187037 research outputs found
Sort by
Impacts of renewable energy integration on military energy independence and operational effectiveness
No full text학위논문(석사) - 한국과학기술원 : 미래전략대학원프로그램, 2025.2,[iii, 52 p. :]본 연구는 재생에너지 도입이 군대의 에너지 자립도와 작전 효율성에 미치는 영향을 분석한다. 군대는 에너지를 대량으로 소비하는 집단으로 안정적인 에너지 공급은 군사 작전의 지속성과 임무 성공에 필수적이다. 그러나 한국의 군대는 화석연료에 크게 의존하고 있어, 에너지 가격 변동성과 공급 불안정으로 인해 작전 지속성을 저해하고 국방 예산에 경제적 부담을 초래하는 취약성을 보인다. 이러한 문제를 해결하기 위해 군사 작전에 재생에너지를 도입하는 방안을 제시한다. 통제집단합성법을 활용하여 재생에너지를 도입한 군 부대와 도입하지 않은 부대를 비교 분석한 결과, 재생에너지를 도입한 부대가 더 높은 에너지 자립도와 작전 효율성을 보였으며 동시에 탄소중립 목표 달성에도 긍정적인 기여를 하는 것으로 나타났다. 이는 기후변화 대응과 탄소중립을 강조하는 전 세계적 흐름에 부합한다는 점에서 중요한 의미를 가진다. 본 연구는 군사 분야에서의 재생에너지 활용이 국가 안보와 경제적 효율성을 동시에 달성할 수 있는 전략적 대안임을 입증하며, 이를 통해 군사 작전의 효율성을 향상시키고 에너지 안보를 강화하기 위한 구체적인 정책적 방향을 제안한다.한국과학기술원 :미래전략대학원프로그램
(A) study on the improvement of governance of the national large research facilities construction project : from a new institutional perspective
No full text학위논문(석사) - 한국과학기술원 : 미래전략대학원프로그램, 2025.2,[iii, 59 p. :]최근 누리호 발사, 4세대 다목적 원형 방사광가속기 신규 구축 추진 등 과학기술분야 대형연구시설 구축사업에 대한 관심이 커지고 중요성도 강조되고 있다. 하지만 한편으로는 여러 사업에서 일정 지연, 예산 초과 등의 문제가 발생하고 있다. 정부는 이러한 문제점을 해결하기 위해 대형연구시설 구축관리 표준지침을 제정하고 국가과학기술심의위원회 산하 전문위원회를 신설하는 등 대형연구시설의 체계적인 구축을 위한 종합사업관리(PM) 제도의 도입을 적극 추진하고 있다. 하지만 PM 제도 도입의 효과가 충분히 나타나지 않는 상황이다. 이 글은 신제도주의의 관점에서 PM 제도 도입을 ‘제도 변화’의 측면에서 분석하여 그 원인을 파악하고 개선점을 제시하고자 하였다. 구체적으로는 ‘아이디어’의 개념을 활용하여 PM 제도 도입 과정을 설명하고 선진국의 PM 제도와 우리나라의 PM 제도 비교 분석함으로써 거버넌스 측면에서 PM 제도가 불완전하게 도입되었음을 밝혀냈고, 프로그램 논리 모형을 활용하여 정책 효과를 분석함으로써 복잡한 거버넌스가 우리나라에서 PM 제도 도입의 효과가 충분히 나타나지 못하게 하고 있음을 밝혔다. 마지막으로는 이러한 문제를 해결하기 위한 거버넌스 측면에서의 개선 방안을 신뢰 기반의 의사결정체계 구축, 권한과 책임의 명확화, 법적·제도적 기반 마련으로 구분하여 제시하였다.한국과학기술원 :미래전략대학원프로그램
전문가 지식을 활용한 검색 기반 무기물 역합성 분석
No full text학위논문(석사) - 한국과학기술원 : 산업및시스템공학과, 2025.2,[iii, 24 p. :]While inorganic retrosynthesis planning is essential in the field of chemical science, the application of machine learning in this area has been notably less explored compared to organic retrosynthesis planning. In this paper, we propose Retrieval- Retro for inorganic retrosynthesis planning, which implicitly extracts the precursor information of reference materials that are retrieved from the knowledge base regarding domain expertise in the field. Specifically, instead of directly employing the precursor information of reference materials, we propose implicitly extracting it with various attention layers, which enables the model to learn novel synthesis recipes more effectively. Moreover, during retrieval, we consider the thermodynamic relationship between target material and precursors, which is essential domain expertise in identifying the most probable precursor set among various options. Extensive experiments demonstrate the superiority of Retrieval-Retro in retrosynthesis planning, especially in discovering novel synthesis recipes, which is crucial for materials discovery.한국과학기술원 :산업및시스템공학과
조직병리학 이미지 분류에서 다중 인스턴스 학습의 예측을 교정하기 위한 불확실성 기반 데이터별 레이블 스무딩
No full text학위논문(석사) - 한국과학기술원 : 산업및시스템공학과, 2025.2,[iv, 31 p. :]Deep neural networks (DNNs) have transformed biomedical image analysis, particularly in histopathology with Whole Slide Images (WSIs) classification. However, training DNNs requires large annotated datasets, which is challenging due to the high heterogeneity and high resolution of WSIs. Multiple Instance Learning (MIL) has become a popular method for weakly supervised classification in this context, training with only slide-level labels. Despite the advancements, ensuring the reliability of model performance is crucial in safety-critical domains including healthcare. Deep learning models in real-world decision-making systems must accurately predict probability estimates to reflect the true likelihood of correctness, known as confidence calibration. This study introduces a novel calibration framework, UDLS, which uses data-wise label smoothing based on predictive uncertainty to improve the calibration of MIL frameworks. This approach involves augmenting WSIs with PatchFeatureDropout, computing predictive uncertainty estimates for original data, and applying these estimates to each sample for label smoothing during model training. Experimental results on benchmark histopathology datasets show noticeable improvements in both calibration and classification performance, highlighting UDLS's potential for enhancing the reliability of predictions from deep learning models in clinical settings.한국과학기술원 :산업및시스템공학과
