Knowledge Base Repository
In addition to research papers, the Design Society is developing several valuable resources for those interested in the study of design. These include a repository of PhD theses, a library of case studies and transcripts of design activities, and an archive of our newsletters. Please note that these resources are accessible exclusively to Design Society members.
Measuring the similarity of finite element models using a hybrid approach with pre-trained transformers
Roos, Aron; Goetz, Stefan; Wartzack, Sandro
Type:
Year:
2026
Editor:
Paetzold-Byhain, K.; Krause, D.; Wartzack, S. and Koch, A.
Author:
Series:
DfX
Institution:
Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany
Page(s):
140-149
DOI number:
Abstract:
AI-based approaches can accelerate product development loops, for example with plausibility checks for finite element (FE) simulations. Their reliability for new FE models depends on their similarity to the training data. Existing approaches to assess this similarity only compare the geometry and neglect FE-specific aspects. This paper addresses this problem in two steps. First, a literature review and benchmark evaluation are conducted, identifying a pre-trained transformer as a suitable approach to measure geometric similarity. Second, we propose a hybrid approach with two separate transformer-based branches for boundary conditions and mesh properties, which are combined in a similarity score. This modular concept enables the use of different networks for the FE-specific aspects.
Keywords:
