Using expert networks as a predictor in manufacturing processes: Initial study

Two topics in artificial intelligence, expert systems and networks, have been used extensively to assist in problems that deal with incomplete or imprecise mathematical models or vague and ambiguous input data. However, these artificial intelligence techniques have drawbacks that limit their use in some intelligent manufacturing applications. Expert networks can represent the next generation of artificial intelligent tools for intelligent manufacturing. The purpose of this research is to look into the feasibility of constructing expert networks appropriate for manufacturing situations where expert systems or artificial neural networks have not been utilized. This research is exploratory in nature due to expert networks representing an artificial intelligence paradigm that has never been fully implemented in manufacturing. Expert networks have never been fully implemented in real applications; therefore there is high risk in the adequacy of the learning and training algorithms developed for such networks to be used in manufacturing applications. This paper will address some of the aspects of developing expert networks in particular for manufacturing processes.