Novel optimization approaches for global fab scheduling in semiconductor manufacturing

Overview & Synopsis

In semiconductor manufacturing, microelectronic components require several hundred operations on several hundred machines grouped into different work centers. Each work center specializes in handling one type of operation, and these are usually very different from one work center to another. These characteristics, along with reentrant flows and long cycle times (6 to 8 weeks) greatly complicate scheduling decisions. It is therefore very difficult to manage in detail all of the scheduling decisions in all the work centers of a factory. Thus, this thesis proposes a global scheduling approach based on a structure in two levels of the operational level (global level and local level). This approach aims at steering scheduling decisions at the work center level using production targets. These production targets are expressed as the quantities of components to be achieved for each operation and for each period over a scheduling horizon. Different mathematical models called global scheduling models (linear programs) proposed in the thesis determine these quantities. These global scheduling models correspond to different global scheduling strategies of the factory such as minimizing variability, controlling cycle times, etc. The local scheduling level aims to achieve the objectives set by the global scheduling models, while optimizing its own criteria and respecting its constraints. The approach is validated by experiments based on a simulation model and industrial data.