- 発表日 2020/07/08
- 17th Annual International CAD Conference（国際会議，オンライン）
- DOIコード 10.14733/cadconfP.2020.297-301
Shape Descriptor-Based Similar Feature Extraction for Finite Element Meshing
Large-scale finite element models (FE models) used in car crash simulations, etc. require high-quality FE meshes in compliance with meshing specifications, such as the location and resolution of the meshing pattern’s features such as ribs and bosses to ensure simulation accuracy. However, the automatic feature-compliant meshing is not fully supported in commercial CAE software, requiring many manual operations and resulting in a high person-hour ratio to the whole CAE process. Extracting feature shapes, such as ribs and bosses, for which the meshing specifications have been strictly assigned from the complex shape of the product CAD model is particularly time consuming. Furthermore, the feature geometries are not uniquely shaped and are usually bounded by smooth and indistinct boundaries, often leading to oversights. Therefore, an automated feature extraction technique targeting a feature-compliant finite element meshing is strongly required where the features whose geometries are not identical to a reference feature shape but are similar to it can be extracted from a product’s CAD model. Researchers have proposed methods to extract feature shapes from CAD models to generate meshes of FE models [2-3][5]. However, three main problems persist. First, the feature extraction does not work robustly when the CAD model includes PDQ issues such as cracked or degenerated geometries. Second, features surrounded by complex and smooth boundaries, commonly found in casted or molded parts, remain difficult to detect. Finally, the extraction algorithm must be designed in an ad-hoc way for different feature types and even for features with similar shapes. Thus, it is difficult to apply these methods of feature extraction when developing a feature-compliant finite element meshing. This work therefore proposes a feature extraction methodology that allows the extraction of features containing nonidentical geometries similar to a reference feature shape from a target shape. In this methodology, the reference feature shape and target shape are represented by a set of shape descriptors defined on triangular meshes; extraction is performed by finding similarities between the descriptors on the reference feature shape and those of the target shape under the projective transformation.
関連記事
