GIST Scholar
Not a member yet
30271 research outputs found
Sort by
Frequency-enhanced network with self-supervised learning for anomaly detection of hydraulic piston pumps
No full textThis paper presents a novel fault diagnosis method to address the challenge posed by the variable operating conditions of hydraulic piston pumps. The proposed method incorporates three key features. First, a frequency-enhanced network (FENet) is developed in this research to consider characteristics from both time and frequency domains simultaneously. FENet overcomes the limitations of conventional convolution operations through its frequency-aware convolution architecture, which enhances feature-extraction capabilities in the frequency domain. Second, a new CutMix-based self-supervised learning approach is proposed to achieve robust generalization performance across varying operating conditions and equipment variations. Third, a novel health index that combines the Mahalanobis distance and the Fisher discriminant ratio is developed to enhance the training stability of high-dimensional latent vector representations during model training. To validate the proposed method, experiments are conducted on industrial hydraulic piston pumps, focusing on wear defects between the slipper shoe and the piston ball under different temperature and pressure conditions. The generalization performance of the proposed method is evaluated through cross-validation across multiple equipment units. Feature map analysis reveals that vibration intensity increases in lower-frequency regions associated with fault severity, while it decreases in a relatively higher-frequency range. These analytical results demonstrate that the proposed method can serve as an effective condition monitoring tool for hydraulic piston pumps in industrial applications. © 2025 Elsevier LtdFALSEsciescopu
Electrochemical defect control of bulky crystalline CuBi2O4 film and the band edge alignment for photoelectrochemical water reduction
No full textHetero-metal oxides have been used in photoelectrochemical (PEC) water splitting systems, but synthetic-originating defects and charge recombination process degrades PEC cell performance. Herein, we studied intrinsic physical properties of pure-phase copper bismuth oxide (CuBi2O4, CBO) photocathode through controlling defects and band edge alignment. Preparation of pure-phase CBO film with large grain sizes (average similar to 290 nm) enabled to investigate the correlation between CBO's crystal structure and charge carrier transport efficiency. The Cu1+-V-o point defects were regulated through electrochemical oxidation or thermal oxygenation under argon, air, and O-2 atmosphere. The thermal treatment in an O-2-saturated environment significantly reduced Cu1+-V-o defects, increasing charge carrier density, thereby reducing band gap, which eventually facilitated the charge transport. Moreover, electrochemical oxidation produced similar band structure to thermal oxygenation under O-2, demonstrating a high level of Cu1+ defect control could be achieved through electrochemical oxidation as well as thermal oxygenation, showing systematic adjustment of the CBO band edge. Additionally, the hole-transfer heterojunction at the CBO film's back side was engineered using copper oxide (CuO) thin film for interfacial band alignment. As a result, band edge-aligned FTO|CuO|CBO heterojunction exhibited a remarkably increased photocurrent density up to 2.63 mA/ cm(2) at 0.4 V vs. RHE in alkaline electrolyte.FALSEsciescopu
Insights from the Incheon Airport Case in South Korea: balancing public safety and individual rights with global scalability analysis
No full textThis article examines the use of facial recognition technology (FRT) in policing by government agencies, exploring the tensions between public safety and individual rights. While FRT enhances law enforcement capabilities in crime prevention and security, it also raises significant concerns regarding privacy invasion and potential misuse. Based on the Incheon Airport case in South Korea, this analysis investigates whether the public interest derived from policing can justify violations of regulations under personal data protection and administrative laws while proposing necessary safeguards. In pursuit of a comprehensive data governance framework for FRT, we are emphasizing the need to establish a more specific and operable data governance structure, strengthen the Privacy by Design principle, and balance security imperatives with individual rights as key alternatives. By shedding light on the Korean case within international discourse, this study underscores the importance of tailoring regulatory frameworks to accommodate local legal traditions and societal values. © The Author(s) 2025.TRUEssciahciscopu
3D LiDAR-Based 6D Pose Estimation for Object Trajectory Prediction in Autonomous Vehicle Applications Minjae Cho School of Mechanical Engineering Gwangju Institute of Science and Technology
