【作者】 曹伟娟；

【导师】 高曙明；

【作者基本信息】 浙江大学 ， 计算机科学与技术， 2013， 博士

【摘要】 在对产品进行CAE分析时,首先需要基于设计结果进行分析几何建模。然而由于分析几何建模的复杂性,目前分析几何建模过程仍离不开工程师的经验和大量的手工交互,阻碍了设计分析过程的快速迭代。为了提高分析几何建模的智能化和自动化程度,本文围绕分析特征建模的若干关键问题展开研究,主要研究工作包括：1、基于特征的分析几何表示提出分析特征模型,利用分析特征表示分析几何中涉及分析语义的局部几何形状,并建立设计几何模型与分析几何模型底层几何元素的关联,以有效支持分析几何模型的自动创建以及设计几何与分析几何的双向互动。2、基于边界表示的分析特征模型生成针对边界表示模型,提出自动生成分析特征模型的方法。该方法通过面对匹配识别可降维区域,充分利用凹边信息提高面对匹配效率、减少分割次数,并基于凹边的连接信息对中面进行合理延拓和裁剪。3、基于特征转换的分析特征模型生成针对特征表示模型,提出基于特征转换的分析特征模型生成方法。利用扫成体将三维的可降维区域识别和简化几何提取问题转化为二维问题以提高效率。此外,基于正特征剩余体分解扫成体,扩大可降维区域识别的范围。最后对可降维区域的误判和漏判情况进行处理,以保证特征转换结果的正确性。4、旋转对称特征的对称单元及其网格的自动生成针对旋转对称模型,提出一种构造其对称单元特征的方法。该方法利用一种改进的Delaunay细分算法,仅对局部网格进行细分,通过在细分过程中加入移动三角形或四面体操作,允许对称边界的变动,以便维护对称边界的一致性,同时保证对称边界上边和面的Delaunay属性。本文证明了二维情况下该方法生成的对称单元网格单元数目是最小数目的常数倍,即对称单元是最优的。更多还原

【Abstract】 During the process of design and analysis, the analysis geometric model is usually generated from the design geometric model by simplification methods. However, due to the absence of an analysis geometric representation capable of supporting the seamless integration of CAD and CAE, and methods for the effective generation of the analysis geometric model, manual work and engineering experiences are still needed in the analysis geometric modeling process. This situation severely hinders the fast iteration of the design-analysis cycle. To solve these problems, a feature based analysis geometric representation is proposed in this paper, and their modeling methods are studied. Main contributions of this paper include:1. Feature based analysis geometric representation.The analysis feature model is proposed. Analysis features are defined to capture analysis semantics involved in the analysis modeling process. Low level associations between the design geometric model and the analysis geometric model are also included in the analysis feature model. With such information, the analysis feature model can enable the automated generation of the analysis model and bi-directional interaction between the design model and the analysis model.2. Generation of analysis feature models from boundary representation models.An approach to automated generation of analysis feature models from boundary representation models is proposed. Thin regions are identified by paring faces. Information about concave edges is utilized to improve the effectiveness of face paring, to reduce the number of times of volume decomposition, and to guide the extending and trimming of mid-surfaces.3. Generation of analysis feature models by feature conversion.An approach to automated generation of analysis feature models from design feature models is proposed. The problems of thin region recognition and idealized geometry extraction in3D are reduced to corresponding2D problems by taking advantage of swept volumes. The range of thin region recognition is enlarged by decomposing remnants of additive features into swept volumes. Missed and wrongly recognized thin regions are also processed, assuring correctness of the results.4. Construction of optimal symmetry cells for symmetry features.An approach is proposed to constructing optimal symmetry cells for rotationally symmetric models. A symmetry cell mesh is constructed using a modified Delaunay refinement mesh generation algorithm. The algorithm operates just on a local mesh. By introducing move of triangles and tetrahedrons, the symmetry boundaries of the local mesh are allowed to change, so as to maintain the consistency of the symmetry boundaries, and to guarantee the Delaunayness of the symmetry edges and faces. The symmetry cell mesh in2D case is proved to be optimal in the sense that its number of elements is within a constant factor of the smallest number.更多还原