187037 research outputs found
Sort by
Optical see-through near-eye display using pixel μ-Lens
No full text학위논문(석사) - 한국과학기술원 : 기계공학과, 2025.2,[vii, 94 p. :]투명 근안 디스플레이(Optical See-Through Near-Eye Display, OST-NED)는 현실과 가상 이미지를 결합하며 몰입감 있는 경험을 제공하는 기술이다. 기존 시스템은 디스플레이가 눈 측면에 위치하고 화면과 외부 환경을 결합하는 광학 결합기가 눈 앞에 배치된 구조로 인해 복잡한 광 경로, 광 손실, 이미지 왜곡, 크기와 무게 증가 등의 문제가 있었다. 본 연구는 이러한 한계를 해결하기 위해 픽셀 마이크로 렌즈와 투명 마이크로 LED 패널을 활용해 디스플레이를 눈앞에 배치하는 새로운 설계를 제안한다. 픽셀 마이크로 렌즈는 마이크로 LED에서 발산하는 빛을 집속하고 방향을 제어함으로써 광학적 손실을 줄인다. 투명 마이크로 LED 패널은 높은 광학 투과율과 전기적 성능을 갖춘 구조로 설계되어, 외부 환경과 가상 이미지를 동시에 명확히 관찰할 수 있도록 한다. 두 광학 요소의 결합을 통해 외부 환경과 가상 이미지를 선명하게 표현할 수 있었으며, 빛의 경로를 단순화해 광 손실과 왜곡을 최소화하고 시스템의 크기와 무게를 줄였다. 이 설계는 제조 공정을 단순화하여 대량 생산 가능성을 높이며, 투명 근안 디스플레이가 가진 기존 한계를 극복할 수 있는 혁신적인 기술로 전망된다.한국과학기술원 :기계공학과
Prediction of anisotropic behavior in metal sheet using discrete plastic potential functions based on a VPSC polycrystal model
No full text학위논문(석사) - 한국과학기술원 : 기계공학과, 2025.2,[vi, 68 p. :]재료의 이방성은 미세구조에 의해 결정되며, 이는 재료의 전반적인 거동에 중대한 영향을 미친다. 따라서 적절한 이방성 모델을 결정하는 것이 중요하다. 본 연구에서는 기존의 현상학적 소성포텐셜 모델을 대체할 수 있는 이산 소성포텐셜 모델을 제안하였다. 점소성 자기일관성 모델을 사용하여 소성포텐셜 곡면의 점 데이터를 생성하였으며, 생성된 점 데이터를 선형 보간하여 평면응력에서의 이산 소성포텐셜 곡면을 완성하였다. 제안한 모델은 기존 모델들과 달리, 모델 내 이방성 계수의 피팅을 필요로 하지 않으며, 재료의 실제 물리적 구조를 반영한다.한국과학기술원 :기계공학과
동력 외골격 로봇의 독립적인 균형 유지를 위한 압력 중심 모델 기반의 발목 제어 전략
No full text학위논문(석사) - 한국과학기술원 : 기계공학과, 2025.2,[v, 79 p. :]Powered exoskeletons for individuals with complete paraplegia stand out as a promising technology to help users overcome the limitations of wheelchair dependence and independently perform various activities in daily life. To provide practical independence, powered exoskeletons must be capable of self-balancing against disturbances. In this study, a CoP-based ankle balancing strategy is proposed to enable self-balancing against disturbances caused by the pilot’s upper body movements. First, the safety range of the CoP was defined to ensure balance could be maintained using only the ankle. Then a stable double stance posture was generated through a human-in-the-loop process based on CoP information. And the disturbances caused by the pilot’s upper body movements were addressed by controlling the ankle position through a CoP model-based controller to maintain balance. The proposed CoP-based ankle balancing strategy was validated in both scenarios: during independent operation of the powered exoskeleton and during pilot tests with an able-bodied individual. Experimental results showed that without the CoP controller, disturbances caused significant CoP oscillations, often exceeding the safety range. In contrast, with the CoP controller, CoP oscillations were significantly reduced and consistently remained within the safety range even under larger disturbances. This study is expected to make a significant contribution to enhancing self-balancing capabilities against disturbances caused by the pilot’s upper body movements in powered exoskeletons.한국과학기술원 :기계공학과
쿠프만 심층 신경망을 이용한 수상선 동역학의 데이터 기반 모델링
No full text학위논문(석사) - 한국과학기술원 : 기계공학과, 2025.2,[iv, 26 p. :]Recently, data-driven dynamics modeling has attracted significant attention from researchers due to its ability to capture complex or hidden dynamic characteristics which conventional models cannot effectively address. Ship dynamics, in particular, exhibit strong nonlinearities and disturbances, primarily due to the dominant influence of hydrodynamic forces. In this context, the Koopman Operator has emerged as a promising approach for deriving globally linearized dynamics models for surface vehicles. This research introduces an effective method for modeling the dynamics of autonomous surface vehicles (ASVs) using a deep learning framework integrated with the Koopman Operator. The proposed network accurately identifies an end-to-end model without requiring any prior knowledge of the vehicle, demonstrating high prediction accuracy for surge, sway and yaw velocities compared with classical ship modeling method and linear/nonlinear autoregressive models. The identified dynamics model was validated through the maneuvering tests in both simulation and field experiment on inland water.한국과학기술원 :기계공학과
