Lot cycle time prediction in a ramping-up semiconductor manufacturing factory with a SOM–FBPN-ensemble approach with multiple buckets and partial normalization

Abstract

“Ramp-up” is the process of increasing the production rate of a semiconductor manufacturing factory from the first lot to whole volume. Lot cycle time estimation and control during production ramp-up in a semiconductor factory is even more important. However, in addition to the large fluctuation, there are also trends and jumps (sudden increases) in the lot cycle time, which makes it a very challenging task. A SOM–FBPN-ensemble approach with multiple buckets and partial normalization is therefore proposed in this study for lot cycle time prediction in a ramping-up semiconductor manufacturing factory, which was seldom investigated in the past studies. The proposed methodology is composed of two parts. In the first part, the multiple-bucket approach is applied to consider the ramp-up plan of the semiconductor manufacturing factory. Subsequently, the SOM–FBPN-ensemble approach is applied to predict the cycle time of every lot in the ramping-up semiconductor manufacturing factory. The buckets obtained in the first part become additional inputs to the SOM–FBPN, and every parameter is normalized into a range narrower than [0, 1] to reflect the difference between the future and the past. To evaluate the effectiveness of the proposed methodology, a production simulation model was employed in this study. According to experimental results, the prediction accuracy of the proposed methodology was significantly better than those of many existing approaches.