MAXIMIZING THROUGHPUT USING DYNAMIC RESOURCE ALLOCATION AND DISCRETE EVENT SIMULATION

This research studies a serial two stage production system with two flexible servers which can be dynamically assigned to either station. This is modeled using discrete event simulation and more specifically the Arena software package by Rockwell. The goal is to determine dynamic allocation policies based upon the inventory level at each station to maximize the throughput of finished goods out of the system. This model adds to previous work by including actual switching time. The effect of the pre-emptive resume assumption is gauged, and the effectiveness of the OptQuest optimization package is also tested. Studies are conducted to determine the throughput of the system using easily implementable heuristics including when workers are together and separate. Additionally, the effect of buffer allocation and buffer sizing are studied, and it is shown that buffer allocation is not sensitive to changes in buffer ratio as long as there is buffer space available at each station while adding buffer space has a diminishing rate of return