PhD defence by Tim Brix Nerenst

Variation is the leading cause of quality loss for unrobust designs and provokes expensive redesigns in late development stages. Even worse, poor robustness can lead to mechanical failures or malfunction with fatal outcomes, such as the case with Toyota’s sticking accelerator pedal in 2010. When worn and in rare conditions, unfortunate variation in the geometry led to the pedal sticking, accelerating the car without the driver’s control. While physical experiments are beneficial for detecting major structural issues such as fractures, it requires an enormous amount of samples to represent the naturally occurring variation from full-scale production and use. Instead, virtual experiments have the ability to control the deterministic and stochastic variables, which can result in a more detailed and cost-effective process for evaluating design robustness compared to physical experiments.

The present research investigates how existing engineering disciplines Computer-Aided Design, Finite Element Analysis, Design of Experiments, Sensitivity Analysis, and Optimization can be combined to create a process for virtually evaluating the structural robustness of mechanical designs, known as “FEA-based variation simulation.” While each engineering discipline thrives individually, their interdisciplinary use is limited in the industry. The present research aims to understand the existing barriers through academic and industrial insight to develop solutions that enable engineers to evaluate the structural robustness cost-effectively in early design phases.
The key results of the present research are; (i) a mapping of technical, practical, and knowledge barriers experienced in the industry when attempting to perform virtual robustness evaluation ([P1], [P2]). (ii) A new sequential framework (sR2BDO) suitable for screening and optimization of design robustness and reliability, with improved computational efficiency ([P3]). (iii) Development of the Robust Sketch Principles to enhance the existing CAD methodology ([P4]). (iv) Explicit guidance for applying FEA-based variation simulation through case studies. With a focus on; robustly configuring the FE-models to avoid component interference, a robust selection of node/element groups for automated result extraction ([P2, P3]), including material variation and detection of kinematic issues (Extended results of [P3]).

Combined, this thesis and the appended articles aim to provide enough information to enable new users to apply the interdisciplinary process of FEA-based variation simulation, making it more accessible for engineers across all sectors to evaluate structural robustness.

Main supervisor:
Associate Professor Kim Lau Nielsen, Dept. of Civil and Mechanical Engineering, DTU

Co-supervisors:
Associate Professor Tobias Eifler, Dept. of Civil and Mechanical Engineering, DTU
Senior Innovation Lead Martin Ebro, Novo Nordisk A/S
Simulation Specialist Morten Hartvigsen Nielsen, Novo Nordisk A/S

Examiners:
Associate Professor Brian Nyvang Legarth, Dept. of Civil and Mechanical Engineering, DTU
Professor Jean-Yves Dantan, École Nationale Supérieure d’Arts et Métier (ENSAM), France
Professor Rikard Söderberg, Chalmers University of Technology, Sweden

Chairman:
Associate Professor Thomas J. Howard, DTU Centre for Technology Entrepreneurship



Tidspunkt

ons 25 maj 22
13:00 - 17:00

Arrangør

DTU Mekanik

Kontaktperson

Hvor

Building 421, Auditorium 074
Technical University of Denmark
https://dtudk.zoom.us/j/63060234299#success