회분식 알루미늄 공정의 폐열 회수를 위한 열에너지 저장 시스템 설계
No full text학위논문(석사) - 한국과학기술원 : 기계공학과, 2025.2,[ix, 65 p. :]This study proposes a unified thermal energy storage system (UTESS) for waste heat recovery in aluminum batch processes, aiming to reduce fuel consumption and greenhouse gas emissions. Thermodynamic analysis was performed to evaluate the fuel-saving effects of preheated combustion air, and a mathematical model was developed to simulate the heat transfer and pressure drop within the UTESS. Based on the thermal gain ratio (TGR), energy analysis results were used to assess the system’s economic and environmental performance. Despite increased electricity costs due to the air blower, the proposed system’s specific life cycle cost (SLCC) was 2.9 USD/ton Al lower, and CO2 emissions were reduced by 10 kg/ton Al compared to the conventional system. These findings suggest the economic and environmental feasibility of the proposed system in aluminum production.한국과학기술원 :기계공학과
크라우드 소싱 데이터 가용성 기반 딥러닝 및 시뮬레이션 모델을 사용한 향상된 교통 상태 추정
No full text학위논문(석사) - 한국과학기술원 : 조천식모빌리티대학원, 2025.2,[iv, 41 p. :]Effective urban traffic management relies on accurate traffic state estimation. However, obtaining comprehensive traffic data from multiple sensors for deep learning (DL) models as well as simulation models is both challenging and resource-intensive. This dissertation explores leveraging crowdsourced data from third-party sources to enhance traffic state estimation for urban arterial networks. A DL model is developed that predicts average traffic speed by utilizing street view images, points of interest, and road network information as graph embedding. This DL model achieved an 82% accuracy rate, demonstrating its capability to provide rapid, approximate speed estimates for high-level traffic management decisions. For enhancing depth and accuracy of analysis, a simulation model calibration technique using nested genetic algorithm (NGA) is proposed to closely replicate near-real-world traffic conditions. The calibration process utilized limited speed data from critical links, supplemented by crowdsourced data, resulting in a 90% accuracy in travel time predictions for a complex road network. The well-calibrated simulation model proved beneficial for analyzing traffic states and generating various metrics within targeted road networks. The findings indicate that while DL models provide quick generalized traffic estimates, simulation models excel in delivering precise, location-specific traffic insights with minimal data. The study also demonstrates a novel data-fusion technique, combining sensor data with passive crowdsourced information, to achieve accurate traffic speed estimation. The values obtained from simulations showed promising proximity to real values, yielding a mean absolute percentage error (MAPE) of 0.29% for a simpler network and 6.31% for the best 10 routes, and 10.33% for all of the 15 priority routes within a complex network. The results highlight that leveraging diverse data sources, including crowdsourced information, can enhance traffic model accuracy and adaptability, particularly in regions with scarce extensive data from conventional sources.한국과학기술원 :조천식모빌리티대학원
