In this work we discuss the development of an expert system with
approximate reasoning which resorts to a new methodology for attribute
selection in knowledge-based systems. First, we make a survey of the
purifying process and its problems, as well as those of conventional
automatic control methods applied to industrial processes. Next, we
establish a definition of the relevance concept for a given set of
attributes, which includes the special case of non-relevant attributes or
nought attributes. A new heuristic is here proposed in such a way that it
finds out the more relevant attributes from those initially selected by
the expert, reducing the cost of the formation & validation of decision
rules and helping to clarify the underlying structure of a non
well-structured domain as are waste-water treatment plants.