UBC Theses and Dissertations
Scheduling techniques for flexible semiconductor manufacturing tools Rostani, Shadi
In this thesis, we introduce Residency Constraints for cluster tools. A residency constraint is a limit on how long raw material can reside on either a processing module or the robot after the process is finished. Like a deadline, it specifies a point in time by which a task must be finished. However, unlike a deadline, the point in time is no longer a fixed point, and is determined dynamically when the system runs based on the completion time of other tasks. Our work is the first to attempt to address different forms of residency constraints in cluster tools. We investigate different models of cluster tools, and provide different scheduling techniques for each model. the first model that we investigate is a single-route model, which assumes that all wafers go through the same processing visits. First, we assume that a residency constraint is only allowed on processing modules, and provide a scheduling technique that searches the time and resource domains for an optimal solution. Then, we allow Residency Constraints on both the processing modules and the robot. For this model, we introduce another technique that uses a special kind of Linear Programming and a branch-and-bound method to search for the optimal schedule. In the next model, we use the buffer module to hold the partially processed materials and free the resources to improve the performance of the system. Our experiment shows that these techniques find the optimal schedule in a reasonable amount of time. The last model that we investigate is a multi-route model that supports the production of different products i n the same tool configuration. We provide a greedy algorithm that finds a near-optimal schedule for residency-free cases. For the residency-aware model, we provide both optimal and nearoptimal scheduling techniques. We use Simulated Annealing for finding the near-optimal schedule, and provide several optimization techniques to improve the quality of the near-optimal solution.
Item Citations and Data