A5101772444- Dynamics and Control of Mechanical Systems
- Hydraulic and Pneumatic Systems
- Topic Modeling
- Natural Language Processing Techniques
- Adhesion, Friction, and Surface Interactions
- Gear and Bearing Dynamics Analysis
- Real-time simulation and control systems
- Tribology and Lubrication Engineering
- Semiconductor Quantum Structures and Devices
- Mechanical stress and fatigue analysis
- Quantum and electron transport phenomena
- Vehicle Dynamics and Control Systems
- Model Reduction and Neural Networks
- Engineering Applied Research
- Speech Recognition and Synthesis
- Mechanical Engineering and Vibrations Research
- Vibration and Dynamic Analysis
- Mechanical and Optical Resonators
- Computational Physics and Python Applications
- Robotic Mechanisms and Dynamics
- Fluid Dynamics Simulations and Interactions
- Data Quality and Management
- Contact Mechanics and Variational Inequalities
- Advanced Mathematical Identities
- Multimodal Machine Learning Applications
eBay (Ireland)
2010-2023
Chung-Ang University
2022
Kyung Hee University
2019
Bay Institute
2016
Seoul National University of Science and Technology
2014
Seoul National University
2009
Kyungpook National University
2007-2009
eBay (South Korea)
2005
In this work, we propose a robust modeling and analysis technique of the piston-lubrication system considering fluid–structure interaction. The proposed schemes are based on combining elastohydrodynamic multi-flexible body dynamics. particular, dynamics can offer highly precise numerical results regarding nonlinear deformation piston skirt cylinder bore, which lead to more accurate film thickness for gaps filled with lubricant relative velocity facing surfaces between block. These dynamic...
We reported the apatite-forming ability of 30CaO∙70SiO 2 scaffolds with 0~100 ppm Ag ions by soaking in simulated body fluid (SBF). This study was to evaluate effects concentrations gels on in-vitro biocompatibility osteoblasts (MC3T3). After seeding cells surface Ag-30CaO∙70SiO scaffold, cellular behaviors were evaluated an assay cell counting kit-8. Cytotoxicity scaffold samples employing extract solutions assays neutral red, MTT and BrdU. In addition, live & dead performed using a gel...
Coordinate reduction has been widely used for efficient simulation of flexible multibody dynamics. To achieve the bodies with reasonable accuracy, appropriate number dominant modes process must be selected. handle this issue, an iterative coordinate strategy is introduced. In iteration step, more are selected than ones in previous step. Among various methods, conventional frequency cut-off rule here considered. As a stop criterion, novel posteriori error estimator that can evaluate relative...
This study proposes a systematic extension of multiflexible-body dynamics (MFBD) formulation that is based on recursive for rigid body dynamics. It extended to include nonlinear plastic and hyperelastic material models the flexible bodies. The bodies in existing MFBD use finite element corotational elements. are coupled using method Lagrange multipliers. extensions add plasticity hyperelasticity outlined. A solid, brick-type shell adapted from literature with material, constant volume...
유연 다물체 동역학은 실제 시스템을 가능한 유사하게 수치화하여 해석할 수 있기 때문에 일반 동역학 연구에 대한 차세대 주제로 각광을 받고 있다. 이러한 동역학에 해석 기술은 리커다인이라는 상용 소프트웨어에 효과적으로 적용되어 있는데, 특히 강체와 유연체를 통합하여 하나의 솔버에서 해석을 할 있는 특징을 가지고 본 논문에서는 리커다인의 솔버의 기술들을 살펴보고자 한다. 기본적으로 해법은 동시회전기법을 사용하는 증분 유한요소 정식화를 기존의 순환공식을 이용한 정식화에 결합함으로써 구현되어 진다. 이 과정에서 유연체 사이의 조인트나 힘 요소 등의 효율적 처리를 위해 가상 바디와 조인트 개념이 사용된다.
Data annotation is an essential step for constructing new datasets. However, the conventional approach of data through crowdsourcing both time-consuming and expensive. In addition, complexity this process increases when dealing with low-resource languages owing to difference in language pool crowdworkers. To address these issues, study proposes autonomous method by utilizing large models, which have been recently demonstrated exhibit remarkable performance. Through our experiments, we...
Rule-based text data augmentation is widely used for NLP tasks due to its simplicity. However, this method can potentially damage the original meaning of text, ultimately hurting performance model. To overcome limitation, we propose a straightforward technique applying soft labels augmented data. We conducted experiments across seven different classification and empirically demonstrated effectiveness our proposed approach. have publicly opened source code reproducibility.
In the field of text data augmentation, rule-based methods are widely adopted for real-world applications owing to their cost-efficiency. However, conventional approaches suffer from possibility losing original semantics given text. We propose a novel augmentation strategy that avoids such phenomena through straightforward deletion adverbs, which play subsidiary role in sentence. Our comprehensive experiments demonstrate efficiency and effectiveness our proposed approach not just single...
The quality of the dataset is crucial for ensuring optimal performance and reliability downstream task models. However, datasets often contain noisy data inadvertently included during construction process. Numerous attempts have been made to correct this issue through human annotators. hiring managing annotators expensive time-consuming. As an alternative, recent studies are exploring use large language models (LLMs) annotation. In study, we present a case study that extends application...
Although pre-trained language models have exhibited great flexibility and versatility with prompt-based few-shot learning, they suffer from the extensive parameter size limited applicability for inference. Recent studies suggested that PLMs be used as dataset generators a tiny task-specific model trained to achieve efficient However, their various domains is because tend generate domain-specific datasets. In this work, we propose novel approach universal domain generalization generates...
Domain generalizability is a crucial aspect of deep learning model since it determines the capability to perform well on data from unseen domains. However, research domain models for vision-language tasks remains limited, primarily because lack required datasets. To address these challenges, we propose VolDoGer: Vision-Language Dataset Generalization, dedicated dataset designed generalization that addresses three tasks: image captioning, visual question answering, and entailment. We...
Large language models (LLMs) have become a dominant approach in natural processing, yet their internal knowledge structures remain largely unexplored. In this paper, we analyze the of LLMs using historical medal tallies from Olympic Games. We task with providing counts for each team and identifying which teams achieved specific rankings. Our results reveal that while state-of-the-art perform remarkably well reporting individual teams, they struggle significantly questions about This suggests...