Research On Crashworthiness Optimization Of Thin-walled Model For Big Simulation Data

Thin-walled structure is a commonly used engineering structure form,which is widely used in automobile,shipbuilding,aerospace and other fields.With the increasing demand for personalized and diversified products,how to improve the efficiency of optimized design has become a hot topic in the field of simulation analysis.Although computer hardware and software are constantly updated and upgraded,simulation calculation and shape optimization still consume a lot of time.How to quickly and accurately obtain the optimization results of complex structures has become an urgent problem.In this paper,on the premise of ensuring the quality of the results,a set of shape optimization process for simulation big data is proposed,and the traditional solver calculation process is replaced by neural network.At the same time,stamping manufacturability is used as a geometric constraint to ensure the shape optimization effect of complex thin-walled structures.In order to improve efficiency,using C ++ programs to realize batch production of simulation data,data processing and annotation.The research contents are as follows:(1)In view of the fact that the B-pillar directly affects the driver's personal safety in the side collision accident of the vehicle,this paper optimizes the shape of the B-pillar thinwalled structure to improve its crashworthiness.Considering that most auto parts are made by stamping process,and the deformation of complex thin-walled models is prone to negative angle structure,this paper proposes a simple and efficient negative angle detection-modification algorithm,and uses professional stamping simulation software Auto Form to verify,This proves that the algorithm has high robustness.Optimal Latin Hypercube Sampling is used to sample design parameters from a specific design domain,and the RBF deformation method is used to generate a variant mesh model.Hypermesh and Ls-Dyna are used to perform collision simulation experiments on a small number of models to establish Training data set.(2)Using the most widely used BP neural network to extract the nonlinear mapping relationship between the deformation parameters of the model and the collision performance,and then obtain the shape solution set that makes the model have the best collision resistance.Integrate the above algorithm into Auto Morpher / TSi MGen assembly,which is convenient for subsequent use.To explore the application of big data in the field of simulation,use the Auto Morpher / TSi MGen assembly of this article to build a part of the 3D mesh model database,and establish the correspondence between the geometric information of the model and the mechanical properties under collision conditions,prepare a good channel to obtain high quality data.(3)It has been verified that the algorithm and program proposed in this paper can achieve the expected goal well,and the crash resistance of the B-pillar model can be improved by more than 15%.Compared with the traditional shape optimization process,the experimental process can save a lot of repetitive labor,and because the algorithm is highly integrated,the user can complete the optimization of the model without having a mechanical or mathematical foundation.