187037 research outputs found
Sort by
Reliability-based design optimization of motor housing to reduce radiated noise under uncertain dynamic eccentricity
No full text학위논문(석사) - 한국과학기술원 : 미래자동차학제전공, 2025.2,[v, 57 p. :]모터 시스템의 소음, 진동을 해석하기 위해서는 전자기해석과 기계진동해석을 통한 다물리해석이 필요하다. 하지만 회전자에 편심이 있는 경우 유한요소해석(FEA)을 통한 전자기해석은 비대칭성으로 인해 전체모델을 사용하며 많은 계산비용이 발생한다. 본 연구는 전체모델을 사용한 정확한 FEA데이터를 고충실도 데이터로, 모터 초기설계 단계에서 이용되는 MEC와 계산비용이 상대적으로 적게 드는 축소모델 FEA를 저충실도 데이터로 동시에 이용하는 다중 충실도 대리모델을 제안한다. 제안된 다중 충실도 대리모델을 각각 다른 저충실도 데이터를 사용한 대리모델과 비교하여 검증하였고, 그 결과 편심을 표현할 수 있는 높은 정확도와 함께 전체모델 FEA를 적게 수행해도 되는 효율성을 입증하였다. 이 대리모델을 통해 생성된 편심에 의한 힘을 진동해석에 사용하여 소음 저감 최적화를 위한 대리모델을 생성한다. 진동해석 대리모델을 통해 신뢰성기반 최적설계(RBDO)를 수행하였고, 그 결과 무게를 16% 줄이며 제약조건인 10% 감소한 ERP 최대 값을 만족시키는 하우징의 최적설계 안을 도출하는 것에 성공했다.한국과학기술원 :미래자동차학제전공
Financing payouts: an empirical study in the Korean stock market
No full text학위논문(석사) - 한국과학기술원 : 금융공학프로그램, 2025.2,[iii, 51 p. :]기업이 내부자금이 아닌 외부의 재원을 조달하여 배당이나 자사주매입을 실시하는 경우는 자금의 유출입 방향이 서로 상이한 지급(Payout)과 외부자본 조달(Financing)이 동시에 발생한다는 점에서 상당히 독특하지만 실제로는 드물지 않게 관찰되는 바, 그러한 활동의 양상과 동인 등에 대해 구체적으로 살펴볼 필요가 있다. 본 연구는 Farre-Mensa et al.(2024)의 방법론에 따라 외부자본 조달 지급활동(Financed Payouts)을 기업이 동일한 회계연도 중 외부자본 조달과 지급활동을 실시한 경우로 정의하고 한국의 상장기업들 사이에서 나타나는 외부자본 조달 지급활동에 대해 계량화하였으며, 세금절감, 현금 및 레버리지 관리가 외부자본 조달과 지급활동의 동기가 된다는 점과 지급의 재원에 따라 투자자의 반응 또한 달라진다는 점을 확인하였다.한국과학기술원 :금융공학프로그램
스마트 팩토리가 기업 재고자산 관리에 미치는 영향 연구
No full text학위논문(석사) - 한국과학기술원 : 경영공학부, 2025.2,[ii, 40 p. :]This study investigates the impact of smart factory adoption on firms' inventory management efficiency. Using manually collected data from the World Economic Forum's 153 Lighthouse designations, I find that the acquisition of a Lighthouse designation significantly reduces inventory levels. The effect is particularly strong among firms with large asset sizes and those receiving their first-time Lighthouse designation. Manufacturing-focused Lighthouses achieve greater inventory reductions compared to End-to-End Lighthouses, underscoring the production-centered value of smart factory strategies. Further, while Lighthouse designation itself does not directly affect future firm performance, it has a significant indirect influence on future sales growth through improvements in inventory management, such as reduced inventory levels, enhanced inventory leanness, and increased inventory turnover ratio. Overall, the findings support the strategic importance of adopting smart technologies to strengthen resilience, sustainability, and competitiveness in modern supply chains.한국과학기술원 :경영공학부
관성모멘트 불확실성이 있는 유연 우주비행체의 진동 억제 및 적응형 성능 보장 자세 제어
No full text학위논문(석사) - 한국과학기술원 : 항공우주공학과, 2025.2,[iv, 42 p. :]In this paper, we propose an adaptive prescribed performance attitude control method for flexible spacecraft using modal observer to strictly adhere to mission requirements defined in advance. Flexible vibrations from lightweight appendages and uncertain moment of inertia tensor lead to performance degradation. To address these challenges, we employ prescribed performance control to meet strict mission requirements. In addition, uncertain parameter is estimated by an adaptive control scheme. A modal observer is employed to estimate the modal states, essential for both an adaptive law and control design effective in vibration suppression. The proposed approach is validated through simulations to showcase the effectiveness in mitigating flexible effects and estimating the uncertainty as well as achieving missions compared to conventional PPCs and PD control. Furthermore, a finite element method (FEM) model is used for the simulation to reflect the unmodeled dynamics and reliability.한국과학기술원 :항공우주공학과
탄성중합체의 손상 및 파괴에 대한 실험 및 연속체 모델링
