"How Can AI Procedures Become More Effective for Manufacturing?"
Proceedings of the Artificial Intelligence and Manufacturing Research Planning Workshop,
published by The AAAI Press, Albuquerque, N.M., June 24-26, pp. 103, 111, 1996.
by Kovalerchuk, B., E. Vityaev, and E. Triantaphyllou
Abstract:
The problems of AI effectiveness in manufacturing for Design, Scheduling, Control and Process Diagnosis
are considered. We have developed an effective dialog procedure for a designer.
This procedure helps him to identify the needed parameters of a designed product, i.e., to distinct
acceptable parameters, non-acceptable parameters and parameters that require new intelligent procedure
to formulate and find an effective schedule. Often such approach can avoid complicated time-consuming computations. In Control we developed simple and robust control procedures, which join the advantages of pure/conventional interpolation and fuzzy control methods for design of quick and cost-effective controllers. In Process Diagnosis we overcome some difficulties of such known methods as neural networks, linear discriminate analysis and the method of nearest neighbors. The main difficulties that we overcome are related to the speed of a dynamic learning process and reliability of diagnosis. We also make the extracted diagnostic regularities easily understood by a manufacturing expert. This approach was successfully used for several tasks related to engineering and medical problems.
