| World semiconductor sales reached an all-time high in 2004, exceeding {dollar}213 billion, according to the Semiconductor Industry Association. With the demand for integrated circuits continuing to grow, semiconductor manufacturers look for any improvements in efficiency that could potentially yield huge savings. These potential savings come from a variety of places within the industry, ranging from the manufacturing process to the supply chain.; This dissertation will focus on an issue in the manufacturing process called the class-constrained lot-to-order matching problem (CLOMP), where individual lots of integrated circuits are matched to specific customer orders, while seeking to optimize multiple objectives. Our inclusion of multiple classes causes the problem's complexity to significantly increase when compared to the single-class version.; We first formulate the problem as an integer program, and then decompose it into two stages. The first stage is a knapsack problem that loads the factory, while the second stage is a modified bin-covering problem that fills the selected orders. By effectively matching supply to demand using our proposed heuristics, we can reduce the costs associated with delivering the product to the customer. Based on solutions to the decomposed model, this dissertation makes recommendations regarding the most appropriate initial inventory level, first stage sorting algorithm and second stage matching heuristic under a variety of operating conditions. Overall, the apparent tardiness cost indexing function and the first-in, first-out with improved endgame matching algorithm produce the best results under most circumstances. |
