Fixture-Driven Hybrid Process Planning

Automated synthesis of manufacturing process plans has been attempted through many methodologies for a wide range of manufacturing problems. In some cases, such as the manufacture of machined parts, these attempts have not been fully successful: to date, it has not been possible to develop purely generative systems to synthesize complete realistic plans for a wide range of realistic machined parts. In this paper we propose a new approach to process planning for machined parts, which integrates case-based reasoning and generative components. Introduction An autolnated process planner providing realistic process plans for a reasonably wide spectrum of products would make a great impact on industrial practice; and mechanical engineering and industrial engineering researchers have done much research on developing process planning systems. However, attempts to build purely generative process planners have had only limited success, primarily because this domain lacks sufficiently well-defined rules and approaches to work across the entire problem domain, and engineers developing process plans have to heavily rely on personal experience and informal manufacturing practices. In contrast, a number of successful semi-automated systems have been developed using an approach called variani process planning. Variant process planning is somewhat similar to case-based planning, in the sense that plans are stored in a database, so that they can be retrieved and modified for use in new planning problems. However, one of the primary differences is that the plan modification is done by the human user rather than by the computer system. If successful ways could be developed to extend the variant approach by using case-based reasoning techniques to do some of the plan modification automatically, this would significantly increase the practical utility of such systems. As a first step in this direction, we give a brief description of a new approach to process planning which reuse old process plans, adapting them to new designs. Background Increasing competition is challenging the manufacturing industry to bring new well-manufactured and competitively priced products to market as quick as possible. Long ago it was recognized that one of the most important steps to this goal is a making of effective process plaus (Chang 1990). A process plan unambiguously describes how a design can be manufactured from a corresponding stock and consists of ordered sequence of descriptions of manufacturing processes, where all relevant parameters of each process are specified. Process plans are similar to plans considered in AI literature--they are synthesized to achieve some goals and manufacturing processes used in them can create or delete preconditions for other processes. Despite the achievements of the past 20 years, the development of a good approach for automating process planning remains a very important and very difficult problem--existing systems are able to reliably handle only very restricted classes of designs (often not producing realistic process plans) and/or require very intensive human interaction. There are two primary approaches to CAPP--the variant and generative approaches. Generative process planning is analogous to plan generation in AI domains: the goal for the process planning system is to develop a complete plan for the proposed product design. Variant process planning is in some respects similar to the plan adaptation and reuse techniques explored by AI researchers, except that although the plan retrieval is done automatically, the plan adaptation is done manually. In process planning practice, variant techniques are the tools of choice: they currently support almost all practical implementations of Computer-Aided Process Planning (CAPP). Several variant systems are commercially available and have provided significant benefits--but despite the popularity of this approach, variant process planning has some well known drawbacks. A great deal of research has been done on generative approaches, and a number of experimental systems have been developed for various aspects of process planning. However, generative process planning has proved quite difficult. Most existing systems 55 From: AAAI Technical Report WS-98-15. Compilation copyright © 1998, AAAI (www.aaai.org). All rights reserved.