• 목록

In a semiconductor wafer fabrication process, it is critical to identify abnormal wafers and to analyze the causes to improve the process. Because a massive number of wafers are fabricated, a large number of wafers are often inspected at the same time. In this case study, we assume a quality control system which displays quality inspection results of several wafers in a single chart. The purpose of this study is to define abnormal trends observed from the charts and further identify abnormal wafers that contribute to the trends. For this purpose, we employed object detection methods such as YOLO. Once the abnormal wafers are identified, their process log data are analyzed by the quality inspection system, so-called commonality analysis, to identify faulty machines that may have resulted in the abnormal wafers. The case study proved that the proposed definition of the abnormal trend is valid and the abnormal wafers identification method works well.