고엔트로피 합금 전기화학 촉매 개발을 위한 열역학 기반 머신러닝 모델 설계 및 나노구조 규명
No full text학위논문(석사) - 한국과학기술원 : 신소재공학과, 2025.2,[v, 156 p. :]Green hydrogen, a type of hydrogen obtained via water electrolysis, is environmentally friendly and is gaining attention due to its large-scale energy storage capability. To produce such green hydrogen, it is essential to develop an electrocatalyst that enables efficient water electrolysis. Furthermore, the development of a system that integrates an air battery with a water electrolysis cell is expected to solve the power imbalance problem inherent to renewable energy characteristics. This thesis will first demonstrate a thermodynamics-informed machine learning model for designing bifunctional high-entropy alloy catalysts capable of both oxygen evolution reaction and hydrogen evolution reaction in water electrolysis. Second, the thesis will cover the design and the characterization of nanostructured trifunctional catalyst capable of oxygen reduction reaction (ORR), oxygen evolution reaction (OER), and hydrogen evolution reaction (HER) necessary for the Zn air battery-driven water electrolysis.
High-entropy alloy (HEA), a multi-principal element alloy with a high configurational entropy, exhibits unique structural, electronic, and physical properties that differ from low-entropy alloys such as binary and ternary alloys. The surface distortions due to their disordered solid solution structure and the tuning of the d-band center position from orbital hybridization are considered advantageous characteristics for electrochemical catalytic reactions. However, since high-entropy alloys typically contain five or more constituent elements, composition optimization through the Edisonian approach is challenging. Thus, recent studies aim to discover the optimal composition of HEAs through machine learning based on density functional theory calculations of candidate elements or by conducting active learning based on experimental data to quickly find reliable optimal points. However, the Gaussian process regression models and Bayesian optimization methodologies that form the basis of such optimization do not recognize the boundary condition that the sum of all constituent element compositions must be 1 when using high-entropy alloy concentrations as variables. As a result, even if the model learns boundary data at xi = 1 (where xi is the composition of the i-th element) once, the uncertainty at the boundary increases again as exploration continues at points far from the boundary.
In this regard, this study proposes an acquisition function that multiplies configurational entropy with the uncertainty term. Since the amount of information obtained varies by compositional location, this unique acquisition function effectively prevents the ill-convergence to local optima (e.g. binary and ternary alloy regions), which was reported in several composition optimization studies of HEAs. This study used manganese, iron, cobalt, nickel, copper, molybdenum, palladium, and platinum as candidate elements, with the noble metal compositions of palladium and platinum fixed at 5 at% to maximize the mass activity and achieve cost-competitiveness. The resulting alloy composition was FeCoNiMoPdPt, which falls within the high-entropy alloy region. This alloy was analyzed through transmission electron microscopy, density functional theory-based adsorption energy calculations, and multiple X-ray diffraction pattern analyses. The HEA with optimal composition showed an oxygen evolution reaction overpotential of 204 mV and a hydrogen evolution reaction overpotential of 33 mV, resulting in an overall water splitting overpotential of 237 mV.
