering, so that errors are accumulated. The model calibration efficiency is low, the determination of the model parameters often de