Data driven optimization for practical large scale operation of automated material handling system in semiconductor FABs

Abstract

This dissertation address the operation optimization of automated material handling system in semiconductor FABs. We focus on the large scale FABs which have high complex process. We suggest an operation that takes into account a number of constraints that arise from application to field. In this dissertation, we discuss three issues: (1) OHT vehicle dispatching problem, (2) MCS routing problem, (3) Storage selection problem. For the first topic, we measure OHT vehicle travel data and adapt the estimator which include the congestion using the reinforcement learning techniques. For the second topic, we define the MCS dynamic routing problem using multiple AMHS. We suggest dynamic routing algorithm using Q function value. For the third topic, we conduct a study that determines which storage is effective. Through data driven optimization for operation, we expect (1) operations beyond human insight and (2) autonomous operations without human operations.