Energy Storage Systems (ESS) are drawing significant interest as a solution to the power imbalance problem, which is regarded as a disadvantage of renewable energy sources. Among various ESSs, air batteries show potential as grid-scale energy storage systems due to their high theoretical energy density. Similarly, hydrogen (H2) is also becoming a subject of interest due to its high energy density (142 MJ kg-1), which is more than three times that of conventional fossil fuels. Nevertheless, current hydrogen production methods mostly involve a process that separates hydrogen from hydrocarbons, causing the problem of carbon dioxide emissions during production. Green hydrogen, a hydrogen that does not require any CO emissions, is generated by water electrolysis using renewable energy sources. However, attempts to produce green hydrogen using renewable energy sources face limitations in terms of energy utilization due to irregular amounts of energy generated over time.
To overcome this, air batteries that can discharge sufficient operating voltage (above 1.23 V) for water electrolysis have recently gained attention as power sources. It is also beneficial since air batteries exhibit high energy densities. Despite the advantages, they show significant limitations regarding power density due to the limited performance of oxygen reduction reaction catalysts and cannot guarantee stable cycle life. This issue is emphasized when the air battery is combined with a water electrolyzer since the two efficiencies have to be multiplied in order to convert the input energy to a form of hydrogen. Hence, it is essential to develop catalysts that are effective for both water electrolysis (oxygen evolution reaction and hydrogen evolution reaction) and it would be beneficial if the catalyst is able to be utilized as a cathode material for air battery, making the charge-discharge reactions (oxygen reduction reaction, oxygen evolution reaction) of air battery efficient.
Based on this, we have synthesized a graphene-sandwiched, heterojunction-containing chalcogenide (G-SHELL) catalyst material. G-SHELL is a trifunctional catalyst material capable of oxygen evolution reaction, hydrogen evolution reaction, and oxygen reduction reaction. When applied to zinc-air battery cathodes and water electrolysis anodes/cathodes, we were able to operate a self-driven water electrolysis cell. In this study, we synthesized a graphene-sandwiched, heterojunction-embedded layered lattice (G-SHELL) catalyst derived from a zeolitic imidazole framework (ZIF) located on the graphene oxide (GO) surface. G-SHELL was found to have conductive graphene layers sandwiched between MoS acting as electron conduction channels, and its hollow form enables fast ion transport. To verify the formation possibility of such structures, density functional theory (DFT) calculations of two heterostructures (CoS/MoS and graphene/MoS) were performed, and Cs-corrected scanning transmission electron microscopy (Cs-STEM), X-ray absorption near edge structure (XANES), and extended X-ray absorption fine structure (EXAFS) analyses were conducted to elucidate G-SHELL's heterojunction and bonding characteristics. Additionally, experimentally observed STEM images were further compared with simulated STEM images of graphene/MoS structures. The induced internal electric fields (IEF) between heterojunctions were confirmed to accelerate electron migration to active sites for the three electrochemical catalytic reactions, resulting in high activity.한국과학기술원 :신소재공학과
전압 조절 자기 이방성 기반의 전기적으로 재구성 가능한 물리적 복제 방지 기능에 대한 연구
No full text학위논문(석사) - 한국과학기술원 : 신소재공학과, 2025.2,[vi, 54 p. :]Physical unclonable functions (PUFs) are security devices that generate encrypted keys by harnessing the innate randomness of a hardware-based system. Here, we demonstrate an electrically reconfigurable spintronic PUF using voltage-controlled magnetic anisotropy (VCMA). The magnetic anisotropy is effectively modulated from perpendicular to in-plane and vice versa through the VCMA effect. It is found that randomly distributed magnetic domains are formed immediately after the magnetic anisotropy is restored from in-plane to the perpendicular direction. We utilize the random domain patterns as an entropy source for a spintronic PUF, exhibiting the ideal PUF characteristics, such as randomness with entropy close to unity and uniqueness. Furthermore, we demonstrate that the domain patterns can be regenerated by repeating the gate voltage application, enabling electrical reconfigurability in our PUFs.한국과학기술원 :신소재공학과
질소도핑 황화구리 이중모드 작동 공기아연전지 촉매 및 유기금속골격체 기반 수전해 전기화학촉매 연구
No full text학위논문(석사) - 한국과학기술원 : 신소재공학과, 2025.2,[v, 87 p. :]The advancement of green energy technologies is imperative to address global environmental challenges and transition toward sustainable energy systems. This study explores the development of advanced electrocatalysts for energy conversion technologies, focusing on zinc-air batteries (ZABs) and alkaline water splitting. Specifically, a nitrogen-doped copper sulfide (N-CuS) was synthesized and evaluated as a bifunctional electrocatalyst for oxygen evolution reaction (OER) and oxygen reduction reaction (ORR) at the cathode of ZABs, while metal-organic framework (MOF)-derived catalysts were investigated for hydrogen evolution reaction (HER) and OER in water splitting electrolyzers.
