A consequence-finding approach for feature recognition in CAPP

Abstract

We present a rewriting approach for a consequence-finding inference of logic programs. Consequence finding restricts the derivations of a logic program to exactly those facts that depend on an explicitly given set of initial facts. The rewriting approach extends the Generalized Magic Sets rewriting, well-known from deductive databases, by an up propagation in addition to the usual down propagation. The initial motivation for this inference was to realize the abstraction phase of a knowledge-based CAPP system for lathe turning. The input to the CAPP system is a detailed description of a workpiece. During the abstraction phase characteristic parts, called features, are recognized for which predefined skeletal plans exist. Consequence finding is a method to restrict the computation such that exactly the features of the actual workpiece are derived. The same inference can also be used for checking integrity constraints: given an update of a deductive database or a logic program, consequence finding applies only those rules that are effected by the update operation.