웹기반 콘크리트 구조물 균열진단 시스템 개발

하주형,조윤구,이석홍,Ha. Ju-Hyung,Cho. Yun-Gu,Lee. Suck-Hong 한국방재학회 2007 한국방재학회 학술발표대회논문집 Vol.2007 No.1

In recent years, the concrete cracks in most construction field have been widely presented and developed. Also it is well known the repair cost of the concrete cracks have been rapidly increased in Korea. So the concrete crack expert system was developed to minimize the repair cost and to support the field engineer. The feature of this web based system for diagnosing the causes of concrete cracks is comprised of comfortable user application.

隋末 儒學者 王通의 思想과 胡漢體制 - 王通의 正統論이 지니는 政治史的 意味를 中心으로

河주亨 서울대학교 동양사학과 2011 서울大 東洋史學科 論集 Vol.35 No.-

2007 국제도로교통박람회 참관기

하주형,Ha, Ju-Hyeong 한국도로교통협회 2007 도로교통 Vol.108 No.-

Multi-task prediction for clinical decision support rule to take head CT

하주형,정광률,차원철 대한응급의학회 2020 대한응급의학회 학술대회초록집 Vol.2020 No.2

Introduction Traumatic bring injury (TBI) is one of the most major medical and socioeconomic problems in the emergency department (ED). For an early and accurate diagnosis of TBI, neuroimaging such as computed tomography (CT) is used as a reference standard. However, CT has its limitation for limited accessibility, high economical cost, and exposure to ionizing radiation, especially for pediatric patients. To avoid unnecessary head CT, researchers have developed rule-based clinical decision support tools. We developed a prediction model using deep neural network (DNN) based multi-task learning (MTL) and the national-level trauma registry (Emergency Department In-depth Injury Surveillance, EDIIS). Material & Method For the sources of data, we utilized the Emergency Department In-depth Injury Surveillance (EDIIS) from 2011 to 2017 and the trauma registry from the Samsung Medical Center (SMC) from 2012 to 2019. We used EDIIS as a development and validation set and SMC as an external validation set. Furthermore, we selected the sub-group validation set to test our algorithm to the group of patients who are more sensitive for head CT scans. For our prediction model, we developed a prediction algorithm using multi-task learning (MTL) based on a deep neural network (DNN). Our multi-task outcomes include multiple outcomes. Our primary outcome is predicting the intracranial hemorrhage. Our secondary outcomes are trauma brain injury, diagnosis codes, visit disposition, and operation related to head injury. To evaluate the performance of our prediction, we used the area under the receiver operating characteristic curve (AUROC), the area under the precision-recall curve (AUPRC), and the frequency-weighted AUROC. Result A total of 777,675 patients\' records are used in our study. For the training set, there are 625,078 records from EDIIS between 2011 and 2016 are used. For the validation, there are 124,948 records in EDIIS from 2017 and 27,649 records in SMC trauma database are used. For the model performance for our primary outcome (intracranial hemorrhage), our model obtained AUROC of 0.94 and 0.91, and AUPRC of 0.63 and 0.36 in EDIIS 2017 and SMC respectively. With sub-group validation set, it obtained AUROC of 0.93 and 0.90, and AUPRC of 0.50 and 0.28. Conclusion We utilized multi-task deep learning to develop a prediction algorithm for predicting multiple outcomes. For the primary outcome, our prediction model obtained AUROC of 0.94.

r면 사파이어 위에 HVPE로 성장된 a면 GaN 에피층의 성장온도 효과 및 1000℃에서의 V/III족 비의 효과

하주형,박미선,이원재,최영준,이혜용,Ha, Ju-Hyung,Park, Mi-Seon,Lee, Won-Jae,Choi, Young-Jun,Lee, Hae-Yong 한국결정성장학회 2015 한국결정성장학회지 Vol.25 No.2

