Plan construction and optimization in a multi-agent framework for computer-aided process planning

Up to now, Computer Aided Process Planning (CAPP) had only little impact on industrial practice. Restricted representations did not allow for capturing all relevant aspects of complex real-life problems, the methods were weak in multi-criteria optimization and, even simple CAPP models became computation-ally prohibitive. Here we present a new, multi-agent based approach that tries to overcome these difficulties. Our method is based on the co-operation of domain-specific geometry and technology related agents and a general purpose optimizer engine. While the specific agents, working on the exact geomet-rical representation of the part, available machines and tools, build up the space of the process plans, the solver extracts solutions from this space. A multi-objective evolutionary algorithm is applied to find Pareto sets of solutions that are best according to several criteria, such as changeover and production costs. The operation of the prototype CAPP system is demonstrated through examples in traditional machining