Three essays on production planning in semiconductor manufacturing

Semiconductor manufacturing is one of today's most complicated manufacturing processes. Most of the recent research in semiconductor manufacturing focuses on demand satisfaction, maximization of equipment utilization, and maximization of throughput with some capacity constraints. In this research, the first chapter presents an overview of the semiconductor manufacturing technology and describes the manufacturing process. It also discusses a real world production planning application in a major U.S. semiconductor company and proposes a new methodology for capacity planning.; The second chapter addresses the assignment problem of wafers to alternative machine groups and the determination of equipment requirements by recognizing capacity limitations of the individual machines as well as reducing operating and investment costs related to the machines. A mathematical model, which is a variation of the well-known generalized assignment problem with continuous/integer variables and fixed charges, is developed for this problem. Given the intractability of the model, lower bounding and heuristic methods to solve the problem are proposed.; The third chapter extends the problem described in Chapter 2 to a dynamic environment where demands vary between time periods and inventory carrying is allowed. We investigate the tradeoff between allocating a constant capacity for the planning horizon and holding inventory and we present a mixed-integer mathematical model to minimize operating and investment costs related to the machines and inventory holding costs. We propose a Lagrangian relaxation based heuristic approach to solve the problem.; The fourth chapter presents an effective algorithm to identify a set of component families in order to effectively to sequence individual printed circuit boards to multiple, identical insertion machines. The objective of the component grouping problem is to maximize the sum of similarities between each component type and a designated component type chosen as the group median in an attempt to minimize the number of machines a printed circuit board needs to visit. Once the component families are formed, two objectives are considered in scheduling the printed circuit boards: minimizing the maximum completion time (makespan) and the mean flow time. Finally, in the fifth chapter we discuss the implications and conclusions of this study and give directions for future research.