http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
반도체 제조 공정에서 식각 종점 탐지를 위한 TadGAN 적용
최준규(Jungyu Choi),김보배(Bobae Kim),홍승모(Seungmo Hong),임성빈(Sungbin Im) 대한전자공학회 2021 대한전자공학회 학술대회 Vol.2021 No.6
Along with rapid computational improvements in computers, machine learning has rapidly evolved and is widely used in various fields. Time series anomaly detection can provide information about the serious situations faced in various fields, from finance and aerospace to IT, security, and healthcare. Likewise, in the semiconductor manufacturing process, much attention is currently being paid to machine learning. Quality management, process monitoring, and maintenance are becoming important as cutting-edge manufacturing technologies become increasingly complex, faster, and more automated. Therefore, this paper proposes a TadGAN-based EPD detection algorithm, a time series anomaly interval detection model using GAN, in the plasma etching process. The performance of the proposed model is evaluated by how accurately we distinguished between non-EPD and EPD over time.
3차 수동소자 상호변조 왜곡을 추정하기 위한 FTDNN와 CVF의 성능 비교
김지혜(ZHIHUI JIN),최준규(Jungyu Choi),임성빈(Sungbin Im) 대한전자공학회 2020 대한전자공학회 학술대회 Vol.2020 No.8
In this paper, the focused time delay neural network (FTDNN) and the cubic Volterra Filter(CVF) are used to model the third—order passive intermodulation distortion (PIMD) caused by the downlink signal in the frequency division duplex (FDD) wireless communication system. Passive intermodulation distortion is an interference signal generated by the non—linearity of a passive element in a wireless communication environment, which causes high noise level of a receiver, and a deterioration of communication quality. Measured PIMD signals and transmitted downlink signals are used to estimate and evaluate the performances of the proposed approaches.
3차 PIMD 모델링을 위한 2-입력 Cubic Volterra Filter의 RLS 구현
임성빈(Sungbin Im),김보배(Bobae Kim),김지혜(Zhihui Jin),송금자(Jinzi Song),최준규(Jungyu Choi) 대한전자공학회 2021 전자공학회논문지 Vol.58 No.12
무선 분산 안테나 시스템의 수동 소자는 서로 다른 주파수 대역의 전송 신호를 비선형적으로 왜곡하여 인접 대역에서 예상치 못한 신호를 초래할 수 있다. 이를 수동 상호 변조 왜곡 (PIMD)이라고 한다. 불행히도 PIMD는 때때로 이동통신망의 상향링크 대역에 간섭으로 입력되어 장비의 수신 성능을 저하시킨다. PIMD를 완화하기 위해서는 하향링크 신호와 PIMD 간의 관계를 설명하는 모델이 필요하다. 본 논문에서는 이러한 모델로 2-입력 입방형 볼테라 필터(CVF)를 사용한다. 또한 수동 부품은 노화 및 작동 온도로 인해 시변이기 때문에 모델에 이러한 속성이 반영되어야 한다. 이러한 이유로 2-입력 CVF의 RLS 기반 재귀 구현을 개발하고 제안된 구현의 성능을 Batch 접근 방식과 비교하여 조사한다. 시뮬레이션 및 측정된 데이터를 사용한 실험을 통하여 제안된 데이터가 배치 접근 방식과 호환됨을 보여준다. Passive elements in a wireless distributed antenna system can nonlinearly distort transmission signals of different frequency bands, which result in unexpected signals in the adjacent bands. This is called passive intermodulation distortion (PIMD). Unfortunately, PIMD sometimes is introduced into the mobile communication network"s uplink band as interference, which lowers the network performance. For mitigating PIMD, the models descrbing the relationship between the downlink signals and the PIMD are required. In this paper, the two-input cubic Volterra filter (CVF) is employed for the model. In addition, since the passive components are time varying due to aging and operating temperature, the model should reflect this property. For this reason, we develop a RLS-based recursive implementation of the 2-input CVF, and investigate the performance of the proposed implementation compared to the batch approach. The experiments with simulated and measured data demonstrate that the proposed one is compatible with the batch approach.