• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2005.11a
• /
• pp.227-230
• /
• 2005

This paper describes the problems in measuring and evaluating the aviation noises and the differences on the noise characteristics between fighter and civil aircraft. From a standpoint of the duration time for aviation noise and the dynamic characteristics related with the sampling time, the measured values for landing and taking-off noises from the fighter are presented and explained in comparison with those from the civil aircraft. And also the aviation noises by the fighter and civil aircraft are evaluated by equivalent sound level and WECPNL, respectively. From the obtained results, it is recommended that the duration time and dynamic characteristics be deeply considered in evaluating and measuring the aviation noises, since they have a great influence on the final assessment results. As the number of flight is not sufficient and the noise source is strong impulse type, moreover, the aviation noise should be assessed by WECPNL.

• Proceedings of the Korean Information Science Society Conference
• /
• 2010.06c
• /
• pp.299-304
• /
• 2010

본 논문에서는 온라인 학습을 이용한 한국어 의존구문분석 방법을 제안한다. CoNLL-X에서 1위를 차지한 그래프 기반 의존구문분석 방법을 한국어에 맞게 변형하고, 한국어의 교착어적 특성을 고려해 한국어에 적합한 자질 집합을 제시하였다. 특히 의존트리의 에지(edge)를 단어와 단어간의 의존관계가 아닌 부분트리(partial tree)와 부분트리의 의존관계로 바라보기 위해 부분트리가 공유하고 있는 기능어 정보를 추가 자질로 사용하였다. 또한 한국어의 지배소 후위(head-final) 언어 특성과 투사성(projectivity)을 이용하여 Eisner(1996) 알고리즘을 사용하지 않고도 O($n^3$)의 CYK알고리즘을 사용할 수 있었고, 이를 이용해 최적의 전역해(global optimum)를 찾을 수 있었다. 각 자질을 위한 최적의 가중치 벡터는 온라인 학습방법 중 하나인 Collins(2002)의 averaged perceptron 알고리즘을 사용함으로써 빠르게 모델을 학습할 수 있었다. 제안 모델을 국어정보베이스(KIBS) 말뭉치에 적용한 결과 어절 단위 정확률 88.42%의 높은 성능을 얻을 수 있었다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2018.05a
• /
• pp.202-204
• /
• 2018

2018년 현재 한국과학기술정보연구원(KISTI)은 계산과학공학 플랫폼 기반으로 온라인 시뮬레이션이 가능한 EDISON(EDucation-research Integration through Simulation On the Net) 서비스를 제공하고 있다. 해당 플랫폼 서비스는 7개 전문분야(전산의학, 전산설계, 전산열유체, 구조동역학, 계산화학, 나노물리, 도시환경)의 학생 및 연구자들이 학습 및 시뮬레이션을 실행하고 결과를 활용할 수 있도록 지원하고 있다. 하지만 EDISON 플랫폼은 시뮬레이션 서비스에 초점을 맞춰 개발과 서비스를 제공했기 때문에 국내외 MOOC(Massive Open Online Course) 서비스들에 비해 아직 교육 콘텐츠가 부족하다. 그리하여 본 논문에서는 계산과학공학 플랫폼에 교육 콘텐츠를 체계적으로 제공하기 위한 커리큘럼 서비스를 설계 및 제안하였다.

• Journal of Radiation Industry
• /
• v.4 no.4
• /
• pp.365-372
• /
• 2010

Irradiation has been recognized as a highly suitable tool to aid in the formation of hydrogel. The radiation process has various advantage, such as easy process control and the lack of necessity for initiators and crosslinker. In this study, the hydrogel containing the citrus fermentation gel for the wound healing were successfully synthesized. The strength of hydrogel was increased as a function of a increasing the concentration of citrus and the irradiation doses. In addition, this hydrogel have been evaluated by the cytotoxicity and animal experiment.

• Textile Coloration and Finishing
• /
• v.34 no.4
• /
• pp.250-260
• /
• 2022

In this research, a high-elasticity acrylic emulsion binder with core-shell polymerization and self-crosslinking system is mixed with a flame-retardant water-dispersed polyurethane (PUD) binder. In addition, finite element analysis was conducted through virtual engineering software ANSYS by applying three representative nonlinear material models. The most suitable nonlinear material model was selected after the relative comparison between the actual experimental values and the predicted values of the properties derived from simulations. The selected nonlinear material model is intended to be used as a nonlinear material model for computational simulation analysis that simulates the experimental environment of the vibration test (ASTM E1399) and the actual fire safety test (ASTM E1966). When the mass fraction of thermally expandable graphite was 0.7%, the thermal and physical properties were the best. Among the nonlinear material models, the simulation result of the Ogden model showed the closest value to the actual result.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.3
• /
• pp.125-132
• /
• 2023

Roundabouts have strengths in traffic flow and safety but can present difficulties for inexperienced drivers. Demand to acquire and analyze drone images has increased to enhance a traffic environment allowing drivers to deal with roundabouts easily. In this paper, we propose a roundabout traffic analysis system that detects, tracks, and analyzes vehicles using a deep learning-based object detection model (YOLOv7) in drone images. About 3600 images for object detection model learning and testing were extracted and labeled from 1 hour of drone video. Through training diverse conditions and evaluating the performance of object detection models, we achieved an average precision (AP) of up to 97.2%. In addition, we utilized SORT (Simple Online and Realtime Tracking) and OC-SORT (Observation-Centric SORT), a real-time object tracking algorithm, which resulted in an average MOTA (Multiple Object Tracking Accuracy) of up to 89.2%. By implementing a method for measuring roundabout entry speed, we achieved an accuracy of 94.5%.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.05a
• /
• pp.494-496
• /
• 2021

