지하공간 연결통로설치에 따른 증분가치산정 연구

권주혁,박환용 대한국토·도시계획학회 2004 國土計劃 Vol.39 No.7

폐탄광 인공습지의 수질과 메탄 발생 특성 연구

권주혁,한영수,조용찬,안주성,임길재 한국자원공학회 2018 한국자원공학회지 Vol.55 No.5

Global warming is accelerating due to methane release to the atmosphere. The contribution of mine areas among the anthropogenic sources of methane can not be ignored. However, the study on mine area as a methane source is insufficient. The purposes of this study are to monitor methane fluxes in constructed wetlands of mine areas and to investigate the effect of water quality on methane production. Water temperature was the most influential factor affecting methane emissions. The methane fluxes measured by a closed chamber method at aerobic wetlands were estimated to be -3.6~12.0 CH4 mg/m2/hr in seven mines in Korea. This range is similar to the methane flux estimated from European free water surface wetlands. As a future work, it is necessary to study the characteristics of the wetland substrates (or sediments), microbes using the substrates and their environmental control factors in order to qualitatively estimate the methane production in mine areas. 대기로 발생하는 메탄가스로 인해 지구온난화는 더욱 가속화되고 있다. 메탄가스의 인위적 발생원 중 광산 이 차지하는 비율은 적지 않다. 그러나 메탄가스의 발생원으로서 광산지역에 대한 연구는 미흡한 실정이다. 본 연구는 광산지역의 인공습지에서 발생하는 메탄에 대한 기초연구로서 메탄발생량을 측정하고 수질특성이 메탄발생에 미치 는 영향을 보고자 하였다. 메탄발생에 영향을 미치는 인공습지의 수질 특성 중 수온이 가장 큰 영향인자였다. 광산지 역의 인공습지 7곳에서 폐쇄형 챔버법을 통해 측정한 메탄 flux는 -3.6~12.0 CH4 mg/m2/hr로 산정되었다. 유럽의 자유 수면 인공습지에서 산정된 메탄 flux와 유사한 범위를 보였다. 향후 광산지역의 인공습지에서 발생하는 메탄에 대해 연구하기 위해 기질의 특성과 그 기질을 이용하는 미생물, 환경조건 조절 등의 정량적인 메탄 발생 특성에 대한 연구가 필요하다.

네트워크 최대 단대단 지연 시간의 최소화를 위한 강화학습 기반 스케줄러

권주혁,류지혜,정진우 한국통신학회 2022 韓國通信學會論文誌 Vol.47 No.9

In this paper, a reinforcement learning(RL)-based scheduler to minimize the maximum network end-to-end latency is implemented in a single agent environment and a multi-agent environment. The RL model used the double deep Q-network (DDQN) with the prioritized experience replay (PER). Since the agents are unable to identify end-to-end latencies in the multi-agent environment, the state and reward were obtained using the estimated end-to-end latencies. Four network topologies were implemented and simulated to compare the reinforcement learning-based scheduler, FIFO, round robin (RR), and a simple heuristic algorithm (HA). As a result of simulation in fixed packet generation scenarios, the RL-based scheduler achieved the minimization of maximum end-to-end latency in all the topologies. The FIFO and RR schedulers could not minimize the maximum end-to-end latency in any of the topology, and the HA could not minimize the maximum end-to-end latency in a single topology. In scenarios with random flows generation, the RL-based scheduler performed better than the FIFO and RR, but performed the same as or worse than the HA, depending on the topology. 본 논문에서는 네트워크 최대 단대단 지연 시간의 최소화를 위한 강화학습 기반 스케줄러를 single agent 환경과 multi agent 환경에서 구현하였다. 강화학습 모델은 double deep Q-network(DDQN)와 prioritized experience replay(PER)를 사용하였다. Multi agent 환경에서는 agent가 정확한 단대단 지연 시간을 알 수 없기에 추정 지연시간을 사용해 상태(state)와 보상(reward)을 구했다. 강화학습 기반 스케줄러의 성능을 확인하기 위해 4가지의 네트워크 토폴러지들을 구현하여 강화학습 기반 스케줄러와 FIFO, round robin(RR), 간단한 heuristic algorithm(HA) 들을 비교하였다. 시뮬레이션 결과 강화학습 기반 스케줄러는 고정적인 패킷 생성 시나리오의 토폴로지들에서 모두 최대 단대단 지연 시간의 최소화를 달성하였고 FIFO, RR은 모두 달성하지 못했으며 HA는 하나의 토폴로지에서 달성하지 못했다. 무작위로 플로우가 생성되는 시나리오에서 강화학습 기반 스케줄러는 FIFO, RR보다 좋은 성능을 보였으나 토폴로지에 따라 HA와 비교하여 같거나 떨어지는 성능을 보였다.

