|About this Abstract
||NUMISHEET 2022: The 12th International Conference on Numerical Simulation of 3D Sheet Metal Forming Processes
||Automotive Sheet Metal Panels Buckling Instability Prediction in the Design Stage Based on Convolutional Neural Network
||Seungro Lee, Luca Quagliato, Donghwi Park, Naksoo Kim
|On-Site Speaker (Planned)
Since metal sheets buckling instability varies according to design styling, robust prediction model for this phenomena ought to be developed. Thus, this study exploits the Convolutional Neural Networks (CNN) theory for an automotive sheet metal panels stiffness prediction model. The data for training was generated by using Finite Element Method (FEM) validated with indentation test and Stiffness Index (SI) was used for quantitative evaluation of the stiffness of the considered panel. The learning CNN model has been developed considering 19 convolutional and 8 pooling layers. The shape of panels and corresponding SI distribution were used as input data. To validate this prediction model, model predicted SI distribution of panel shape unused for training and result was compared with FEM and experiment data. This CNN based prediction model can be properly utilized for automotive sheet metal panels stiffness prediction already in the design stage.
||Definite: At-meeting proceedings