In intelligent manufacturing systems design the concept of intelligence is closely related to that of integration. Using feature-based product models is thought to make complete CAD-CAM integration possible. However, this integration is difficultbecause some relevant manufacturing information is lost during the transfer of the part design into the CAD model. One way to overcome this problem is to consider manufacturing feature recognition problem as an interfacing task between CAD and CAMapplications. This article presents a plex grammar approach for feature recognition. The basic idea is to search the rnanufacturing feature logical latent structure, considering that generally engineering knowledge has a very strong structure. Theapproach is based on the recursive and generic concept of features, which are considered as structures having multiple attaching points. This approach performs at three levels. In the first level, the features primitives are searched, based on thehypothesis that robust relationships between some elementary feature structures better represent the feature structure from the local point of view. In the second level, features are clustered based on their primitives. Using the notion of dataaggregation, the ordering relations are discovered in each cluster among the primitives of each feature and the features of each cluster. Based on the precedent levels, the grammars are generated in the third level. Primitives searching, featuresclustering, and ordering relations allow one to determine a potential group of features that might be generated by the same grammar. Another advantage of decomposition, clustering, and ordering relations construction is that they provide a systematic wayof considering the problem of feature recognition for capturing the logical latent structure in the case of interactions. An application directed to polyhedral features shows the interest of this approach.
展开▼
