Research On Multidisciplinary Optimization Design Of Frame Body Front Impact Structure Based On Machine Learning

Vehicle lightweight is one of the key technologies to achieve the requirement of carbon emission reduction in the Energy-saving and New Energy Vehicle Technology Roadmap 2.0.The body lightweight can not only aim at the goal of the lowest mass,but also improve the automotive safety.The research on improving active and passive safety while reducing the weight of the body is very complex.Therefore,the frame body structure and materials is researched to lightweight design using Multidisciplinary Design Optimization(MDO)method.To solve the shortcomings of the approximate modeling method and optimization strategy in the MDO framework,the basic principles of machine learning and dynamic system consistency constraints methods are explored.The approximate modeling technique and optimization strategy are improved respectively by Gaussian Process Regression(GPR)approximation model and dynamic relaxation factor constraint.MDO framework for the front crash structure of frame body is built based on machine learning.The main research contents of this paper are as follows:(1)Building a multidisciplinary design optimization framework for front impact structure of frame body based on machine learning.To solve the complex problem of improving active and passive safety while lightweight the body,the MDO model is established and the basic principles are analyzed.The influence of approximate modeling techniques on accuracy are explored,and the influence of multidisciplinary optimization strategies on consistency constraints in the MDO of the process are analyzed.To build a multidisciplinary optimization design framework for front impact structure of frame body based on machine learning,the mass model of front impact structure is seen as the system level to achieve the lightweight;the crash safety model is seen as the subsystem level to reduce the peak acceleration of front impact structure and the intrusion of front baffle;the aerodynamics model is seen as the subsystem level to increase the sufficient down force to improve handling stability.This Framework could improve both active and passive safety,while satisfying lightweight design requirements.(2)To build a GPR approximation model based on particle swarm optimization algorithm.The GPR approximation model based on particle swarm algorithm is established by analyzing the principle of Gaussian Process(one of the machine learning algorithms),and the characteristics of Gaussian Process kernel function are explored.The GPR approximation model is tested by standard mathematical examples and the high accuracy of this novel model is verified to meet the design requirements.(3)Analytical Target Cascading(ATC)method based on dynamic relaxation factor constraint is proposed.The basic principle of ATC and the constraint mechanism of relaxation factor are researched to analyze the influences of dynamic relaxation factor on the consistency of MDO.The differences in convergence accuracy and iteration efficiency between dynamic relaxation factor constrained ATC and traditional ATC are analyzed using mathematical examples.The comparison results show that the dynamic relaxation factor constrained ATC iterates faster and has higher convergence accuracy.(4)Multidisciplinary lightweight design based on Gaussian process regression with particle swarm optimization algorithm and dynamic relaxation factor constrained ATC is carried out.The front impact structure of frame body is optimized by using Multidisciplinary Design Optimization method.Firstly,the approximation model between the design variables and the performance response are established by GPR.Secondly,the mathematical model of this lightweight problem is established by dynamic relaxation factor constrained ATC.Finally,the optimization is carried out by particle swarm algorithm.The optimization results show that compared with the traditional ATC method,the dynamic relaxation factor constrained ATC method reduce the body mass by 3.87%,peak acceleration by 12.97%,intrusion at front baffle by4.11%,air lift coefficient by 16.28%,and the iteration efficiency of the proposed method is improved by 34.0%.This novel framework efficiently and accurately improve the active and passive safety of the whole vehicle in the process of lightweight that has theoretical value and engineering application significant.