MS06 - Data-driven dynamics

Keywords: Data-driven analysis and modeling of dynamics; Physics-constrained learning; Learning-based system identification; Data-enhanced physical simulations; Industrial applications of deep learning for engineering dynamics; Data Science; Data-based control; Error bounds and guarantees in data-based methods

Organizers:
Charlotte Geier (1) – charlotte.geier@tuhh.de
Malte Grube (1) – malte.grube@tuhh.de
Annika Hayn (2,3) – annika.hayn@de.bosch.com
Lukas Lanza (4) – lukas.lanza@tu-ilmenau.de

Affiliations:
(1) Hamburg University of Technology, Germany
(2) Robert Bosch GmbH, Germany
(3) University of Kassel, Germany
(4) Technische Universität Ilmenau, Germany

Abstract:
Data-driven approaches are transforming the analysis, modeling, and prediction of dynamical systems. By integrating mathematical and statistical principles in machine learning approaches, significant progress is being made in addressing key challenges in engineering. Particularly phenomena that limit the effectiveness of traditional physics-based simulations, such as nonlinearities, transient loads and model uncertainties, are being addressed with promise. A focal point in this area is the development of hybrid methods that seamlessly combine physical insights into system dynamics with data-driven strategies.

In computational mechanics, data-based methods complement classical approaches in applications such as predictive maintenance, real-time system optimization, and advanced control techniques. Contributions will provide insights into the effective integration of data-driven methods with traditional engineering practices to enhance performance and optimize outcomes. Case studies demonstrating successful implementations of hybrid models in various branches of engineering will be highlighted, showcasing tangible benefits and practical insights. We also welcome contributions that address the theoretical foundations of data-driven approaches, with a focus on error bounds and guarantees based on the data collected and the models derived from it.

This symposium aims to bring together researchers from the interdisciplinary fields of machine learning and data-driven methods with mechanical dynamics to foster collaboration and promote the adoption of hybrid algorithms in engineering applications. We anticipate presentations from academic institutions and industry, reflecting the growing prominence of data-driven approaches in science-based decision-making. Additionally, the symposium will serve as a platform for discussing the latest advancements and future directions in this rapidly evolving field. Topics of interest to this symposium include, but are not limited to the ones mentioned a keywords, with an emphasis on innovative approaches that push the boundaries of current engineering practices.