Genericity in Expert Process Planning Systems

Abstract

From an AI point of view, process planning is a typical planning problem that can be formulated as follows: given the initial state which is the raw material, the goal state which is the manufactured part, the repertory of permissible actions (operations) and the available resources (machines, tools,...), the problem is to determine the sequence of actions (process plan) which enables the goal state to be reached starting from the initial state. The fact that the process planning know-how required is typically ill-defined and intuitive explains why the expert system paradigm is so well suited to automate such a process. However, process planning being a vast collection of problem areas (e.g. machining, assembly, moulding, forging,..) and subareas (e.g. prismatic parts, rotational parts, sheet metal parts,... for the machining area), situations where different problem areas/subareas involving slightly different problem-solving methodologies would require distinct traditional expert systems. Generic expert systems may spare us this wasteful duplication of efforts by encompassing a broader scope of problem areas. We are presently developing a generic process planning expert system XPS-E whose scope includes subareas of machining and assembly. It is based on experience gained from GARI, a prismatic parts process planning expert system. In this paper, we describe a generic structure for internal representation and manipulation of process plans which confers a high degree of domain parametrizability to the system built around it.