Performance analysis of scheduling policies for make-to-order systems producing multiple part types

Abstract

This thesis carries out a performance analysis of numerous scheduling policies for make-to-order systems, with a particular emphasis on semiconductor FAB lines. It also introduces the applications of a new developed scheduling policy, Control Point Policy (CPP) for semiconductor wafer FAB lines and compares its performance with popular make-to-order scheduling policies including Earliest Due Date (EDD), Minimum Slack (MS) and Critical Ratio (CR). Discrete event modeling is utilized to analyze the performance of polices for three important performance measures; cycle times, waiting times and inventory levels. New insights for system performance for CPP are developed by implementing it at multiple stations such as bottleneck stations with the use of finite size buffers between the workstations. Our simulation results demonstrate the capability of CPP to give the best system performance i.e lowest cycle times, minimum standard deviation of cycle times, low waiting times, minimum inventory levels and high service rates in an environment where parts are characterized by numerous ranking criteria. This ranking criterion determines the processing order for parts, when parts at numerous stages of production compete for limited resources (machines) simultaneously. It also addresses the introduction of “Hot Lots” in semiconductor manufacturing. Furthermore, in environments, when the products are explicitly classified based on an organization’s priorities such as new products, most profitable, most valuable customers etc. besides due date, CPP tends to give the best performance for the prioritized products while maintaining the best balanced performance for the remaining products. The results are demonstrated by our simulation experiments. This research also presents new opportunities for future developmental work for CPP.