http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
A Bayesian Mixture Inference of RUL Integrating Reliability Information
( Junhyun Byun ),( Suhong Min ),( Jihoon Kang ) 한국품질경영학회 2022 한국품질경영학회 학술대회 Vol.2022 No.1
현대 제조 공정이 복잡해짐에 따라 기존의 고장물리(Failure Physics) 기반 잔여수명(Remained Useful Lifecycle : RUL) 예측 방법론의 일반적 활용에 한계가 있다. 이에, 센서 데이터를 기반으로 한 머신러닝 모델을 활용하는 잔여 수명 예측 방법론이 광범위하게 개발되었으나, 대표적인 데이터 기반의 잔여수명 예측 방법론은 주로 단일모델의 예측성능에 의존되는 경향이 있어 이로 인한 예측 불확실성이 존재한다. 이에, 본 연구에서는 기존 방법론의 한계점을 해결하기 위해 기존 신뢰성 분포를 활용해 다양한 가상패턴을 생성하고 실제 데이터와 가상패턴 분포 간 적합도를 확률 가중치로 조합하는 혼합형 모델을 제안한다. 또한 예측 불확실성을 줄이기 위해 적합도 확률 가중치에 베이지안 업데이트 기법을 적용하여 실시간으로 확률적 적합도 가중치를 업데이트하는 방법을 제안한다. 본 수명예측 방법론은 성능 검토를 위해 일반적으로 널리 활용되는 지수함수 가중치를 반영한 가중 선형회귀(Exponentially Weighted Linear Regression) 모델과 비교해서 정확도를 검증했다.
Hojin Kang,Yongjae Kim,Junhyun Kim,Heeyeop Chae 한국진공학회 2021 한국진공학회 학술발표회초록집 Vol.2021 No.2
High selective etching process precision control is required in semiconductor devices. Perfluorocarbon compounds have intrinsic global warming issue. In this work, reactive ion etching process was conducted for SiO<sub>2</sub> and a-C in a dual frequency superimposed capacitively coupled plasmas (DFS-CCPs) reactor with C<sub>4</sub>H<sub>3</sub>F<sub>7</sub>O ether and alcohol plasmas. The etch rate and radical density showed that ether was higher than alcohol. Selectivity was superior when the low frequency power ratio was higher than high frequency power ratio.
Cho, Junhyun,Shin, Hyungki,Cho, Jongjae,Kang, Young-Seok,Ra, Ho-Sang,Roh, Chulwoo,Lee, Beomjoon,Lee, Gilbong,Kim, Byunghui,Baik, Young-Jin Springer-Verlag 2017 Frontiers in energy Vol.11 No.4
<P>Research on applying a supercritical carbon dioxide power cycle (S-CO2) to concentrating solar power (CSP) instead of a steam Rankine cycle or an air Brayton cycle has been recently conducted. An S-CO2 system is suitable for CSP owing to its compactness, higher efficiency, and dry-cooling capability. At the Korea Institute of Energy Research (KIER), to implement an S-CO2 system, a 10 kWe class test loop with a turbinealternator-compressor (TAC) using gas foil bearings was developed. A basic sub-kWe class test loop with a highspeed radial type turbo-generator and a test loop with a capability of tens of kWe with an axial type turbogenerator were then developed. To solve the technical bottleneck of S-CO2 turbomachinery, a partial admission nozzle and oil-lubrication bearings were used in the turbogenerators. To experience the closed-power cycle and develop an operational strategy of S-CO2 operated at high pressure, an organic Rankine cycle (ORC) operating test using a refrigerant as the working fluid was conducted owing to its operational capability at relatively low-pressure conditions of approximately 30 to 40 bar. By operating the sub-kWe class test loop using R134a as the working fluid instead of CO2, an average turbine power of 400 W was obtained.</P>
변준현(Junhyun Byun),민수홍(Suhong Min),강지훈(Jihoon Kang) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11
현대에 이르러 산업이 빠른 발전속도로 인해 제조 공정이 복잡해짐에 따라 기존의 고장물리(Failure Physics) 기반 잔여수명 (Remained Useful Lifecycle : RUL) 예측 방법론의 일반적 활용에 한계가 있다. 이에 센서 데이터를 기반으로 한 머신러닝 모델링을 활용하는 잔여수명 예측 방법론이 광범위하게 개발되었으나, 데이터 기반의 잔여수명 예측 방법론은 주로 단일모델의 예측성능에 의존되는 경향이 있어 이로 인한 불확실성이 존재한다. 이에, 본 연구에서는 해당 한계를 극복하기 위해 확률적 수명분포를 활용한 다양한 가상패턴을 생성하고 실제 데이터와 가상패턴 분포의 적합도를 확률 가중치로 조합하는 혼합모델을 제안한다. 뿐만 아니라, 베이지안 업데이트 기법을 접목한 실시간 적합도 확률 업데이트 열화 패턴 변화에 대한 불확실성을 줄였다. 본 수명예측 방법론의 성능 검토를 위해 일반적으로 널리 활용되는 지수함수 가중치를 반영한 가중 선형 회귀(Exponentially Weighted Liner Regression) 모델과 비교해서 정확도를 검증하였다. As the manufacturing process becomes more complicated due to rapid development of the industry, Existing RUL (Remained Useful Lifecycle) prediction method based on failure physics that used in general was hard to use. Thus, although machine learning model with sensor data in RUL prediction has been widely developed, data-driven RUL prediction methodology tends to depend primarily on the predictive performance of a single model and have uncertainty of degradation patterns change. In this paper, we proposed a mixture model combining several artificial hazard functions with their fitness of the performance. Then Bayesian update is adopted in real time to reduce uncertainty of degradation pattern change. We compared the performance of proposed idea, with existing Exponentially weighted linear regression (EWLR). Experiments with the simulations and real datasets demonstrate the effectiveness and usefulness of the proposed algorithm.
