Analysis and design of manufacturing control systems

Manufacturing control policies such as Kanban or Constant Work-in-process (CONWIP) are implemented to control the flow of information and materials in manufacturing systems so as to influence the performance of the system. This performance may be measured in terms of average work-in-process inventory, average finished goods inventory carried, cycle time, customer order fill rates, and system throughput. These measures usually conflict - lowering work-in-process inventory (WIP) will negatively impact the system's ability to respond to random demands, and thus have a negative impact on customer fill rates. We have developed a framework for the analysis and selection of control strategies for manufacturing systems. This framework consists of several elements: the Production Authorization Card System (Buzacott and Shanthikumar, 1992), a unified framework for modeling manufacturing systems; a simulation model capable of simulating complex manufacturing systems operating with this system; neural networks to approximate the simulation expected value functions; and an approach to developing optimal policy curves, a management tool to help identify the best policies in terms of multiple measures of system performance. In the research proposed here, we want to apply this framework to address some open production problems. These include the comparison of control policies on large, complex systems with several variants; the study of the impact of using imperfect advance demand information (forecasts) for both push and pull type systems; including priority schemes other than first-in, first-out as policy variables; and studying the impact of batching (lot-sizing). We also propose to continue the extenstion of the simulation model to incorporate more real world system complexities such as imperfect yield.