In the semiconductor manufacturing process, scheduling lots can have a dramatic effect on a fab’s efficiency. This type of scheduling process is called flow shop with job re-entry and is notoriously difficult to model and solve efficiently. The problem stems from tuning parameter values, as well as finding an optimal combination for every fab condition. The dispatch rule logic has numerous parameters, and its performance relies heavily on the ability to fine-tune parameter values according to work-in-progress and station availability.
To solve this challenge, we developed an efficient parameter tuning method based on Evolutionary Optimization combined with Simulated Annealing. Our results show that the approach outperformed other benchmarks and can be successfully used to dynamically optimize a dispatch rule’s parameter values.