No full textAn autonomous driving system is a technology that enables a vehicle to navigate itself to a given destination without human intervention. This technology is being ac- tively researched due to its promise to reduce tra!c accidents, improve transportation e!ciency, and save time and energy. In autonomous driving, vision sensors such as cameras and LiDAR detect surrounding objects and predict their trajectories. Camera sensors are currently the most commonly used, but they do not provide direct distance information, which limits their use for safe driving. LiDAR sensors, on the other hand, provide direct distance information through 3D point cloud data, enabling safer driving. To achieve high safety and e!ciency in autonomous driving systems, it is necessary to predict the trajectories of objects. Currently, LiDAR-based 3D object detection is widely used to predict the trajectory of objects. These techniques are mainly catego- rized into voxel-based, point-based, and fusion-based methods. Voxel-based methods convert sparse 3D point cloud data into regular 3D voxels or 2D grids and perform 3D sparse convolutional neural networks (CNNs), which are computationally e!cient but have the limitation of low location accuracy. Point-based methods focus on all raw point cloud data to preserve accurate location information but could be more compu- tationally e!cient because they focus on all points. Fusion-based methods combine the advantages of both, preserving computational e!ciency and location information. Conventional LiDAR-based 3D object detection techniques estimate an object’s class, size, and z-axis rotation, ignoring the x-axis and y-axis rotation information. However, 6D pose estimation is essential to accurately predict the trajectories of all objects, including two-wheelers and pedestrians. In this paper, I propose two meth- ods for estimating a 6D pose of objects. First, I propose constructing a synthetic 6D pose dataset in a simulation environment for training the network. Second, I propose Point-Voxel based-Pose estimation networks (PV-Pose) for 6D pose estimation on point clouds by applying transfer learning based on the PV-RCNN++ model. ©2025 Minjae Cho ALL RIGHTS RESERVEDDoctorAbstract i
List of Contents iii
List of Tables v
List of Figures vi
I. INTRODUCTION 1
1.1. Motivation 1
1.2. Contributions 4
1.3. Overview of Thesis 5
II. RELATED WORK 7
2.1. Development synthetic dataset 7
2.2. Pose estimation 8
2.2.1. Pose estimation from RGB-(D) data 8
2.2.2. Pose estimation from LiDAR data 8
2.3. 3D Object detection with Point clouds 9
2.3.1. Voxel-based method 9
2.3.2. Point-based method 9
2.3.3. Fusion point-based and voxel-based method 10
III. DEVELOPMENT OF A SYNTHETIC 6D POSE DATASET FOR
OBJECTS IN A SIMULATION ENVIRONMENT 11
3.1. Introduction 11
3.2. Data annotation using 3D object models 12
3.2.1. Initialization of 3D object models 13
3.2.2. Scene Generations and data annotations 15
3.3. Synthetic point cloud generation in a simulation environment 18
3.3.1. 3D LiDAR model generation in simulation environment 19
3.3.2. Synthetic point cloud data generation 21
3.4. Conclusion 23
– iii –
IV. 3D LiDAR-BASED 6D OBJECT POSE ESTIMATION MODEL 24
4.1. Introduction 24
4.2. Model 29
4.3. Training loss 31
4.4. Training procedure 33
4.5. Conclusion 34
V. EXPERIMENTS 35
5.1. Experimental setup 35
5.1.1. Datasets 35
5.1.2. Evaluation metrics 36
5.2. Experimental results 38
5.2.1. Synthetic 6D pose dataset 38
5.2.2. 6D pose estimation results 39
5.2.3. Comparison with 3D object detection method 44
5.3. Ablation study 46
5.4. Discussion 47
VI. CONCLUSION 51
VII. FUTURE WORK 53
References 54
Acknowledgements 65
– iv
Stimulated emission tomography for frequency-degenerate orbital angular momentum qubits
No full textStimulated parametric down-conversion(StimPDC) can not only effectively mimic quantum correlations of spontaneous parametric down-conversion(SPDC) but also offers amplification of photons in several orders. With this advantage, StimPDC has been exploited in tasks that require higher brightness, such as quantum sensing and quantum state tomography. However, most of the StimPDC research only considers frequency non-degenerate processes. Since many photonic quantum protocols, such as measurement device-independent quantum key distribution(MDI QKD), and entanglement distillation utilizes Hong-Ou-Mandel(HOM) interference, frequency-degenerate photons are preferred. To adress this, we study frequency-degenerate StimPDC and demonstrate stimulated emission tomography(SET) for frequency-degenerate orbital angular momentum qubits. The biggest obstacle in the frequency-degenerate process was noise photons emitted from the seed laser, however, with TCSPC we could separate StimPDC photons and background noise photons. Our results agree with the theoretical prediction of the spiral bandwidth of entangled biphotons
Structural Studies of Phospholipid Biosynthesis and tRNA Modification
No full text본 학위논문은 미생물 효소의 구조적 역학과 분자 간 상호작용을 심층적으로 탐구한다. 특히 단백질-지질과 단백질-tRNA 상호작용에 초점을 맞추어 효소의 기질 인식, 촉매 메커니즘, 그리고 생물학적 기능을 구조적 관점에서 분석한다.
