A hybrid feature recognition system using feature hints, graph manipulation
s and artificial neural networks for the recognition of overlapping machini
ng features is presented. Based on the enhanced attributed adjacency graph
(EAAG) representation and the virtual link graph (VLG) of a designed part,
the face loops (F-loops) are defined as the generalized feature hints. They
are then extracted From the EAAG using vector calculations, and the relati
onships between the F-loops are established. Next, the F-loops are manipula
ted according to the six types of the relationship between F-loops to build
the F-loop subgraphs (FLGs), which are potential features. Finally, these
FLGs are presented to a trained artificial neural network using various ove
rlapping feature cases to be classified into different types of feature. By
utilizing the characteristics of three intelligent techniques in the diffe
rent subtasks of the feature recognition process, the system can recognize
complex overlapping machining features with planar faces and quadric surfac
es efficiently. The system is open and has the capability to recognize new
types of overlapping feature from the learning ability of the artificial ne
ural networks.