부분 유입 노즐을 적용한 초임계 이산화탄소 발전용 초고속 터보발전기 개발 연구
조준현(Junhyun Cho),신형기(Hyung-ki Shin),강영석(Young-Seok Kang),김병휘(Byunghui Kim),이길봉(Gilbong Lee),백영진(Young-Jin Baik) 대한기계학회 2017 大韓機械學會論文集B Vol.41 No.4
초임계 이산화탄소 발전사이클의 다양한 특성을 분석하기 위하여 Sub-kWe급의 소형 실험장치를 설계, 제작하였으며, 터보발전기를 개발하였다. 초임계 이산화탄소 발전용 터빈에서는 팽창비가 작고, 유량이 작기 때문에 터보발전기의 회전수가 높아지게 되고, 이에 따라 회전 부품의 선정, 터빈 공력설계, 축력 및 회전체 동역학 설계가 어려워지게 된다. 이에 터보발전기의 회전수를 줄이기 위하여 노즐의 여러 채널 중 1개의 노즐만 사용하는 부분유입 방법을 세계 최초로 초임계 이산화탄소 발전용 터보발전기에 적용하였으며, 회전체의 진동을 측정하여 부분유입 노즐을 적용함에도 회전체 안정성은 허용 범위내에 있음을 확인하였다. A Sub-kWe small-scale experimental test loop was manufactured to investigate characteristics of the supercritical carbon dioxide power cycle. A high-speed turbo-generator was also designed and manufactured. The designed rotational speed of this turbo-generator was 200,000 rpm. Because of the low expansion ratio through the turbine and low mass flowrate, the rotational speed of the turbo-generator was high. Therefore, it was difficult to select the rotating parts and design the turbine wheel, axial force balance and rotor dynamics in the lab-scale experimental test loop. Using only one channel of the nozzle, the partial admission method was adapted to reduce the rotational speed of the rotor. This was the world’s first approach to the supercritical carbon dioxide turbo-generator. A cold-run test using nitrogen gas under an atmospheric condition was conducted to observe the effect of the partial admission nozzle on the rotor dynamics. The vibration level of the rotor was obtained using a gap sensor, and the results showed that the effect of the partial admission nozzle on the rotor dynamics was allowable.
Experimental Study on the Element-Specific Positron Annihilation Spectroscopy
김용민,Junhyun Kwon,HamdyF. Mohamed,MoonJoo Hyun,Changsun Kang 한국물리학회 2009 THE JOURNAL OF THE KOREAN PHYSICAL SOCIETY Vol.54 No.6
Coincidence Doppler-broadening (CDB) and positron annihilation lifetime (PAL) techniques have been applied to study 12 pure elements. The existence of open-type defects was investigated in the elements. The characteristic features of the high-momentum part of the CDB spectra for elemental metals with/without defects were investigated. By normalizing the CDB spectra to the CDB spectrum of well-annealed Ti, ratio-curves are presented. The moment region range (12×10-3 moc < PL < 20 × 10-3 moc) was elucidated and provided important information on the elemental specificity. An empirical relation that linked the positron anities and the atomic number to the shape of the high-momentum part of the CDB spectra for 3-d elements was derived. The results imply that elements (closely located in the periodic table) can be identified from an analysis of the CDB spectra. Coincidence Doppler-broadening (CDB) and positron annihilation lifetime (PAL) techniques have been applied to study 12 pure elements. The existence of open-type defects was investigated in the elements. The characteristic features of the high-momentum part of the CDB spectra for elemental metals with/without defects were investigated. By normalizing the CDB spectra to the CDB spectrum of well-annealed Ti, ratio-curves are presented. The moment region range (12×10-3 moc < PL < 20 × 10-3 moc) was elucidated and provided important information on the elemental specificity. An empirical relation that linked the positron anities and the atomic number to the shape of the high-momentum part of the CDB spectra for 3-d elements was derived. The results imply that elements (closely located in the periodic table) can be identified from an analysis of the CDB spectra.
재귀적 베이지안 앙상블 모델링 기법을 이용한 사회경제지표의 다변량 예측
변준현(Junhyun Byun),민수홍(Suhong Min),강지훈(Jihoon Kang) 한국정보기술학회 2024 한국정보기술학회논문지 Vol.22 No.2
After the COVID-19 pandemic, there has been an increase in uncertainty in both domestic and international socio-economic conditions, highlighting the necessity of accurate predictions for socio-economic indicators(SEIs). Historically, equations, timeseries, and machine learning models have been widely used for forecasting SEIs, but, their predictive accuracy has been constrained by the characteristics of SEIs and limitations in prediction models. Thus we proposed the Recursive Bayesian Ensemble Model(RBEM) to reliably predict SEIs that exhibit high uncertainty over time. The proposed model probabilistically combines various prediction models with different forecasting principles, such as timeseries models and machine learning, and sequentially adopts recursive Bayesian update techniques. The predictive performance and stability of the proposed model have been demonstrated through the prediction of SEIs.