• 목록

Forecasting yield output in complex multistage manufacturing process poses significant difficulties. To address this challenge, we propose a methodology to predict yield with minimal available data - specifically, only the production path sequence and the corresponding yield results. First, our approach starts by mapping the multistage manufacturing process onto a graph structure. We then employ a graph convolutional network (GCN) to develop a yield prediction model. This methodology is expected to enable yield predictions in scenarios where only machine sequence information is available, with all other data being unknown or inaccessible.