A manufacturing system characterized by its stochastic nature, is defined by both qualitative and quantitative variables. Often there exists a situation when a performance measure such as throughput, work-in-process or cycle time of the system needs to be optimized with respect to some decision variables. It is generally convenient to express a manufacturing system in the form of an analytical model, to get the solutions as quickly as possible. However, as the complexity of the system increases, it gets more and more difficult to accommodate that complexity into the analytical model due to the uncertainty involved. In such situations, we resort to simulation modeling as an effective alternative.

Equipment selection forms a separate class of problems in the domain of manufacturing systems. It assumes a high significance for capital-intensive industry, especially the semiconductor industry whose equipment cost comprises a significant amount of the total budget spent. For semiconductor wafer fabs that incorporate complex product flows of multiple product families, a reduction in the cycle time through the choice of appropriate equipment could result in significant profits.

This thesis focuses on the equipment selection problem, which selects tools for the workstations with a choice of different tool types at each workstation. The objective is to minimize the average cycle time of a wafer lot in a semiconductor fab, subject to throughput and budget constraints. To solve the problem, we implement five simulation-based algorithms and an analytical algorithm. The simulation-based algorithms include the hill climbing algorithm, two gradient-based algorithms biggest leap and safer leap, and two versions of the nested partitions algorithm.

We compare the performance of the simulation-based algorithms against that of the analytical algorithm and discuss the advantages of prior knowledge of the problem structure for the selection of a suitable algorithm.