No full text학위논문(석사) - 한국과학기술원 : 항공우주공학과, 2025.2,[iv, 48 p. :]Elastomeric materials display a complicated set of stretchability and fracture properties that strongly depend on the flaw size, temperature, and deformation rate, which has long been of interest to engineers and material scientists. In this thesis, we explore fracture in elastomers which depends on the flaw size and the loading rate. We first combine experiments and numerical simulations for a comprehensive understanding of the nonlocal, size-dependent features of fracture in elastomers. We quantitatively describe the size-dependent fracture behavior using a nonlocal continuum model. The key ingredient of the nonlocal model is the use of an intrinsic length scale associated with a finite fracture process zone, which is inferred from experiments. Applications to a wide range of synthetic elastomers that exhibit moderate (~ 200 %) to extreme stretchability (~ 1200%) are presented, which is also used to demonstrate the applicability of our approach in elastomeric specimens with complex geometries. Next, we study the rate-dependent deformation and fracture behaviors in elastomers. Significant efforts have been made to deal with the rate-dependent fracture using the phase-field model, however, there are limitations to account for the fracture of elastomeric networks. We aim to describe the rate-dependent fracture using the internal energy-driven fracture criterion which is used in the rate-independent fracture in elastomers. To this end, we design a new constitutive model for viscoelasticity based on the bond stretch. We argue that the rate-independent and -dependent fracture in elastomeric materials can be effectively accounted for by utilizing the internal energy-driven fracture criterion.한국과학기술원 :항공우주공학과
곡면 오리가미 적용 코일러블 붐의 구조성능 개선
No full text학위논문(석사) - 한국과학기술원 : 항공우주공학과, 2025.2,[iv, 56 p. :]Deployable structures are used in space applications when large structures are required as the fairing dimensions of launch vehicles are limited. One type of deployable structures are coilable tubular booms like STEM and CTM, which are simple in their design and have excellent packaging ratio compared to other deployable boom types. However, they also are usually less stiff and have less strength, which limits their applicability. Thin-walled booms under a bending load are prone to ovaling and overlapping of the cross-section. These effects reduce the second moment of area and make them more susceptible to buckling. Additionally, in the transition zone, where the boom’s cross-section changes from flat to the deployed cross-section, the boom is weakened. The transition zone is typically reinforced by a support structure that mitigates outward deformation. However, the support structure interacts with the thin shell of the boom under load, which can further alter the cross-section. To mitigate these effects, this thesis investigates the use of curved origami diaphragms to hinder inward deformation and isolate the transition zone, thereby stiffening and strengthening coilable tubular booms. The design of the curved origami diaphragm-reinforced booms is presented. A prototype was designed and manufactured to demonstrate the feasibility of deployment of this new design. Experiments were conducted to characterize the structural properties of booms using diaphragms. A simulation model was set up to further investigate the structural properties.
The experiment and simulation data shows that the diaphragms can significantly increase the strength and stiffness during loading for a bending load case. It was shown that diaphragms can be used to efficiently increase the stiffness with minimal increase of coiled volume.
