Dr. Jan Schubnell, born in 1987, received his BSc in Mechanical Engineering from Furtwangen University in 2012 and his MSc from Offenburg University of Applied Science, Germany in 2016. Since 2016, he has worked in the group “Fatigue” of the business unit “Component safety and Lightweight Construction” at the Fraunhofer Institute for Mechanics of Materials IWM in Freiburg. He received his PhD from the Faculty of Mechanical Engineering of Karlsruhe Institute of Technology (KIT), Germany in 2021. He continued his work in the field of fatigue of welded joints and residual stresses at the Fraunhofer Institute for Mechanics of Materials IWM in Freiburg.
Abstract
The surface geometry and the stress concentration of welded joints shows a large variation and is individual for each joint at each position. This is one reason for the conservative fatigue assessment of welded joints. In the past the determination of stress concentration factors (SCF) by Finite Element (FE) simulation was based on the idealized surface geometry of the welded, defined by weld toe radius and flank angle, or other geometrical parameters. In this work a new approach is presented to directly determine SCFs of welded joints based on the real shape of the surface from 3D-surface scans. For this, different types of artificial neuronal networks (ANN) are used. As input parameter for the ANN single 2D-cuts evaluated from 3D-scans were used. The ANN was trained SCFs determined from FE-simulation based on the real weld geometry and idealized weld geometry (approximated by weld toe radius and flank angle). The developed SCF solution was compared to commonly used SCF solution based on idealized weld geometry. The error of the proposed solution based on the real weld geometry was quite smaller compared to solutions based on the idealized geometry.Session
| Room | Date | Hour | Subject |
|---|---|---|---|
| Room 6 | Wednesday 29th November | 15:00-15:30 | Jan Schubnell S02-2 Big Data and Artificial Intelligence 89 - Determination of stress concentration factors of welded joints from 3D-surface scans by artificial neural networks |
