| 摘要: | 網格在有限元素分析中為重要的技術，是直接影響模擬分析的準確度和計算效率，而建構網格卻是最重要也最耗時的工程，尤其在射出成型中，因模型通常為薄件，邊界處的性質又變化劇烈，若使用常見的Tetrahadron建立網格，雖然自動快速但品質較差。若使用六面體網格或三角柱網格雖然適合於薄件中，但需手動建立且難以產生。因此本研究利用自動特徵辨識技術，將特徵擷取後，依不同特徵建立一套對應的網格化模式，將特徵處均建立六面體網格和三角柱網格的結構化網格，而模型剩餘的部份則使用BLM (Boundary layer meshes)填滿。本研究先進行下列常見且難以網格化的特徵進行開發辨識，包括肋、Tube和Boss，並提供每組特徵獨立的資訊，且模型的拓樸資訊、倒圓角等資訊也同時掌握。另外本研究整合多個特徵辨識於一系統，使多人可同時開發且不互相衝突，計算之資料又能互相使用的環境，包含前置處理、倒圓角、孔洞和倒圓角簡化等技術。另外本研究開發自動規劃肋的切割位置及建立切割面，將肋分解成數塊規則的區域和提供網格排列的方式，以避免網格相交的問題。最後本研究以手動的方式建立結構化網格，所有網格品質指標均有改善，其中Orthogonality改善最為明顯。綜合以上技術，整合特徵辨識與特徵分解，以輔助網格化功能自動建立排列規則且高品質的結構化網格。;In finite element analysis, meshing is an important step as it affects the accuracy and efficiency of the desired analysis and simulation. Mesh generation is usually a time-consuming process, especially in injection molding which involves a lot of thin-shell plastic parts where the shape on the boundary changes dramatically. Tetrahedral meshes are commonly used as they can easily be generated. But, their quality near thin-shell regions is often poor. By contrast, hexahedral and prismatic meshes are more suitable to describe thin-shell shape. But, they are difficult to be generated automatically. This study proposed an approach for recognizing features from a CAD model and decomposing them into regions that can be meshed with hexahedral and prismatic meshes; the rest of the CAD model was then meshed with boundary layer meshes (BLM). This study emphasizes the recognition of rib, tube, and boss features as they frequently exist on CAD models and are difficult to be meshed. All geometric and topological data related to each of these features were addressed and evaluated, including the information of fillets, if they exist. A software platform for the development of various feature recognition algorithms was established, which can provide preliminary data, fillets, holes, chamfers and simplification data for the use in different feature recognition algorithms. A rib decomposition algorithm was also developed to manage undercut regions, generate slicing faces, and decompose the ribs into segment and transition regions. The proposed algorithm provides a method of mesh arrangement to match the nodes between the features and the rest of the model. Finally, mesh generation in terms of the features recognized was conducted manually for the generation of hexahedral and prismatic meshes. The result indicated that all criteria of mesh quality have been improved significantly, with the orthogonality index improved the most. In conclusion, this study developed feature recognition and decomposition algorithms to extract several main features from a CAD model so that a mix of hexahedral and prismatic meshes can be applied for improving the mesh quality of the features recognized. |
