| Summary: | 碩士 === 元智工學院 === 工業工程學系 === 84 === A method for recognition of machining features from a
boundary-representation
solid modeler is presented. This new approach is developed by
applying three methods, face-adjacency matrix,
neural network and fuzzy theory. The
first step is to convert the B-rep. data of the part into a pre-
designed face-adjacency matrix. The face-adjacency
matrix then can be input to the neural network. If there
are more than two recognized results with the same machined
volumes or the machined volumes are close to each other, the
developed fuzzy rules are used to obtain a better
result where features are recognized with maximum
volume extent.
The contribution of this thesis mainly solves the principal
drawbacks of previous feature-recognition
approaches including template comparing,
unrecognizable intersecting features. Moreover, the recognized
features are made to be suitable from manufacturing
point of view by using fuzzy method by considering
five manufacturing factors : number of features, number of
dispartured features, number of closed corners, number of
freedom directions and number of non-continuous
features.
Finally, the system is implemented with programming in C++ with
the assist of the ACIS solid modeler software on SGI
Indy workstation machine. Several example parts
are modeled and tested. The result shows that the developed new
method is capable of recognizing complex interacting features
and is suitable from manufacturing point of view.
|
|---|