The N-CuS cathode was synthesized using an electrochemical anodizing process followed by low-temperature nitrogen plasma treatment, resulting in a highly mesoporous structure with the 75-fold increased surface area compared to the pristine copper cathode. ZABs equipped with the N-CuS cathode achieved a capacity of 788 mAh·g⁻¹ and an energy density of 916 Wh·kg⁻¹ under aerobic conditions. Remarkably, the N-CuS cathode enabled efficient ORR/OER with the dual-mode operations of zinc-air and zinc-copper modes, with sustained capacity under oxygen-deficient conditions. Detailed mechanistic insights were gained through ex-situ and in-situ characterization techniques. The N-CuS ZAB demonstrated a six-fold longer cycle life compared to conventional Pt/C + RuO₂ systems, highlighting its potential for renewable energy storage applications.
In parallel, MOF-derived electrocatalysts were developed to address the challenges of alkaline water splitting. Dual metal-decorated nitrogen-doped carbon catalysts, including h-NiFe/NC and h-PtCo/NC, were synthesized via acid etching and pyrolysis of zeolitic imidazole frameworks (ZIFs). These catalysts achieved efficient HER and OER, with the h-NiFe/NC catalyst exhibiting an OER overpotential of 201 mV and the best combination achieves an overall water-splitting overpotential of 321 mV. Additionally, the catalysts demonstrated high durability, without significant degradation over extended operation. The results underscore the potential of MOF-derived catalysts as promising alternatives to platinum group metals (PGMs) for green hydrogen production.
Together, these findings contribute to the advancement of cost-effective, scalable, and efficient electrocatalysts, paving the way for renewable energy solutions. The N-CuS and MOF-derived catalysts developed in this study offer promising pathways to realize sustainable energy storage and hydrogen production systems, which are vital for many advanced applications across a variety of fields.한국과학기술원 :신소재공학과
고방사선 환경에서의 저전력 무선 방사선 계측 시스템을 위한 SiPM multiplexing 기법 및 내방사선 10-bit R-2R DAC
No full text학위논문(석사) - 한국과학기술원 : 원자력및양자공학과, 2025.2,[iii, 39 p. :]This study developed key technologies for a low-power radiation detection and radiation-hardened wireless communication system capable of real-time monitoring of radiation leakage outside nuclear power plants under extreme conditions such as severe accidents. First, to address the detection area limitations of silicon photomultipliers (SiPMs), a radiation sensor, caused by their small single-cell size, a detector capacitance compensation circuit utilizing the Miller effect was designed. This circuit effectively improved the signal-to-noise ratio (SNR) and energy degradation associated with parallel SiPM connections. Second, to mitigate the reliability degradation of digital-to-analog converters (DACs), a critical component of wireless communication, under radiation environments, an R+Ron - 2R+2Ron structure was proposed to compensate for changes in switch on-resistance (Ron) caused by total ionizing dose (TID) effects. This system enhances the reliability of radiation monitoring during severe accidents, enabling prompt and accurate responses, and is expected to be applicable to advanced medical imaging devices, gamma cameras, and other fields.한국과학기술원 :원자력및양자공학과