제1장에서는 대장균(Escherichia coli)의 인지질 생합성에 중요한 포스파티딜세린 탈탄산효소(PSD)를 연구한다. 고해상도 X-선 결정학적 방법을 통해 PSD의 비결합 상태와 인지질과의 복합체 구조를 규명하였으며, 이를 통해 PSD의 고유한 막 결합 특성과 기질 상호작용, 자가절단 및 촉매 과정에 관여하는 핵심 아미노산 잔기를 밝혀냈다. 본 연구는 세균 막 생합성에 대한 이해를 심화시키는 데 기여한다.
제2장에서는 황색포도상구균(Staphylococcus aureus)의 전사 RNA(tRNA) 변형 효소인 MnmM의 구조적 동역학을 연구한다. tRNA와의 복합체 결정 구조 분석을 통해 MnmM이 tRNA를 인식하고 상호작용하는 메커니즘을 밝혀냈으며, 이는 그람 양성균에서의 효소 진화적 적응을 이해하는 데 중요한 통찰을 제공한다.
부록에서는 고초균(Bacillus subtilis)과 대장균의 II형 tRNA의 결정 구조를 최초로 제시한다. 구조 분석 결과, tRNA의 안티코돈 줄기-고리 구조(anticodon stem loop)가 표준적 구조를 유지하면서도 가변 루프에서 독특한 구조적 유연성을 보임을 발견하였다. 이러한 구조적 가변성은 II형 tRNA의 상호작용과 기능에 중요한 의미를 시사한다.
본 학위논문은 미생물의 막 생합성, tRNA 변형 경로, 그리고 tRNA의 구조적 동역학에 대한 심도 있는 이해를 제공함으로써 핵심적인 생물학적 효소 반응에 대한 중요한 통찰을 제공한다.|The complex structure of an enzyme bound to its substrate reveals the molecular basis of selective substrate recognition, discrimination from non-substrates, and catalytic mechanism, illustrating how enzymes reduce energy barriers to drive reactions. This dissertation examines key intermolecular interactions—specifically protein-lipid and protein-tRNA—that underpin these processes in bacterial systems.
Chapter I investigates the enzyme phosphatidylserine decarboxylase (PSD), critical for phosphatidylethanolamine (PE) synthesis in Escherichia coli. High-resolution X-ray crystallography reveals both apo-PSD and its complex with PE, mimicking the Schiff base intermediate. These structures elucidate unique membrane association and substrate interactions of PSD, highlighting essential residues involved in auto-cleavage and catalysis through site-directed mutagenesis. This comprehensive structural study enhances understanding of bacterial membrane biogenesis.
Chapter II presents the structural dynamics of Staphylococcus aureus MnmM, a novel tRNA modifying methyltransferase, through its crystal structures in three states: apo, ligand-bound, and tRNA-ligand-bound. These structures provide unique insights into how S. aureus MnmM recognizes and interacts with tRNA, revealing distinctive substrate recognition and catalytic mechanisms and underscoring the evolutionary adaptation of this enzyme in Gram-positive bacteria.