Ga source 채널의 HCl flow가 700 sccm, 그리고 V/III족 비가 10으로 고정되었을 때, r-면 사파이어 위에 HVPE로 성장된 a-면 GaN 에피층 특성에 대한 성장 온도 영향을 연구하였다. 추가적으로 성장온도가 $1000^{\circ}C$, 그리고 Ga source 채널의 HCl flow가 700 sccm으로 고정되었을 때, 공급가스에 대한 V/III족 비 영향에 대하여 연구하였다. 성장온도가 높아지면서, a-면 GaN 에피층에 대한 (11-20) 면의 Rocking curve(RC)의 반치폭 값이 감소하였고 a-면 GaN 에피층의 성장두께는 증가하였다. $1000^{\circ}C$에서 V/III족 비가 높아짐에 따라, (11-20) 면의 RC의 반치폭 값이 감소하였고, a-면 GaN 에피층의 성장두께가 증가하였다. $1000^{\circ}C$와 V/III족 비=10에서 성장된 a-면 GaN 에피층이 (11-20) 면에서 가장 낮은 RC 반치폭인 734 arcsec을 보이며, RC측정을 통한 (11-20) 면의 방위각 가장 작은 영향을 보여준다.

r-Plane sapphire 위에 HVPE에 의해 성장한 a-plane GaN에피텍셜층의 V/III족 ratio에 따른 특성 변화

하주형,박미선,이원재,최영준,이혜용,Ha, Ju-Hyung,Park, Mi-Seon,Lee, Won-Jae,Choi, Young-Jun,Lee, Hae-Yong 한국결정성장학회 2014 한국결정성장학회지 Vol.24 No.3

V/III족 ratio의 변화에 따른 r-plane의 sapphire 위에 HVPE로 성장한 a-plane GaN 에피텍셜층의 특성변화를 연구하였다. V/III족 ratio가 증가함에 따라서, a-plane (11-20) GaN에 대한 Rocking Curve의 FWHM의 값이 감소하며, 성장된 GaN의 표면 거칠기도 감소하고, 성장성도는 증가하다가 V/III족 ratio 7까지는 증가하다가 다시 감소하는 경향을 보여준다. 즉 V/III족 ratio 10에서 a-plane (11-20) GaN에 대한 Rocking Curve의 FWHM의 가장 작은 829 arcsec값을 보이고, 표면거칠기도 가장 작은 1.58 nm 값을 보인다. 또한 광학현미경상에서 관찰되는 내부 Crack 또는 void가 가장 적게 발생하였다. 그리고 M모양의 Azimuth angle 의존도를 전 샘플에서 보이며, V/III족 ratio 10에서 FWHM 최대값과 최소값의 편차값이 439 arcscec로 가장 작은 차이를 보였다.

절삭력 물리 모델 기반 가공용로봇 생산 시스템 디지털 트윈 플랫폼

하주형,김동민 한국정밀공학회 2024 한국정밀공학회지 Vol.41 No.6

Digital twin technology offers the advantage of monitoring the status of equipment, systems, and more in a virtualenvironment, allowing validation through simulation. This technology has found numerous applications in the industrialrobotics field, driven by recent advancements in the manufacturing industry. Consequently, predicting machining qualityusing digital twin technology is imperative for ensuring high-quality processed goods. In this study, we developed a digitaltwin program based on a cutting-force physical model and created a performance enhancement module that allows thevisualization of material removal for user convenience. The predicted cutting forces from both conventional CNC and thephysical model demonstrate a high accuracy of within 2%. Within the digital twin environment, the error rate for the roboticdrilling process is 13.5%. Building upon this, we developed and validated a module for material removal visualization,aiming to increase convenience for on-site operators.

