7.0 SUMMARY AND CONCLUSIONS
In the current research, the CAD IGES model is converted to a Noodles solid model. Features are extracted from the Noodles representation and made accessible within a feature graph. PIMES makes extensive use of these features, along with knowledge-based productions encoding manufacturing heuristics and standards. For a limited set of features, PIMES can successfully detect violations of design leading to forming defects.
The feature recognition process is currently a bottleneck within the DFM environment because more complex features are required for design evaluation. Feature recognition is the subject of active research. Alternative approaches under investigation include Medial Axis Transformations and precedence graph grammars. Expansion of the knowledge base with feature-oriented heuristics continues. Even if feature recognition difficulties remain unresolved, the system can be used with little modification under a design-with-features scenario.
The domain of machine size determination and machine selection needs only the values of volume and surface area intrinsic within the Noodles solid model. The extension of PIMES into the domain of cost assessment, using part volume and material density information already available, is only natural.
It has been seen how object technology allows symbols to be fashioned within a computer system to resemble those acted upon by manufacturing and design experts. Expert system heuristics are structured as rules in the form of productions. These productions use object list differentiating processes similar to those employed by Fourth Generation Languages (4GLs). The single-tiered representations used within PIMES effectively serve the characteristic processes of production systems, 4GLs, and object-orientated programming.
In the future, the identified features and the results of the feature critiques could be stored back into the CAD model. A feature critique code could either be stored within a database along with elaborating message strings or stand on its own as a meaningful piece of data. Constraint violations can then effectively enhance the original geometric model; another advising process that is able to understand the standards involved can retrieve and make further use of these codes in another, perhaps unforeseen, evaluation.