Convolutional Neural Networks for Automatic Virtual Metrology

To ensure stable manufacturing and high yield of production, factories (e.g., semiconductor or TFT-LCD fabs) conduct quality inspection on workpieces. They tend to adopt sampling inspection in consideration of reducing cost and cycle time, yet that fails to achieve real-time and online total inspection because of the sampling strategy and metrology delay. Automatic Virtual Metrology (AVM) is the best solution to tackle the problem mentioned above, due to the fact that it can convert sampling inspection with metrology delay into on-line and real-time total inspection. However, with the advancement of science and technology, the processes become more and more sophisticated, and the requirement for the accuracy of virtual metrology becomes higher. The current AVM prediction algorithm is the traditional machine learning method, Back-Propagation Neural Networks (BPNN). However, even if the amount of data in this method increases, the performance improvement has its limits, and it requires a strict and time-consuming feature selection process. To improve the prediction accuracy, this work proposes the deep learning method, Convolutional Neural Networks (CNN), for the AVM server. The accuracy of CNN improves as the amount of data grows. In other words, if there are sufficient data, the current accuracy limit of machine learning can be enhanced. Experimental results reveal that CNN can automatically extract highly informative features from the data and improves the original AVM accuracy.