The Appendix presents the first crystal structures of free-state Type-II tRNAs, specifically tRNALeu from Bacillus subtilis and Escherichia coli. Structural analysis shows the anticodon stem-loop retains its canonical conformation while revealing unique conformational flexibility in the variable loop. This structural plasticity suggests significant implications for Type-II tRNA interactions and function.
Collectively, this dissertation contributes to a deeper understanding of bacterial membrane biogenesis, tRNA modification pathways, and the structural dynamics of tRNA, enhancing insights into fundamental biological processes.DoctorAbstract i
List of contents ․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․.․․․․․․․․․․․․․․․․․․․․․․․ iii
List of tables ․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․.․․․․․․․․․․․․․․․․․․․․ vi
List of figures ․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․.․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․․ vii
Chapter I １
Structural Insights into Phosphatidylethanolamine Formation in Bacterial Membrane Biogenesis
ㅤ
1.1. Abstract ２
1.2. Introduction ３
1.2.1. Functions of phospholipid
1.2.2. Bacterial phosphatidylethanolamine biosynthesis
1.2.3. Eukaryotic phosphatidylethanolamine biosynthesis
1.2.4. Phosphatidylserine decarboxylase in E. coli
1.2.5. Recent structural studies of PSD
1.2.6. Research goal
1.3. Materials and Methods ７
1.3.1. Materials
1.3.2. Cloning
1.3.3. Expression and purification
1.3.4. Ligand conjugation by reduction for structural characterization
1.3.5. Crystallization and structural determination
1.3.6. SEC-MALS analysis
1.3.7. Membrane association assay
1.3.8. LC-MS analysis of functional activity
1.3.9. Ligand conjugation by reduction of H144 mutants
1.3.10. MALDI-TOF for conjugation efficiency analysis
1.3.11. Auto-activation assay
1.3.12. Data availability
1.4. Results １２
1.4.1. Validation of activity
1.4.2. Overall features of apo structures
1.4.3. Identification of the membrane association domain
1.4.4. Phospholipid-bound structures
1.4.5. Recognition of phosphoserine moiety
1.4.6. Mutagenesis studies on auto-cleavage mechanism
1.5. Discussion ３４
1.5.1. Substrate specificity and phospholipid recognition
1.5.2. Membrane association domain
1.5.3. Mechanism of auto-cleavage
1.5.4. Therapeutic implications
ㅤ
Chapter II ３８
Structural Insights into 5-Aminomethyl-2-thiouridine Methyltransferase in Staphylococcus aureus
tRNA Modification
ㅤ
2.1. Abstract ３９
2.2. Introduction ４０
2.2.1. Overview of tRNA modifications
2.2.2 Wobble position modifications
2.2.3 Biosynthesis of mnm5(s2)U in Escherichia coli and Gram-positive bacteria
2.2.4 Discovery of MnmM as a novel methyltransferase in Bacillus subtilis
2.2.5 Research goal
2.3. Materials and Methods ４４
2.3.1. Biological resources
2.3.2. Cloning and protein expression
2.3.3. Bacterial bulk tRNA extraction
2.3.4. In vitro assay
2.3.5. Crystallization and structural determination
2.3.6. Data availability
2.4. Results ４７
2.4.1. In vitro methyl transfer activity of recombinant MnmM orthologs
2.4.2. X-ray crystal structures of S. aureus MnmM
2.4.3. Recognition of anticodon stem-loop at the tRNA binding surface of MnmM
2.5. Discussion ６４
2.5.1. Substrate specificity of S. aureus MnmM
2.5.2. A proposed catalytic mechanism of S. aureus MnmM
ㅤ
Appendix ６８
Structural Insights into the Conformational Flexibility of Type-II tRNALeu in the Free State
ㅤ
A.1. Abstract ６９
A.2. Introduction ７０
A.2.1. The structural and functional role of tRNA
A.2.2. Classification of tRNA types and the unique role of Type-II
A.2.3. Research goal
A.3. Materials and Methods ７２
A.3.1. tRNA preparation
A.3.2. Crystallization and structure determination
A.3.3. Data availability
A.4. Results ７４
A.4.1. Overall structures of free tRNALeu
A.4.2. Anticodon stem loop
A.4.3. Variable arm
A.4.4. Comparison with Type-I tRNA
A.4.5. Structural comparison of free BstRNALeu
UAA with protein-complexed tRNALeu
A.4.6. Structural comparison of free EctRNALeu
UAG with complexed tRNALeu
A.5. Discussion ９９
ㅤ
References １０１
Curriculum Vitae １１１
Acknowledgements １１３
국문 초록 １１
