Ph.D. Thesis Summary

Stable, Distributed Real-Time Scheduling of Flexible Manufacturing Systems: An Energy Approach

Juergen Fenchel

March 1995

Dynamic market demands in terms of product lead time, quality, and manufacturing flexibility incur significant planning complexity for production strategies. Flexible Manufacturing Systems (FMS) have become widely accepted for medium batch quantity, medium part type (PT) variety production processes. In order to explore the FMS capabilities, it is necessary to develop scheduling policies which are both real-time capable, and flexible. Real-time capability prohibits applying methodologies of significant computational complexity at the systems level. Such methods may become feasible for a single machine in isolation with increasing computational capabilities. Flexibility of a scheduling policy pertains to the fact that different machines tend to require different strategies for PT selection in order to achieve their objectives, e.g. bottleneck server and dedicated single PT server require different approaches. A generic system property which allows to provide real-time decision making capabilities addresses the boundedness, i.e. stability, of the part type inventory levels.

In this dissertation, two principal approaches are used to analyze stability of real-time scheduling policies: (1) workload balance, and (2) Lyapunov function. Workload balance is used to develop criteria for stability when the setup time between PT is non-zero, for parameter values of existing policies, and for a new policy. The Lyapunov function approach is applied for the development of stable policies for different system topologies, and a model extension to incorporate performance into a stable scheduling policy.

Developed policies distinguish the number of processing states simultaneously present within a considered system to model different topologies. A process stage is a particular segment in the sequence of transformation procedures of a PT. Single-stage policies are developed which maintain a safety stock level, and branch arriving PTs into a set of distinct output models. Multi-stage policies separate between internal, and external reentrant PT flows between servers. An internal flow represents for example a rework condition for a PT. External reentrant flows incur to resolve server coupling and transportation delay issues. Furthermore, a policy is developed which allows for a PT to enter a server at different stages within the entire production process. A topology of particular interest considers parallel stages of servers. Multiple servers are arranged in parallel and the assigned workload between servers is to be balanced. Two categories of servers are separated in the context of the routing policy: base- (BSC), and peak (PCS) capacity servers. Whereas BCS allow several PTs to be assigned to the server for processing, PCS consider strictly one PT. The servers have identical operating conditions but different processing capabilities.