Simulation Modeling and Analysis
Simulation Modeling and Analysis
| Hi-index | 0.00 |
The oxidation and diffusion ovens in wafer fabrication are batch processors, where only jobs belonging to identical job families can be processed together. In this paper, we compare the performance of a proposed online heuristic based on Model Predictive Control against a popular look-ahead method called NACHM. Simulation results show that the MPC-based heuristic, with properly selected parameters, can have up to 16.67% shorter mean cycle time than NACHM under uncorrelated job arrivals. Under positively correlated job arrivals, the mean cycle time of jobs passing through the batch processor is almost always significantly reduced for both the MPC-based policy and NACHM. The simulation results also suggest that increased correlation generates less improvement for policies that foresee events longer into the future, as NACHM improves at a faster rate than the MPC-based heuristic. Thus, when the correlation is sufficiently high (0.7) and the traffic intensity is low (0.5), the MPC-based heuristic, which considers events that occur farther into the future, has higher mean cycle time (from 1.92% to 9.47%) than NACHM. Controlling processors in front of the batch processor with the anticipated needs of the batch processor, successive job arrivals to the batch processor may result in positively correlated job families. Our results highlight the potential benefits of constraining the production of the upstream processor according to the anticipated needs of the batch processor.