Robust and Automated Geometry Processing Workflows for Engineering Applications
Vijai Kumar Suriyababu (TU Delft - Electrical Engineering, Mathematics and Computer Science)
M. Möller – Promotor (TU Delft - Electrical Engineering, Mathematics and Computer Science)
C. Vuik – Promotor (TU Delft - Electrical Engineering, Mathematics and Computer Science)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Modern computer-aided engineering (CAE) depends on discrete 3D data such as meshes and point clouds, yet raw geometric models are rarely ready for direct use. Industrial CAD assemblies and scans often contain holes, topological ambiguities, irregular sampling, and feature loss that can disrupt downstream simulation and design workflows. This thesis addresses these practical gaps by developing multiple geometry-processing workflows that prioritize robustness, automation, and reproducibility.
One contribution focuses on topological queries and cleanup. A curvature-aware, low-parameter approach detects hole boundaries and supports through-hole removal when needed. The same chapter also presents extraction of negative- volume regions as a separate topology-processing task. Independently, the thesis introduces Series of Local Triangulations (SOLT), an intermediate representation for stable resampling of non-uniform point clouds into uniform, high-fidelity sets while preserving geometric detail, and explores a preliminary application to CADfree higher-order surface reconstruction.
For simulation-ready models, the thesis proposes a shrink-wrap mesh generation workflow on octrees that combines distance fields and adaptive morphology. This framework enables reliable topology simplification, controlled genus reduction, and robust wrap mesh generation for complex real-world geometries. In a separate application-focused workflow, the thesis also presents a heat-method-based algorithm for mean camber line extraction and structured multiblock decomposition, offering a practical and implementation-friendly alternative to traditional geometric constructions.
These algorithmic contributions culminate in tgLang, a strongly typed domain-specific programming language for scientific computing and geometry processing. By elevating geometric entities to first-class language constructs and coupling them with deterministic execution, runtime support, and native deployment, tgLang transforms complex mesh workflows into concise and reproducible programs. Together, the methods and language presented in this thesis provide a flexible toolbox that guides imperfect input geometry toward reliable, application-ready computational models.