Nested Logit Model에서의 군집 구조 학습: 병원 선택 경향 분석을 위한 데이터 기반 접근법
No full text학위논문(석사) - 한국과학기술원 : 산업및시스템공학과, 2025.2,[v, 52 p. :]The Korean healthcare delivery system is characterized by patients’ independent and unrestricted choice of medical institutions, which leads to issues such as the overconcentration of patients in Seoul’s tertiary hospitals. To effectively address this problem, understanding patients’ decision-making process is imperative. This research employs the Nested Logit Model (NLM) to derive a patient-perspective hospital classification structure from data, grouping hospitals that are in competitive relationships and are chosen by individuals with similar preferences and needs. Acknowledging the lack of existing methods for directly extracting optimal nest structures from data, we propose a novel two-stage algorithm combining a genetic algorithm (GA) and a silhouette-based evaluation. Our results confirm that patients’ hospital choice behavior is better explained at the group level rather than at the individual hospital level. Regional analyses of Chungcheong, Jeolla, and Gyeongsang patients reveal significant heterogeneity in hospital classification. The derived nest structures challenge the prevailing belief that Seoul’s “Big 5” hospitals are viewed as interchangeable, and suggest an alternative grouping that includes the two most visited hospitals in Seoul and a Gyeongggi-based hosptial. We also find that higher-income patients, younger patients and patients with more severe diseases are more likely to choose Seoul hospitals. These findings provide valuable insights that could help policy-makers in designing targeted strategies to mitigate the patient overconcentration problem.한국과학기술원 :산업및시스템공학과
Theoretical study on compensation for employees’ inventions -A focus on corporate legal precedents and standard model-
No full text학위논문(석사) - 한국과학기술원 : 지식재산대학원프로그램, 2025.2,[iv, 49 p. :]최근 국내 기업의 직무발명 보상금 청구 소송이 늘어나고 있다. 이러한 소모적인 직무발명 보상금 청구 분쟁을 미연에 방지하여 사용자인 기업이 직무발명 보상금에 대한 예측 가능성을 높이면서도, 종업원에게는 정당한 보상이 돌아가는 방향으로 제도 개선이 이루어져야 하겠다. 따라서 본 학위논문에서는 직무발명 보상과 관련된 법률 및 주요 판례 등을 검토하고, 이를 기초로 현행 ‘직무발명 보상 규정 표준모델’을 보완하여, 앞으로 직무발명 보상 규정을 도입할 기업에게, 구체적이고 세밀한 보상 기준을 제시할 수 있는 표준모델 개선안을 제안하고자 한다.한국과학기술원 :지식재산대학원프로그램
공정한 배치 오더 블록체인 시스템에 대한 프론트러닝 공격
No full text학위논문(석사) - 한국과학기술원 : 정보보호대학원, 2025.2,[iv, 36 p. :]In timing-sensitive blockchain applications, such as decentralized finance (DeFi), maintaining the fair ordering of transactions is more important than ever. This paper examines recent batch-order-fair systems (e.g., Themis [1] and SpeedyFair [2]) designed to provide fair ordering of transactions against frontrunning attacks and reveals their vulnerability to a new type of frontrunning attack called the Ambush attack. We demonstrate that these fair ordering systems can be compromised by a single malicious replica that simply submits a few transactions to the other replicas in a carefully crafted order. The Ambush attack exploits ambiguities in transaction ordering within Condorcet cycles, achieving frontrunning success rates up to near 100% against leading systems in practical settings (e.g., over 100 replicas handling thousands transactions per second) at a cost of approximately one dollar per attack while remaining undetectable. Our analysis provides detailed insights into the attack’s reliability and offers several key intuitions on how the attack strategies can be heavily optimized in practice. We discuss two countermeasures that can potentially mitigate the Ambush attack to near zero success rateshowever, these countermeasures come with non-trivial trade-offs in terms of performance and operational complexity.한국과학기술원 :정보보호대학원
카메라 기반 촉각 측정과 상대 운동 추적을 통한 외부 접촉 감지를 이용한 파지된 물체 형상 재구성
