A hybrid ANN-FIR system for lot output time prediction and achievability evaluation in a wafer fab

Abstract

A hybrid artificial neural network (ANN)-fuzzy inference rules (FIR) system is constructed in this study for lot output time prediction and achievability evaluation in a fabrication plant (wafer fab), which are critical tasks to the wafer fab. At first, a hybrid and recurrent ANN, i.e. self-organization map (SOM) and fuzzy back propagation network (FBPN), is proposed to predict the output time of a wafer lot. According to experimental results, the prediction accuracy of the hybrid ANN was significantly better than those of some existing approaches. Subsequently, a set of fuzzy inference rules is established to evaluate the achievability of an output time forecast. © 2007 Springer-Verlag Berlin Heidelberg.