Abstract
In semiconductor manufacturing, the process is one of the most complicated manufacturing processes in the world. The process usually consists of more than 300 manufacturing steps that have complex interaction with each other. Frequent sampling can lead to redundant tests and increased cost, while infrequent sampling makes the quality of final products doubtable. There is a nontrivial tradeoff. How to decide on the optimal sampling policy is a critical, but challenging problem, in semiconductor manufacturing. In this thesis, we study the sampling problem in semiconductor industry and develop a mathematical model to characterize it. We use the simulation optimization technique to solve this model, and compare with multiple start schemes. We also derive the optimal sampling policy that can achieve minimum cost. A numerical experiment and an empirical study are conducted to verify the viability of the proposed model.