Correlation between lipid profiles and cardiovascular outcomes in CKD patients: KNOW-CKD study

하주형,남기헌,박정탁,유태현,강신욱,최규헌,장태익,오국환,안규리,한승혁 대한내과학회 2018 대한내과학회 추계학술발표논문집 Vol.2018 No.1

고정익 수직이착륙 무인비행기용 직렬하이브리드 추진 시스템의 발전 제어 시뮬레이션

하주형,박규성,문창모,이근호 한국항공우주학회 2015 한국항공우주학회 학술발표회 논문집 Vol.2015 No.4

무인 비행기의 항속 시간을 증가시키기 위해 여러 연구들이 진행되어 왔다. 본 연구에서는 수직이착륙에 필요한 큰 에너지를 에너지 밀도가 높은 배터리로 공급하고, 기존의 엔진보다 무게가 작은 엔진을 전기모터와 직렬 연결하여 배터리 충전을 위한 직렬하이브리드 추진 시스템을 구성하였다. 추진 동력원에는 BLDC모터가 사용된다. 이착륙 시에는 큰 전력이 필요하므로 배터리와 발전기에서 전력을 공급하고, 순항 시에는 발전기에서 공급한 에너지로 순항에 필요한 에너지와 충전을 위한 에너지를 공급하도록 알고리즘을 구성하였다. 이 시스템의 시뮬레이션을 통하여 발전 제어 알고리즘을 검증하고자 한다. Several studies have been conducted for increasing the endurance time. In this paper, the supplied energy for VTOL should be the large energy by the high energy density battery. The engine connect to the electric motor in series that is smaller than pervious one to be composed of Serial hybrid propulsion system for battery charging. BLDC motor is employed as propulsive power source. This algorithm is consist of VTOL mode and cruising mode according to the power supply pattern. In case of VTOL mode, high power is indispensable that are supplied by generator and battery. And during the cruise, the energies that are needed for generator, airplane cruising and battery charging should be taken into account. this simulation was used for verifying generating control algorithm.

Effects of the Group V/III Ratio and a GaN Inter-layer on the Crystal Quality of InN Grown by Using the Hydride Vapor-phase Epitaxy Method

하주형,왕주안,이원재,최영준,Hae-Yong Lee,Jung-Gon Kim,Hiroshi Harima 한국물리학회 2015 THE JOURNAL OF THE KOREAN PHYSICAL SOCIETY Vol.66 No.6

The hydride vapor-phase epitaxy (HVPE) method was used to deposit high-quality InN layerson GaN inter-layer/sapphire (0001) structures that had been fabricated by using either the HVPEmethod or the metal-organic chemical-phase deposition (MOCVD) method. The effects of the groupV/III ratio and different GaN inter-layers on the crystal quality of the InN layers were systemicallyinvestigated. The InN layer grown at a low Group V/III ratio revealed a high crystal qualitywith a two-dimensional (2D) growth mode. Also, the 110.7-nm-thick InN layer grown by usingHVPE on a GaN inter-layer/sapphire (0001) substrate structure that had been fabricated by usingMOCVD had a high crystal quality, with the full width at half maximum (FWHM) of the InNX-ray diffraction (XRD) peak being about 844 arcsec, and a smooth surface with an atomic forcemicroscopy (AFM) roughness of about 0.07 nm. On the other hand, the 145.7-nm-thick InN layergrown by using HVPE on a GaN inter-layer/sapphire (0001) substrate structure that had beenfabricated by using the HVPE method had a lower crystal quality, a FWHM value for the InN(0002) peak of about 2772 arcsec, and a surface roughness of about 3.73 nm. In addition, the peakof the E2 (high) phonon mode for the 110.7-nm-thick InN layer grown by using HVPE on a GaNinter-layer/sapphire (0001) structure that had been fabricated by using MOCVD was detected at491 cm−1 and had a FWHM of 9.9 cm−1. As a result, InN layers grown by using HVPE on GaNinter-layer/sapphire (0001) substrate structures fabricated by using MOCVD have a high crystalquality and a reduced Raman value, which agrees well with the results of the XRD analysis.