UBC Theses and Dissertations
Variational direct modeling for computer-aided design Qiang, Zou
This thesis presents a new computer-aided design (CAD) modeling approach for three-dimensional objects. Improving CAD modeling efficiency has always been a central topic in the CAD domain, especially for model editing. Currently, two CAD modeling paradigms, each with its own capabilities and limitations, dominate this subject. The parametric modeling paradigm offers great flexibility for global model edits involving preplanned parametrics but becomes very rigid for unplanned model edits. The very recent direct modeling paradigm provides flexible local model edits but barely supports parametric (global) edits. In order to improve modeling efficiency, flexible local and global model editing need to coexist. This work proposes a novel modeling approach for this purpose, which integrates variational modeling with direct modeling. It is for this reason that this approach is named variational direct modeling. The underlying problem for variational direct modeling is information inconsistency resolution. There are three layers of information in a model: geometry, topology, and constraint. When an information layer is edited, the changes are not reflected in the others automatically. As a result, the consistency of the three information layers in the pre-edit model is broken, and an invalid model is generated. There often exist many options for resolving such inconsistencies, and the fundamental challenge lies in ensuring the validity of resulting models, which requires systematic decision-making among the options. Unfortunately, there has not been much existing research work on such decision-making. The main contributions of this thesis include a thorough analysis of the information inconsistencies and novel, systematic decision-making methods to resolve them. The analysis primarily discusses what forms the inconsistencies take, based on which, effective methods are proposed to take out these inconsistencies and to rethink the relevant information. The presented methods yield a modeling result that (1) is guaranteed to be valid (being solid and well-constrained) and (2) attains a continuous model shape variation for direct edits and (3) exhibits a minimal model variation for parametric edits. These methods have been validated through a series of case studies.
Item Citations and Data
Attribution-NonCommercial-NoDerivatives 4.0 International