머신러닝 이용한 슈퍼접합 모스펫의 최적화 연구

초록

In this work, the the adoption of machine learning for optimization of superjunction MOSFET is investigated. Abundant data (on-resistance(RON), breakdown voltage(BV)) with various process parameters is earned by technology computer-aided design (TCAD) simulation. We also compare the prediction accuracy between eXtreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LightGBM). XGBoost shows higher accuracy than LightGBM. The use of machine learning is very effective way to reduce the cost and time of superjunction MOSFET development. Copyright ? The Korean Institute of Electrical Engineers.

키워드