진공 저온 분사 공정을 통해 형성된 Fe계 비정질 재료의 적층거동 및 미세구조 변화 관찰

권주혁,박형권,이일주,이창희,Kwon, Juhyuk,Park, Hyungkwon,Lee, Illjoo,Lee, Changhee 한국재료학회 2014 한국재료학회지 Vol.24 No.1

Fe-based amorphous coatings were fabricated on a soda-lime glass substrate by the vacuum kinetic spray method. The effect of the gas flow rate, which determines particle velocity, on the deposition behavior of the particle and microstructure of the resultant films was investigated. The as-fabricated microstructure of the film was studied by field emission scanning electron microscopy (FE-SEM) and high resolution transmission electron microscopy (HR-TEM). Although the activation energy for transformation from the amorphous phase to crystalline phase was lowered by severe plastic deformation and particle fracturing under a high strain rate, the crystalline phases could not be found in the coating layer. Incompletely fractured and small fragments 100~300 nm in size, which are smaller than initial feedstock material, were found on the coating surface and inside of the coating. Also, some pores and voids occurred between particle-particle interfaces. In the case of brittle Fe-based amorphous alloy, particles fail in fragmentation fracture mode through initiation and propagation of the numerous small cracks rather than shear fracture mode under compressive stress. It could be deduced that amorphous alloy underwent particle fracturing in a vacuum kinetic spray process. Also, it is considered that surface energy caused by the formation of new surfaces and friction energy contributed to the bonding of fragments.

비동기식 지터 상한 보장 네트워크의 구현

권주혁(Juhyeok Kwon),정진우(Jinoo Joung) 한국통신학회 2021 韓國通信學會論文誌 Vol.46 No.12

사상 최대 레이테 해전의 전투경과 분석과 전투 교훈 고찰

권주혁 ( Kwon Ju Hyuk ) 미래군사학회 2017 한국군사학논총 Vol.6 No.1

As the greatest naval battle in the history, the battle of Leyte Gulf in Oct 1944 has very significant meanings in the Pacific War. Compared with the battle of Jutland which has a place as the second largest battle of all, the battle of Leyte Gulf has shown not only its bigger size but also all sort of the available warships and warfare machines in surface, under sea and sky. From the point of views in the tides of Pacific War, this battle has marked as the decisive battle which the back bones of IJN(Imperial Japanese Navy) was completely broken down and confirmed US naval power far higher level to IJN. IJN`s fleet nearly achieved its long-desired victory because of poor US command decisions. Instead, on the brink of success, the Japanese commander turned away and US Navy won another major victory. After naval victory at the battle of Leyte Gulf, US Force managed to press on Luzon, Okinawa, Iwojima and then finally made Japan unconditionally surrender in Aug 1945. This study focused in detail on the each phase of the five main battles in the battle of Leyte Gulf, namely at Palawan Passage, Sibuyan Sea, Surigao Strait, Samar Island and Cape Engano. Through this study, we can make some analysis what sort of factors influence on the results of the decisive battles, and also we can learn the valuable combat lessons which can be useful for the implementation of future combat operation which might take place in Korean peninsula.

