On Updating a Virtual Metrology Model in Semiconductor Manufacturing via Transfer Learning

Abstract

Manufacturing processes are often subject to drifts over production cycles. This study applies the paradigm of Transfer Learning (TL) to support the updating of a Virtual Metrology (VM) model based on a Convolutional Neural Network (CNN). The VM is applied to a Chemical Mechanical Planarization (CMP) to predict the average material removal rate. Through the prism of a benchmark case study, this paper empirically investigates how transfer learning can improve the updatability of a VM CNN-based model.