No full text학위논문(석사) - 한국과학기술원 : 기계공학과, 2025.2,[iv, 46 p. :]In contact-rich tasks, such as assembly or tool use, humans can utilize their sense of touch to estimate the shape of a grasped object from its interaction with the environment. However, robots must estimate the contact between the hand, object, and environment through a limited bandwidth tactile sensor, which limits their ability to utilize extrinsic contact to reconstruct the shape of an object. This research proposes a method to utilize vision-based tactile sensors that visually represent tactile information to classify the type of contact between the grasped object and the environment, estimate the contact location, and integrate them to reconstruct the object's shape. First, by analyzing the deformation of the vision-based tactile sensor, the contact type (point contact or line contact) is classified. Second, the contact detection problem is formulated as a constraint-based prediction framework, and the location of extrinsic contacts is estimated by solving the constraints for each type of contact. Third, the geometric properties and parameters of the bottom surface of the grasped object are estimated by integrating the estimated point and line information. Utilizing extrinsic contact information to reconstruct the object geometry improves the robot's ability to interact with the gripped object and its environment. The proposed method will improve the robot's manipulation strategy planning ability in contact-rich manipulation tasks.한국과학기술원 :기계공학과
8절점 솔리드-쉘 유한요소의 잠김 현상 개선
No full text학위논문(석사) - 한국과학기술원 : 기계공학과, 2025.2,[iv, 43 p. :]The objective of this study is to develop an 8-node solid-shell finite element to alleviate the Membrane locking phenomenon and improve numerical stability in the analysis of thin-walled structures. The proposed element incorporates two key concepts: the membrane assumed strain field derived from shell elements and the geometry-dependent Gauss integration. Additionally, a modified material law is introduced to enhance numerical stability in thin-structure analysis. Unlike previous solid-shell elements that introduced additional degrees of freedom to address locking issues, the proposed element minimizes the need for such additional degrees of freedom, thereby reducing computational cost and eliminating numerical instability caused by the extra degrees of freedom. Furthermore, the adoption of the assumed strain field and geometry-dependent Gauss integration ensures smoother solutions and better convergence, enabling the use of the proposed element in relatively coarse meshes. The new solid-shell element successfully passes three fundamental numerical tests: zero energy mode, isotropy, and patch tests. Moreover, the proposed element demonstrates its effectiveness and reliability in linear analyses through various numerical examples.한국과학기술원 :기계공학과
회전 원판 위 돌출 구조물들에 의한 액막 분열 현상
No full text학위논문(석사) - 한국과학기술원 : 기계공학과, 2025.2,[iii, 17 p. :]In industrial applications, solid surfaces with protruded structures are often employed to enhance heat transfer and atomization systems. Therefore, extensive research has been conducted on the interactions between liquid film contact lines and the protruded structures on solid surfaces. However, most of these studies have primarily focused on the flow modifications of the liquid film caused by the protruded structures, with less attention given to the influence of these structures on liquid film disintegration. Disintegration of liquid film greatly affects the performance of industry applications, such as enhanced heat transfer, reacting and atomization systems. Therefore, it is essential to investigate the breakup phenomena arising from the interactions between liquid film flow and protrusions on solid surfaces. In this study, the disintegration of liquid film is investigated when film flow generated by an impinging jet on a rotating disk interacts with protruded structures on a rotating disk. Four modes of disintegration were observed based on flow variables, such as flow rate and angular velocities. The mechanism of disintegration on the structures is identified by observing disintegration mode transitions, which is varied with centrifugal force and flow rate around the protruded structure. To quantitatively analyze the transition boundaries between modes, the relation between diameter of generated droplets and the Weber number was investigated. To find transition boundaries between the disintegration modes, the flow rate around the structure was defined and mode transition boundary model according to the mechanism was derived.한국과학기술원 :기계공학과