Off-line learning based adaptive dispatching rule for semiconductor wafer fabrication facility

Abstract

Uncertainties in semiconductor manufacturing fabrication facilities (FABs) requires scheduling methods adaptive to real-time environments. This paper presents an off-line learning based adaptive dispatching rule (ADR) whose parameters are tuned dynamically with real-time information relevant to scheduling. First, we introduce the framework of ADR composed of dynamic dispatching, learning machine, and simulation model. Secondly, we discuss the workflow of the dispatching rule in detail. Thirdly, we use an immune cloning selection algorithm (ICSA) to find the relations between weighting parameters of the dispatching rule and real-time status information to adapt these parameters dynamically to the real-time environment. Finally, a real fab simulation model is used to demonstrate the proposed method. The simulation results show that ADR with changing parameters tracking real-time production information over time is more robust than ADR with constant ones, and improve the movements of WIP by about 2%.