Generalized Job Shop Scheduling for the Diffusion Area in Semiconductor Manufacturing

International audienceScheduling decisions in the diffusion area of a semiconductor manufacturing facilityhave an important impact on the overall performance of a plant. Consequently, we wantto optimize those decisions while taking real-world constraints into account. The problemwe address is a generalization of a job shop with multiple resources per operation. Weaim to optimize several objectives that comprise throughput and waiting time. The operations to be performed in the diffusion area are cleaning or furnace operationsthat need specific machines. To obtain practicable schedules, we consider complexmachines of that area in detail and include their components as resources in our model.Complex machines consist of tubes, boats and a load port. A tube is the place whereprocesses are conducted. A boat is a moveable carrier for wafers needed for all operationsin a machine. The load port is the place where wafers are loaded and unloaded.For those resources, we are given synchronization constraints that can be explained usingtwo characteristic examples: Firstly, if a loading operation uses the load port ofa machine, the corresponding unloading operation must use the load port of the samemachine. Secondly, after a specific boat is loaded, it cannot be used elsewhere beforebeing unloaded. So, a resource can be blocked even if no operation is currently using it. Many approaches from literature extend the job shop problem to the needs of realworldproblems. However, we are not aware of any work that addresses resource dependenciesas described above. For the given problem, we present a construction algorithmto find an initial solution. We propose a suitable disjunctive graph representation thatadapts ideas from Dauzère-Pérès and Paulli (1997) and introduce moves that form abasis for different metaheuristics. Finally, we present results of an implementation.