PHREEQC를 활용한 인회석의 AMD 처리 효율 검증

권주혁(Ju-Hyeok Kwon),최정찬(Jung Chan Choi) 대한지질학회 2017 대한지질학회 학술대회 Vol.2017 No.10

지연시간의 통계적 분석을 통한 FAIR 프레임워크의 타당성 연구

권주혁(Juhyeok Kwon),정진우(Jinoo Joung) 한국통신학회 2021 韓國通信學會論文誌 Vol.46 No.2

진공상온분사(VKS) 공정에서의 비행입자 가속 기구 및 속도제어를 위한 가스 유량 효과에 관한 연구

박형권,권주혁,이일주,이창희,Park, Hyungkwon,Kwon, Juhyuk,Lee, Illjoo,Lee, Changhee 한국재료학회 2014 한국재료학회지 Vol.24 No.2

Vacuum kinetic spray(VKS) is a relatively advanced process for fabricating thin/thick and dense ceramic coatings via submicron-sized particle impact at room temperature. However, unfortunately, the particle velocity, which is an important value for investigating the deposition mechanism, has not been clarified yet. Thus, in this research, VKS average particle velocities were derived by numerical analysis method(CFD: computational fluid dynamics) connected with an experimental approach(SCM: slit cell method). When the process gas or powder particles are accelerated by a compressive force generated by gas pressure in kinetic spraying, a tensile force generated by the vacuum in the VKS system accelerates the process gas. As a result, the gas is able to reach supersonic speed even though only 0.6MPa gas pressure is used in VKS. In addition, small size powders can be accelerated up to supersonic velocity by means of the drag-force of the low pressure process gas flow. Furthermore, in this process, the increase of gas flow makes the drag-force stronger and gas distribution more homogenized in the pipe, by which the total particle average velocity becomes higher and the difference between max. and min. particle velocity decreases. Consequently, the control of particle size and gas flow rate are important factors in making the velocity of particles high enough for successful deposition in the VKS system.

DDQN을 활용한 강화학습 기반 타임슬롯 스케줄링

류지혜,권주혁,정진우 한국통신학회 2022 韓國通信學會論文誌 Vol.47 No.7

To adopt reinforcement learning in the network scheduling area is getting more attention than ever, because of its flexibility to adapt to the dynamic changes in network traffic specifications and their requirements. In this study, a timeslot scheduling algorithm based on priority is designed using Double deep q-network (DDQN), a reinforcement learning algorithm. To evaluate the behavior of the DDQN agent, a reward function is defined based on the difference between the estimated delay and the deadline of packets transmitted at timeslot; and on the priority of packets. The simulation showed that the designed scheduling algorithm performs better than the existing algorithms such as the strict priority (SP) or weighted round robin (WRR) scheduler, in the sense that more packets have arrived within the deadline. By using the proposed DDQN-based scheduler, it is expected that autonomous network scheduling can be realized in the upcoming network framework. 트래픽 특성과 요구사항의 다양한 변화에 네트워크 자율적으로 유연하게 적응하고 대처하는 방안으로, 강화학습을 적용한 네트워크 스케줄러가 최근 주목받고 있다. 본 연구에서는 딥러닝을 적용한 강화학습 모델인 double deep q-network(DDQN)을 사용해 우선순위 기반의 타임슬롯 스케줄링을 구현한다. DDQN 에이전트의 행동에 대한 가치를 평가하기 위해 reward는 timeslot에서 전송된 패킷의 추정 delay와 deadline의 차이, 그리고 패킷의 우선순위에 기반해 지급하는 function으로 정의하였다. 시뮬레이션 결과, 학습된 스케줄러가 strict priority (SP) 혹은weighted round robin(WRR)과 같은 기존 알고리즘으로 스케줄링했을 때 우려되는 문제점을 극복한 것을 확인할수 있었다. 또한, 기존 스케줄러보다 높은 누적 보상의 합인 score를 기록하며, deadline 내에 더 많은 packet이 도착함을 확인하였다. 본 연구를 통해서 대규모 유선 네트워크에서 자율 네트워크 스케줄링 기능 실현의 가능성을타진하였다. 특히 제안하는 DDQN 기반 강화학습 에이전트를 사용하면 자율성과 성능을 모두 개선할 수 있을 것으로 기대한다.