关键词: 相交特征, 特征分割, 规则驱动, 痕迹识别

Abstract: To address the challenge of effectively segmenting and classifying complex machining features in B-Rep models,where features are often interlaced,nested,and exhibit ambiguous boundaries,a parts feature segmentation method that integrates rule-driven and trace recognition was proposed. The method treated convex edges between machining features as generalizable interaction traces,independent of specific feature types,and built a corresponding set of segmentation rules. Initially,transitional features including fillets and chamfers were identified and removed to simplify the model′s topology and extract initial feature face groups. Convex edge detection was then employed to identify intersecting features,followed by a preliminary classification of feature faces based on concave adjacency. For faces with multiple possible classifications,a geometric similarity metric was introduced to guide optimal assignment. In scenarios where multiple features share a single face,three typical fusion patterns were distinguished based on the fusion face type and convex edge location,with corresponding segmentation rules established to achieve effective segmentation of different fusion structures. Finally,the merging of similar face groups and the consolidation of isolated faces further improve the completeness and stability of the feature segmentation results. Experimental validation on representative parts with complex structures demonstrates that the proposed method offers strong generality and practical applicability in handling the segmentation of highly coupled machining features.

Key words: intersecting features, feature segmentation, rule-driven, trace recognition