반도체 패키지 검사 공정의 데이터 변화 감지를 통한 불량 예측 모델의 갱신

Abstract

In semiconductor manufacturing, it is difficult to maintain a high level of quality due to the miniaturization of the process and the large capacity of the product. Also the quality requirement of the customer is increasing. Observing and managing the quality is an essential element in the semiconductor manufacturing process. Because the data distribution changes as the manufacturing process and inspection conditions change, the predictive model generated from the previous data does not match the new data, so the relevant model must be updated. In this paper, we propose a method to determine the predictive model update by detecting the change of the importance of variables in the semiconductor package test. The proposed method can classify the lots efficiently and with high accuracy in a continuously changing data distribution.