Process changes in manufacturing are often done by trial and error, even when experienced domain personnel are involved. This is mainly due to the fact that in many domains the number of parameters involved is large and there exists only a partial understanding of interrelationships between them. This paper describes a framework for keeping track of process change experiments, before they qualify as full cases. Process changes happen as a result of diagnosis done by the expert, following which some therapy is decided. The paper also presents an algorithm for diagnosis and therapy based on induction on discrimination trees constructed on specific views on the set of problem parameters.
展开▼
