Resnet networks for plausibility detection in finite element simulations
DS 118: Proceedings of NordDesign 2022, Copenhagen, Denmark, 16th - 18th August 2022
Year: 2022
Editor: Mortensen, N.H.; Hansen, C.T. and Deininger, M.
Author: Bickel, Sebastian; Schleich, Benjamin; Wartzack, Sandro
Series: NordDESIGN
Institution: Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU)
Section: Product Validation
Page(s): 10
DOI number: https://doi.org/10.35199/NORDDESIGN2022.4
ISBN: 9781912254170
Abstract
Today, product design without Finite Element (FE) simulation is hardly conceivable. In addition, the market promotes shorter development times and a greater product variety. This combination poses the risk of inexperienced users performing simulation tasks. An idea to support these is to check their FE models for plausibility. One approach utilizes calculated simulations to train a Deep Learning model that classifies the new simulations. However, a high recognition accuracy is required. Therefore, this paper investigates the ability of a ResNets to check the plausibility of simulations.
Keywords: Data mining, structural analysis, data driven design, Deep Learning, Plausibility checks