Essays on production planning and control

This dissertation explores several issues in production planning and control, each of which was motivated, to some degree, by on-site research conducted at Intel Corporation's semiconductor assembly and test facility in Chandler, Arizona.; Production planning in a large manufacturing company is a complex process, involving many people, tools, and functional organizations. I detail a series of regularly scheduled meetings at which decisions are made that transform an aggregate production plan into increasingly specific production schedules, and highlight the way Intel has adapted the drum-buffer-rope (DBR) scheduling policy for their manufacturing system. In the course of this description I identify a number of researchable issues, several of which receive detailed study in this dissertation.; A significant portion of my dissertation concerns work release; that is, the decision regarding when to release a new job onto the factory floor. The focus of this analysis is on a tandem queueing system. For a system with a single bottleneck resource, it is shown that operating the system as a closed queueing network from the beginning of the process to the bottleneck resource provides better system performance than using either a closed queueing model for the entire process (CONWIP), or any static release rule. I further explore via simulation the case of multiple bottlenecks and discover that a release rule that considers simultaneously the number of jobs before both bottlenecks significantly outperforms rules based on the status of either bottleneck independently.; The final chapters of my dissertation study the sequencing of jobs in a closed queueing network. For a given number of jobs in the system, the objective of sequencing rules should be to minimize server idleness, which translates directly into higher levels of throughput. By analyzing the Brownian model that approximates a queueing network in heavy traffic, one can gain insight into scheduling rules that will perform well in the original system. I identify a priority scheme that I call the Polytope-Filling policy, and use both explicit calculations and simulation results to show that, for several networks I have analyzed, this prioritization method outperforms other scheduling rules proposed in the literature.