Learning data-driven reduced inelastic models of spot-welded patches
Please login to view abstract download link
Components joined by spot-welds, a technique widely employed in car manufacturing, demonstrate complex and rich space- and time-localized behaviors. Spot-weld rupture has an important effect on the crash response of lightweight automotive bodies and must be considered in any realistic vehicle model. Simplified models commonly used for addressing the spot-weld mechanical behavior consist mainly of rigid or flexible beams enriched with a few simplified structural elements within the patch. However, these models do not capture the important 3D effects (rupture, damage, …) exhibited during crash simulation and their associated effects on the whole car structure. This motivates the need for considering the detailed 3D finite elements of the spot-welds into car models. Including a single local detailed (3D) spot-weld in the global model of the vehicle reduce drastically the timestep to satisfy the CFL stability criterion of the explicit time-integration scheme, which scales with the size of the smallest element. This in turn leads to a significant increase in computational effort. Furthermore, a typical vehicle will include around 4000 spot welds. Thus, including a 3D detailed model for all the spot-welds in the vehicle make the crash simulation computationally intractable. To overcome these difficulties, in the present work, we propose a data-driven technique for learning the rich behavior of a local 3D patch and integrate it into a standard global description at the structure level. Thus, the localized effects express in the global response, but their explicit online solution is no longer required.
complas_sec@cimne.upc.edu/ Telf. + 34 - 93 405 46 94
Copyright © 2022 CIMNE, All Rights Reserved. Terms of service