Explainable Deep Learning for Multi-Cohort Alzheimer's Disease Classification using MRI: Insights into Pathological Brain Regions
No full textBACKGROUND: Magnetic Resonance Imaging (MRI) plays a crucial role in the early diagnosis and monitoring of Alzheimer's disease, offering detailed insights into structural and functional brain changes. However, variations in MRI data across national cohorts present significant challenges, complicating consistent and reliable diagnosis. Explainable AI with attention-map visualization techniques, can overcome these challenges by enhancing diagnostic accuracy and enabling interpretable insights into Alzheimer's disease progression through MRI analysis. In this study, we propose a deep-learning approach to visualize key pathological brain regions associated with AD versus Cognitive Normal (CN) classification across multi-cohort datasets. METHOD: The data utilized in this study were collected from two cohorts: the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (American cohort) and the Gwangju Alzheimer's and Related Dementia (GARD) database (Korean cohort). To investigate critical pathological brain regions associated with Alzheimer's disease, we trained a classification model to distinguish Alzheimer's disease (AD) from cognitively normal (CN) individuals. The model employed a ResNet-50 backbone, integrated with an attention mechanism to enhance spatial feature extraction. To ensure the robustness of the approach, attention maps were visualized in the sagittal, axial, and coronal planes of MRI scans. RESULT: Our proposed method demonstrates state-of-the-art performance in distinguishing AD from CN individuals, achieving 98.18% accuracy, 96.73% specificity, and 98.97% sensitivity. Attention map analysis indicates that the model primarily targets the hippocampus and temporal lobe, consistent with prior research on key pathological regions associated with Alzheimer's disease. Additionally, the model exhibits consistent performance on the GARD cohort, underscoring its generalizability and reliability. CONCLUSION: In summary, our method achieves outstanding performance in AD vs. CN classification, accurately identifying key pathological regions such as the hippocampus and temporal lobe. The consistent results across diverse cohorts demonstrate its robustness and potential for broader clinical implementation. © 2025 The Alzheimer's Association. Alzheimer's & Dementia published by Wiley Periodicals LLC on behalf of Alzheimer's Association.TRUEsciescopu
ReD shoes: actuated footwear for multisensory redirected walking in virtual reality
No full textA major challenge in virtual reality (VR) is enabling users to navigate expansive virtual environments within confined real-world spaces. Although redirected walking (RDW) addresses this by manipulating users’ walking paths, it is hindered by visual-vestibular inconsistencies, leading to reduced immersion, discomfort, and constrained detection thresholds (DTs). We introduce ReD Shoes, redirecting shoes for VR locomotion with dynamically adjustable inclination designed to enhance multisensory integration by providing haptic feedback directly to the feet. This feedback mitigates visual-vestibular inconsistencies by enhancing proprioceptive cues and aligning tactile sensations with visual stimuli, thereby improving self-motion perception, presence, and gait stability. We conducted two experiments with 30 participants: 10 in a pilot study to optimize inclination parameters and 20 in a main study to evaluate the DT, gait stability, and user experience under No Inclination, Low Inclination (1 cm), and High Inclination (1.5 cm) conditions. Low Inclination considerably expanded the DT (43.64%), enhanced the user’s feel of presence, and maintained walking stability through effective multisensory alignment. In comparison, High Inclination showed a smaller DT expansion (33.86%), increased discomfort, reduced stability, and aggravated oculomotor symptoms due to intensified sensory conflicts. This study demonstrates the use of ReD Shoes in addressing key RDW limitations by enhancing DT, stability, and user presence through active haptic feedback. Our findings offer insights for developing advanced VR mobility solutions, supporting adaptive haptic footwear that improves RDW functionality and extends immersive VR applications in gaming, training, and rehabilitation.TRUEsciescopu
