Boundary Representation Language Modeling Repairs CAD Geometry Through Geometry Tokens
Global Media and Social Sciences Research Journal · 20.06.2026 · 7 (2) · 150-160
DOI: 10.71465/gmssrj206 ↗Kurz gefasst
Fachartikel in der Zeitschrift Global Media and Social Sciences Research Journal, online erschienen am 20.06.2026. Verfasst von Laura Ferrari und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The seamless translation of computer-aided design models across diverse software environments remains a persistent challenge in modern engineering workflows, primarily due to topological inconsistencies and geometric translation errors. Boundary representation structures, the industry standard for modeling three-dimensional solid objects, often suffer from missing faces, disconnected edges, and overlapping surfaces during file conversion or parametric manipulation. Traditional heuristic algorithms and localized geometric healing tools frequently fail to resolve complex topological corruption, necessitating labor-intensive manual intervention. This paper proposes a novel framework that reformulates boundary representation data as a structured sequence of topological and geometric tokens, enabling the application of advanced language modeling architectures to automate parametric computer-aided design repair operations. By projecting vertices, edges, co-edges, and faces into a high-dimensional continuous embedding space, we leverage transformer-based self-attention mechanisms to predict and reconstruct missing or corrupted topological entities. The methodology involves the development of a custom tokenization strategy that captures both the hierarchical dependency of boundary representation structures and the underlying mathematical surface definitions. Through extensive experimentation on a synthetically corrupted dataset of industrial components, the proposed language model demonstrates a profound ability to infer design intent and restore manifold solidity. The approach significantly outperforms conventional graph-based and rule-based repair algorithms in terms of topological accuracy and geometric deviation. These findings establish a foundational bridge between natural language processing techniques and geometric deep learning, offering scalable solutions for automated design validation, interoperability, and intelligent geometric processing in manufacturing ecosystems.
Zitieren
APA
Ferrari, L., & Yamamoto, S. (2026). Boundary Representation Language Modeling Repairs CAD Geometry Through Geometry Tokens. Global Media and Social Sciences Research Journal, 7(2), 150-160. https://doi.org/10.71465/gmssrj206
Vancouver
Ferrari L, Yamamoto S. Boundary Representation Language Modeling Repairs CAD Geometry Through Geometry Tokens. Global Media and Social Sciences Research Journal. 2026;7(2):150-160. doi:10.71465/gmssrj206
BibTeX
@article{Ferrari2026,
author = {Ferrari, Laura and Yamamoto, Sayaka},
title = {Boundary Representation Language Modeling Repairs CAD Geometry Through Geometry Tokens},
journal = {Global Media and Social Sciences Research Journal},
year = {2026},
volume = {7},
number = {2},
pages = {150--160},
doi = {10.71465/gmssrj206}
}
EUVIMED