최근 경량 딥러닝 기반 키워드 음성인식은 가전, 완구, 키오스크 등 다양한 응용에 음성 인터페이스를 쉽게 적용할 수 있는 기술로서 주목받고 있다. 키워드 음성인식은 일부 키워드만 인식 가능한 음성인식 기술로서 저성능 디바이스에서 활용 가능한 장점이 있다. 그러나 응용에 따라 필요한 키워드에 대하여 다시 음성데이터를 수집해야하고 이를 학습하여 모델을 새로 준비해야하는 단점이 있다. 따라서 본 연구에서는 음성데이터 수집 없이 음성합성을 통해 생성한 음성으로만 키워드 음성인식 모델을 학습하는 음성합성 기반 자동 학습 기법을 제안하였다. 생성한 음성데이터를 활용하고자하는 시도가 활발히 이루어지고 있으나, 기존 연구에서는 정확도를 유지하기 위하여 수집한 실제 음성데이터가 필요한 한계가 있다. 제안한 자동 학습 기법은 생성한 음성데이터에 대해 복합 데이터 증대 기법을 적용하여 실제 음성데이터 없이 키워드 음성인식의 정확도를 높였다. 제안한 기법에 대하여 상용 음성합성 서비스를 기반으로 수집한 한국어 키워드 데이터세트를 활용하여 성능평가를 진행하였다. 20개 한국어 키워드에 대해 실험한 결과, 제안한 기법을 적용하여 학습시킨 키워드 음성인식 모델의 정확도는 86.44%임을 확인하였다.

• Journal of the Society of Naval Architects of Korea
• /
• v.60 no.5
• /
• pp.332-340
• /
• 2023

The present study established hull form optimization for small catamaran based on variations of knuckle lines. Four knuckle lines below the free surface were employed as design variables. Knuckle lines were independently transformed within remaining the main dimensions of the existing hull. For the hull form optimization, the SHERPA algorithm of HEEDS was utilized. Computational fluid dynamics was employed to estimate the resistance performance. The optimal hull showed the improvement of resistance performance of 9.3% than that of existing hull. The improvement of wave and pressure distributions for optimal hull was confirmed. Throughout the present study, it is expected that established optimization method can be applied for various small vessels such as fishing and leisure boats.

• Journal of the Korean Vacuum Society
• /
• v.17 no.4
• /
• pp.359-364
• /
• 2008

We have investigated optical and structural properties of $Al_{0.3}Ga_{0.7}N$/GaN and $Al_{0.3}Ga_{0.7}N/GaN/Al_{0.15}Ga_{0.85}N/GaN$ heterostructures (HSs) grown by metal-organic chemical vapor deposition, by means of Hall measurement, high-resolution X-ray diffraction, and temperature- and excitation power-dependent photoluminescence (PL) spectroscopy. A strong GaN band edge emission and its longitudinal optical phonon replicas were observed for all the samples. At 10 K, a 2DEG-related PL peak located at ${\sim}\;3.445\;eV$ was observed for $Al_{0.3}Ga_{0.7}N$/GaN HS, while two 2DEG peaks at ${\sim}\;3.42$ and ${\sim}\;3.445\;eV$ were observed for $Al_{0.3}Ga_{0.7}N/GaN/Al_{0.15}Ga_{0.85}N/GaN$ HS due to the additional $Al_{0.15}Ga_{0.85}N$ layers. Moreover, the emission intensity of the 2DEG peak was higher in $Al_{0.3}Ga_{0.7}N/GaN/Al_{0.15}Ga_{0.85}N/GaN$ HS than in $Al_{0.3}Ga_{0.7}N$/GaN HS probably due to an effective confinement of the photo-excited holes by the additional $Al_{0.15}Ga_{0.85}N$ layers. The 2DEG-related emission intensity decreased with increasing temperature and disappeared at temperatures above 150 K. To investigate the origin of the new 2DEG peaks, the energy-band structure for multiple AlGaN/GaN HSs were simulated and compared with the experimental data. As a result, the observed high- and low-energy peaks of 2DEG can be attributed to the spatially-separated 2DEG emissions formed at different AlGaN/GaN heterointerfaces.

• The Korean Journal of Rheology
• /
• v.11 no.1
• /
• pp.34-43
• /
• 1999

Flow-induced voids during resin impregnation and poor fiber wetting have known to be highly detrimental to the performance of composite parts manufactured by resin transfer molding(RTM) process. In this study, in order to overcome these serious problems encountered in RTM, the effects of surface modification by using silane coupling agent as a surface modifier on the flow characteristics, the wetting between resin and fiber, and void content were investigated. For the experiments of microscopic flow visualization and curing in a beam mold, glass fiber mats having plain weaving structure and epoxy resin were used. Modifying the fiber surface was found to result in a significant decrease of dynamic contact angle between resin and fiber and increase of wicking rate. Therefore, it was confirmed that the surface modification employed in this study could improve the wettability of reinforcing fibers as well as micro flow behavior. In addition, It was revealed that high temperature and low penetration rate of the resin are more favorable processing conditions to reduce the dynamic contact angle. However, surface modified fiber mat was found to have lower permeability than the unmodified one, which may be explained in terms of the decrease of contact time between resin and fiber owing to improvement of wetting. It was also exhibited that surface modification had a significant influence on void formation in RTM process, resulting in a decrease of overall void content due to the improvement of wetting in cured composite parts.