Learning-based release control of semiconductor wafer fabrication facilities

Release control plays an important role in the performance of a semiconductor wafer fabrication facility. A new release control policy based on extreme learning machine (abbreviated as RPELM) is proposed. Its main idea is to regulate the release sequence of the lots in a daily release plan subject to the attributes of the lots and running states of the fab. Firstly, the workflow of RPELM is introduced. Secondly, correlation coefficient method is used to select running states of the fab closely related to its performance. Finally, RPELM is validated and verified by a benchmark model (fab6 of MIMAC) and an actual 6 inch fab model (called BL), respectively. The simulation results show that RPELM performs better than common release policy with higher on-time delivery rate, especially for that of hot lots, without sacrificing the throughput and cycle time performance.