Research On Model Simplification And Validation Method For Auto Side Impact Simulation

In the development of modern automobile industry,energy and safety have become the basic requirements for automobile design.How to achieve weight reduction on the premise of guaranteeing the crashworthiness has been a hot topic in automobile optimization area.Automobile is commonly seen as a complex engineering system and computer simulation technology has commonly employed for improving design efficiency and saving development cost in automobile structure optimization.With the growing development of computational capability,the element number of vehicle simulation model has become more and more.The simulation accuracy has also been greatly improved,while the simulation efficiency has been decreased.Hence,how to improve optimization efficiency without lose necessary accuracy becomes a problem to be settled urgently.During the conversion process from geometrical models to finite element models,the impacts of many detail removals on analysis results are very small.So,the analysis oriented model simplification plays a key role on reducing the analysis data and improving the computation efficiency.In order to ensure the reliability of the simulation model and the accuracy of the simulation result,model verification and validation is emerged for quantifying the model precision.In the optimization process of automobile structure,lots of iteration simulation of finite element analysis is still time consuming.By using surrogate models constructed based on the samples and responses,the process of obtaining the responses through finite element analysis can be skipped,and the optimization efficiency can also be improved.Therefore,the surrogate models are widely employed to replace the finite element models in automobile structure optimization.In this paper,the following three aspects are researched:1.An analysis oriented error evaluation system for model simplification is established based on the traditional finite element model simplification methods.The simplified features are extracted and parameterized firstly.Then,based on the feature parameters and analysis results of simplified models,machine learning algorithms are applied to train the error evaluation system.Engineers can utilize this system to evaluate the impacts of simplified features on analysis results before CAE simulation.This system can be used as a tool for constructing accurate and efficient simplified models in finite element simulation.2.In the study of vehicle side impact,the most intuitive manifestation of lateral crashworthiness is the deformation of the door surface.A deformed surfaces comparison based model validation metric is developed for studying the deformation status in vehicle side impact simulation.Combined with dynamic response model validation method,an integrated validation framework for vehicle side impact model is raised in this paper.The accuracy of CAE model can be validated more comprehensively through this proposed method.3.By approximating the high complex simulation process with meta-modelling technology,the time of simulation based optimization design can be reduced a lot.The surrogate models for B pillar structure optimization are constructed based on the simplified side impact simulation model.Considering the discrepancy existed between the simplified model and original model,the bias corrections based on the response errors of high-fidelity model and low fidelity model are added on the optimization constraints.This correction process can improve the accuracy of optimal design based on simplified model.Finally,the proposed methods are applied into a B-pillar lightweight design.The optimal structure is of nice crashworthiness and the weight of optimal B pillar is reduced.And the results shows that the optimization efficiency can be improved while design accuracy can be also guaranteed by the methods proposed in this paper.