Reduction of 3D-FE models to 1D beam models for AI-based simulations

Treschau M, Witzgall C, Wartzack S (2026)

Publication Language: English

Publication Type: Conference contribution, Conference Contribution

Publication year: 2026

Pages Range: 011-020

Conference Proceedings Title: DS 151: Proceedings of the 37th Symposium Design for X (DFX2026)

Event location: Radebeul

URI: https://www.designsociety.org/publication/49297/reduction_of_3d-fe_models_to_1d_beam_models_for_ai-based_simulations

DOI: 10.35199/dfx2026.2

Open Access Link: https://www.designsociety.org/publication/49297/reduction_of_3d-fe_models_to_1d_beam_models_for_ai-based_simulations

Abstract

Especially in early stages of product development, it is necessary to evaluate designs using fast and automated simulations. While conventional finite element analysis is computationally expensive, physics-informed neural networks enable real-time prediction. However, they require problem-tailored architectures and therefore physically consistent training data. To address this issue, we present a procedure for automatically reducing 3D-FE models to physically consistent, backward-compatible 1D beam models. This method builds on geometric skeletonisation by mapping mechanical state variables onto the medial axis to preserve geometry, topology and physical information. The resulting models can be used to generate synthetic training data for machine learning models.

Authors with CRIS profile

Moritz Treschau Chair of Engineering Design (KTmfk) Christian Witzgall Chair of Engineering Design (KTmfk) Sandro Wartzack Chair of Engineering Design (KTmfk)

How to cite

APA:

Treschau, M., Witzgall, C., & Wartzack, S. (2026). Reduction of 3D-FE models to 1D beam models for AI-based simulations. In Paetzold-Byhain, K.; Krause, D.; Wartzack, S. and Koch, A. (Eds.), DS 151: Proceedings of the 37th Symposium Design for X (DFX2026) (pp. 011-020). Radebeul, DE.

MLA:

Treschau, Moritz, Christian Witzgall, and Sandro Wartzack. "Reduction of 3D-FE models to 1D beam models for AI-based simulations." Proceedings of the DfX Symposium 2026, Radebeul Ed. Paetzold-Byhain, K.; Krause, D.; Wartzack, S. and Koch, A., 2026. 011-020.

BibTeX: Download