반도체 제조라인 내 물류자동화시스템의 처리능력 향상을 위한 딥러닝 기반 디스패칭 방법론

초록

We present a deep-learning-based prediction method for the machine allocation problem of production scheduling in semiconductor manufacturing fabrication (FAB). This method is devised to improve the throughput capacity of the automated material handling system (AMHS). A prediction method is applied to determine the machine to perform the next process after a lot completes a process. Selecting the proper machine for the next process can shorten the travel distance of the overhead hoist transfers (OHTs), and this will eventually lead to reduced utilization and increased throughput capacity of the AMHS. The results confirm that the accuracy of our deep-learning-based machine selecting method is quite high and that it outperforms the other machine learning methods.

키워드