http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
이규원,김문석,김동호,이상근,Lee K. W,Kim M. S,Kim D. H,Lee S. G 한국초전도학회 2004 Progress in superconductivity Vol.6 No.1
Magnetic $T_{c}$ of composite superconductors has been studied for providing a standard method. Various magnetization-temperature curves of NbTi, $Nb_3$Sn and Bi-2223 wires were measured using a SQUID magnetometer. Magnetization-temperature curve of zero-field-cooled procedure showed larger values than fie Id-cooled procedure. To obtain higher resolution near the onset temperature, we employed a two-field-direction method which measures a magnetization-temperature curve of a specimen first in positive and then negative fields. Analytical comparison of the magnetic $T_{c}$, with the resistive T$_{c}$ was accomplished for three specimens. The magnetic $T_{c}$/ mettled showed more detailed information on superconducting state of a specimen than the resistive$T_{c}$/ method. We have also studied the field dependence of the magnetic $T_{c}$ from 5 Oe to 120 Oe, however, no significant difference on field strength was found in our three specimensns
기기 이상 감지를 위한 LSTM Autoencoder 구현
전승현(Jeon S.H.),박채린(Park C.L.),이규원(Lee G.W.),김성종(Kim S.J.),구본근(Gu B.G.) 한국정보기술학회 2022 Proceedings of KIIT Conference Vol.2022 No.12
산업 기기의 고장 여부를 미리 알 수 있다면 인력적으로나 경제적으로도 손실을 막을 수 있을 것이다. 실제 장비고장 예측을 위한 많은 기계학습 알고리즘이 연구되고 있지만, 기계의 이상 데이터가 많지 않아 모델을 학습시키기에 어렵다는 문제점이 있다. 본 논문에서는 산업 기기의 고장을 탐지할 수 있는 모델을 제안하고 정상들 사이에 고장 값이 섞여 있는 비슷한 데이터를 가져와 직접 시험해 본다. 모델 학습의 준비작업으로 비지도 학습 방법론 중 하나인 LSTM Autoencoder를 사용해 펌프 센서 데이터의 이상을 탐지하는 모델을 생성했고, 최종 실험 결과는 분류문제에 대한 성능을 측정할 수 있는 AUC(Area Under the Curve)가 0.834, 테스트 정확도가 0.999587을 보였다. If industrial equipment failure can be known in advance, loss in manpower and economically can be prevented. Many machine learning algorithms for predicting actual equipment failure are being studied, but there is a problem that it is difficult to train the model because there is not much abnormal data of the machine. In this paper, we propose a model that can detect failures of industrial equipment and test them directly by bringing similar data with failure values mixed between normals. As a preparation for model learning, a model that detects anomalies in pump sensor data is implemented using LSTM Autoencoder, one of the unsupervised learning, and results of the experiments showed that, the accuracy of AUC (Area Under the Curve) is 0.834, and the accuracy of test data is 0.9996.
김기웅,이용호,권혁찬,김진목,김인선,박용기,이규원,Kim K.,Lee Y. H,Kwon H,Kim J. M,Kim I. S,Park Y. K,Lee K. W 한국초전도학회 2004 Progress in superconductivity Vol.6 No.1
We consider design factors for a SQUID sensor array to construct a 52-channel magnetocardiogram (MCG) system that can be used to measure tangential components of the cardiac magnetic fields. Nowadays, full-size multichannel MCG systems, which cover the whole signal area of a heart, are developed to improve the clinical analysis with high accuracy and to provide patients with comfort in the course of measurement. To design the full-size MCG system, we have to make a compromise between cost and performance. The cost is involved with the number of sensors, the number of the electronics, the size of a cooling dewar, the consumption of refrigerants for maintenance, and etc. The performance is the capability of covering the whole heart volume at once and of localizing current sources with a small error. In this study, we design the cost-effective arrangement of sensors for MCG by considering an adequate sensor interval and the confidence region of a tolerable localization error, which covers the heart. In order to fit the detector array on the cylindrical dewar economically, we removed the detectors that were located at the corners of the array square. Through simulations using the confidence region method, we verified that our design of the detector array was good enough to obtain whole information from the heart at a time. A result of the simulation also suggested that tangential-component MCG measurement could localize deeper current dipoles than normal-component MCG measurement with the same confidence volume; therefore, we conclude that measurement of the tangential component is more suitable to an MCG system than measurement of the normal component.
정창훈,이규원 한국대기환경학회 2003 한국대기환경학회지 Vol.19 No.6
The flow field in multiple fluid sphere systems was studied analytically. The expanded zero vorticity cell model based on Kuwabara's theory (1959) was applied and the effects of gas slippage at the collecting surface were considered. Also, the solid sphere system was extended to fluid sphere including the effects of the induced internal circulation inside the liquid droplet spheres or gas bubble systems. As a result, the obtained analytic solution was converged to the existing solutions for flow field around solid and bubble sphere systems with proper boundary conditions. Based on the resolved flow field, the terminal velocity around the collecting fluid spheres was obtained. Subsequently, this study evaluated the most general solution for flow field around the multiple fluid sphere systems. The obtained flow field in multiple fluid sphere could be used as a fundamental consideration of wet scrubber design and devices for removing particles by fluid-fluid interactions.