Development of a real-time link-based predictive energy management strategy for extending FCEV lifespan using an experiment-driven degradation model
No full textFuel cell electric vehicles (FCEVs) face durability challenges primarily due to cell degradation influenced by power variations and operational ranges. This issue can be mitigated through an energy management strategy (EMS), with many durability-focused studies employing predictive EMS (P-EMS) for high performance. However, existing strategies often rely on highly uncertain future vehicle trajectories, such as velocity or power demand, leading to a shortened horizon length and significant loss of optimality. This study proposes a novel link-based, durability-focused P-EMS optimized on a per-link basis, achieving near-optimal performance. The key innovation lies in reformulating the problem from trajectory optimization to parameter optimization, expressed as a quadratic programming (QP) problem, which enables real-time implementation. The degradation model consists of dynamic and quasi-static operations, where the quasi-static model is developed based on experimental data. A multi-objective optimal control problem is then formulated, revealing a Pareto optimal relationship between degradation and system efficiency through a dynamic programming (DP) algorithm that ensures global optimality. Building on insights from DP results, the proposed approach analytically reformulates the problem, requiring easily predictable driving parameters such as travel time and energy demand that represent link conditions. The simulation results reveal that, when prioritizing cell degradation protection, the proposed method achieves high performance comparable to DP, with a minimal loss of optimality (1.5 % in fuel economy and 6.7 % in fuel cell degradation) while showing an impressive average computational time of merely 2.5 ms.FALSEsciescopu
Investigation of Surface Oxidation States in Fe3GeTe2 via Ultrafast and Nonlinear Optical Technique
No full textFe3GeTe2, with a high Curie temperature of about 200 K, is attracting attention as a next- generation spintronic material due to its interesting phenomena. However, the effect of surface oxidation on the electronic structure and magnetic properties of Fe3GeTe2 remains elusive. In this thesis, we investigated the influence of surface oxidation on the electronic, lattice, and magnetic dynamics in Fe3GeTe2 using various optical techniques. The oxidation was induced by annealing in ambient conditions. We found that the transient reflectivity and photo-induced demagnetization dynamics are significantly modified by surface oxidation. Additionally, the second harmonic generation (SHG) intensity has increased with oxidation. These results highlight the complex interaction among charge, spin, and lattice under surface oxidation. Our results suggest that oxidation serves as a crucial tuning parameter for controlling two-dimensional ferromagnetic materials. It offers insights into the design of spintronic devices based on Dzyaloshinskii-Moriya interaction or spin-orbit torque effects.MasterList of Contents
Abstract i
List of Contents ii
List of Figures vi
List of Tables ⅴ
I. Introduction 1
1. 1. Van der Waals ferromagnet Fe3GeTe2 1
1. 2. Ultrafast carrier dynamics in Fe3GeTe2 2
1. 2. 1. Transient reflectivity and Phonon oscillation 2
1. 2. 2. Ultrafast spin dynamics in Fe3GeTe2 3
1. 3. Fe3GeTe2 heterostructure 3
Ⅱ. Experimental methods 5
2. 1. Sample preparation 5
2. 1. 1. Mechanical exfoliation 5
2. 1. 2. Oxidation method 6
2. 2. Optical techniques 6
2. 2. 1. Optical Pump Optical Probe (OPOP) 7
2. 2. 2. Second Harmonic Generation (SHG) 8
2. 2. 3. Time-resolved Magneto-optic Kerr effect (TR-MOKE) 9
Ⅲ. Results and discussion 11
3. 1. Sample preparation and Structural characterization 11
3. 2. Ultrafast dynamics in Fe3GeTe2 14
3. 3. Oxidation-induced symmetry change 17
3. 4. Magnetization dynamics and oxidation effects 21
Summary 25
References 2