The use of curved origami diaphragm could, therefore, increase the range of applications in which coilable tubular booms can be used.한국과학기술원 :항공우주공학과
얼굴 안티스푸핑에서 도메인 일반화를 위한 디퓨전 기반 텍스트-이미지 합성
No full text학위논문(석사) - 한국과학기술원 : 김재철AI대학원, 2025.2,[iv, 27 p. :]This paper addresses the challenge of developing robust Face Anti-Spoofing (FAS) models for face recognition systems. Traditional FAS protocols are limited by a lack of diversity in subject identities and environmental conditions, restricting generalization to real-world scenarios. Recent advancements in spoof image synthesis have mitigated data scarcity but still fail to capture the full range of facial attributes and environmental variability needed for effective domain generalization. To address this, we propose a novel framework capable of generating diverse, realistic facial images with text-guided control. We fine-tune Stable Diffusion to extract real facial features and specifically train LoRA layers to capture detailed spoof patterns. Additionally, the text-guided control of attributes helps overcome the lack of diversity seen in previous methods. Extensive experiments demonstrate that our text-to-image-based synthetic data generation significantly enhances the robustness of FAS models, establishing a new benchmark for domain-independent and reliable anti-spoofing systems.한국과학기술원 :김재철AI대학원
평생 학습 언어 모델에서 동적으로 변화하는 지식 평가에 관한 연구
No full text학위논문(석사) - 한국과학기술원 : 김재철AI대학원, 2025.2,[iv, 25 p. :]The dynamic nature of knowledge in an ever-changing world presents challenges for language models trained on static datathe model in the real world often requires not only acquiring new knowledge but also overwriting outdated information into updated ones. To study the ability of language models for these time-dependent dynamics in human language, we introduce a novel task, EvolvingQA, a temporally evolving question-answering benchmark designed for training and evaluating LMs on an evolving Wikipedia database. The construction of EvolvingQA is automated with our pipeline using large language models. We uncover that existing continual learning baselines suffer from updating and removing outdated knowledge. Our analysis suggests that models fail to rectify knowledge due to small weight gradients. In addition, we elucidate that language models particularly struggle to reflect the change of numerical or temporal information. Our work aims to model the dynamic nature of real-world information, suggesting faithful evaluations of the evolution-adaptability of language models.한국과학기술원 :김재철AI대학원
경로 적분 제어를 통한 추론에서 생성 모델 가이던스
No full text학위논문(석사) - 한국과학기술원 : 김재철AI대학원, 2025.2,[iv, 40 p. :]Despite the remarkable progress in generative models, producing outputs that align with user expectations remains a significant challenge. In the case of text-to-video generative models, complex temporal dependencies hinder the generation of videos aligned with the given text. Similarly, large language models, despite their ability to understand and generate text, often fail to accurately solve mathematical problems requiring complex logical reasoning. To address these issues, this paper proposes a method to control model generation using path integral control derived from stochastic optimization. This approach is particularly suitable for large-scale models such as video generation models and large language models, as it eliminates the need for additional training or internal gradient computation. We validate the path integral control approach on two types of generative models. First, we demonstrate that the proposed method can guide the video generation trajectory of a diffusion model to produce videos that are spatially and temporally aligned with the given text. Second, we show that for large language models, selecting optimal inference steps among multiple reasoning trajectories can significantly improve the accuracy of solving complex mathematical problems.한국과학기술원 :김재철AI대학원
A Wireless and Reusable OLED Patch for Accelerated In Vivo Skin Wound Healing
No full textA fully wireless and reusable red organic light-emitting diode (OLED) patch was developed as a noninvasive platform for skin wound phototherapy. The device integrates a USB-C rechargeable battery, a flexible printed circuit board, and a red OLED emitter into a soft, skin-conformal form factor. The OLED delivers uniform 645 nm light with minimal heat (similar to 21 degrees C) and stable radiance (21.3 W/sr/m2) over 500 h. In a murine full-thickness wound model, a single session of irradiation (2.6-3.9 J/cm2) significantly accelerated wound closure, increased collagen synthesis, and enhanced epidermal differentiation. Immunohistochemical analysis further confirmed reduced inflammation and improved tissue maturation. Unlike previous OLED-based systems that required rigid encapsulation or wired power, this patch enables convenient, home-based light therapy without clinical supervision. The platform demonstrates strong potential as an electroceutical device for personalized dermatologic care and regenerative medicine.