Open AccessA rule-based approach with multi-level feature taxonomy for recognition of machining features from 3D solid models
Author(s) -
Andri Pratama,
Riona Ihsan Media
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1450/1/012128
Subject(s) - adjacency list , machining , feature (linguistics) , feature recognition , computer science , pattern recognition (psychology) , artificial intelligence , face (sociological concept) , taxonomy (biology) , software , visibility , data mining , algorithm , engineering , geography , mechanical engineering , linguistics , philosophy , botany , biology , social science , sociology , meteorology , programming language
Numerous approaches in recognition of intersecting and isolated features have been proposed in the last several decades. However, they are limited to features with topologically fixed shapes and restricted to isolated machining features since they are dependent on the predefined patterns or rules. In the present work, a rule-based approach is developed to accommodate intersecting features with variable topology shapes. The proposed approach classifies the features according to the multi-level feature taxonomy. In the first level, features are categorized into three groups of primitive features according to their loops and edges types. In the second level, pockets and holes are identified from their primitive feature attributes while visibility maps are adopted to recognize slots and steps features. Intersecting features are identified based on adjacency relationships among face members of the features. On the other hand, pre-defined rules are still utilized in restricted application to identify special machining features. In addition to that, the proposed approach has been implemented to recognize machining features in the industrial parts model in the b-rep format. The implementation result shows the proposed methodology has enlarged the number of identified features up to 55.7 percent compare to the existing method in a commercial software.