• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.88-88
• /
• 2021

최근 수자원과 수질관리 분야에 자료기반 머신러닝 모델과 딥러닝 모델의 활용이 급증하고 있다. 그러나 딥러닝 모델은 Blackbox 모델의 특성상 고전적인 질량, 운동량, 에너지 보존법칙을 고려하지 않고, 데이터에 내재된 패턴과 관계를 해석하기 때문에 물리적 법칙을 만족하지 않는 예측결과를 가져올 수 있다. 또한, 딥러닝 모델의 예측 성능은 학습데이터의 양과 변수 선정에 크게 영향을 받는 모델이기 때문에 양질의 데이터가 제공되지 않으면 모델의 bias와 variation이 클 수 있으며 정확도 높은 예측이 어렵다. 최근 이러한 자료기반 모델링 방법의 단점을 보완하기 위해 프로세스 기반 수치모델과 딥러닝 모델을 결합하여 두 모델링 방법의 장점을 활용하는 연구가 활발히 진행되고 있다(Read et al., 2019). Process-Guided Deep Learning (PGDL) 방법은 물리적 법칙을 반영하여 딥러닝 모델을 훈련시킴으로써 순수한 딥러닝 모델의 물리적 법칙 결여성 문제를 해결할 수 있는 대안으로 활용되고 있다. PGDL 모델은 딥러닝 모델에 물리적인 법칙을 해석할 수 있는 추가변수를 도입하며, 딥러닝 모델의 매개변수 최적화 과정에서 Cost 함수에 물리적 법칙을 위반하는 경우 Penalty를 추가하는 알고리즘을 도입하여 물리적 보존법칙을 만족하도록 모델을 훈련시킨다. 본 연구의 목적은 대청호의 수심별 수온을 예측하기 위해 역학적 모델과 딥러닝 모델을 융합한 PGDL 모델을 개발하고 적용성을 평가하는데 있다. 역학적 모델은 2차원 횡방향 평균 수리·수질 모델인 CE-QUAL-W2을 사용하였으며, 대청호를 대상으로 2017년부터 2018년까지 총 2년간 수온과 에너지 수지를 모의하였다. 기상(기온, 이슬점온도, 풍향, 풍속, 운량), 수문(저수위, 유입·유출 유량), 수온자료를 수집하여 CE-QUAL-W2 모델을 구축하고 보정하였으며, 모델은 저수위 변화, 수온의 수심별 시계열 변동 특성을 적절하게 재현하였다. 또한, 동일기간 대청호 수심별 수온 예측을 위한 순환 신경망 모델인 LSTM(Long Short-Term Memory)을 개발하였으며, 종속변수는 수온계 체인을 통해 수집한 수심별 고빈도 수온 자료를 사용하고 독립 변수는 기온, 풍속, 상대습도, 강수량, 단파복사에너지, 장파복사에너지를 사용하였다. LSTM 모델의 매개변수 최적화는 지도학습을 통해 예측값과 실측값의 RMSE가 최소화 되로록 훈련하였다. PGDL 모델은 동일 기간 LSTM 모델과 동일 입력 자료를 사용하여 구축하였으며, 역학적 모델에서 얻은 에너지 수지를 만족하지 않는 경우 Cost Function에 Penalty를 추가하여 물리적 보존법칙을 만족하도록 훈련하고 수심별 수온 예측결과를 비교·분석하였다.

• Nuclear Engineering and Technology
• /
• v.28 no.2
• /
• pp.103-117
• /
• 1996

A mathematical model of vapor explosion propagation is presented. The model predict two-dimensional, transient flow fields and energies of the four fluid phases of melt drop, fragmented debris, liquid coolant and vapor coolant by solving a set of governing equations with the relevant constitutive relations. These relations include melt fragmentation, coolant-phase-change, and heat and momentum exchange models. To allow thermodynamic non-equilibrium between the coolant liquid and vapor, an equation of state for oater is uniquely formulated. A multiphase code, TRACER, has been developed based on this mathematical formulation. A set of base calculations for tin/water explosions show that the model predicts the explosion propagation speed and peak pressure in a reasonable degree although the quantitative agreement relies strongly on the parameters in the constitutive relations. A set of calculations for sensitivity studies on these parameters have identified the important initial conditions and relations. These are melt fragmentation rate, momentum exchange function, heat transfer function and coolant phase change model as well as local vapor fractions and fuel fractions.

• Solar Energy
• /
• v.16 no.2
• /
• pp.79-86
• /
• 1996

In order to reduce damages due to cavitation effects and to improve performance of fluid machinery, supercavitation around the cascade and the hydraulic characteristics of supercavitating cascade must be analyzed accurately. And the study on the effects of cavitation on fluid machinery and analysis on the performances of supercavitating hydrofoil through various elements governing flow field are critically important. In this study comparison of experiment results with the computed results of linear theory using singularity method was obtainable. Specially singularity points like sources and vortexes on hydrofoil and freestreamline were distributed to analyze two dimensional flow field of supercavitating cascade, and governing equations of flow field were derived and hydraulic characteristics of cascade were calculated by numerical analysis of the governing equations.

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• v.25 no.10
• /
• pp.562-567
• /
• 2013

1D-numerical analysis of the network algorithm with the orifice equation for the relationship between pressure difference and flowrate has been mostly used to analyse leakage flow at the gap. In this study, a 3D-numerical method applying momentum loss model to the gap region in the computational domain is represented to reflect effectively the effect of leakage flow by determining the proportion of pressure difference to air passage velocity. While the 3D-numerical method is verified through the computation of the two compartments model, the numerical analysis of the stack effect in a building stairway is performed. As the temperature of air outside drops, the pressure in the upper stairway and leakage flowrate through the gap in the door rise. The change of gap area does not have an effect on pressure in the stairway for the analysis conditions.

• Journal of the Korean Society of Propulsion Engineers
• /
• v.14 no.2
• /
• pp.1-9
• /
• 2010

LES(Large eddy simulation) of breakup and droplet atomization of a liquid jet into cross turbulent flow was performed. Two phase flow of gas and liquid phases were modeled by the mixed numerical scheme of both Eulerian and Lagrangian methods for gas and liquid droplet respectively. The breakup process of a liquid column and droplets was observed by implementing the blob-KH wave breakup model. The penetration depth into cross flow was comparable with experimental data for several variants of the liquid-gas momentum flux ratio by varying liquid injection velocity. SMD(Sauter Mean Diameter) distribution downstream of jet was analyzed.

• Journal of the Korean Society of Propulsion Engineers
• /
• v.15 no.3
• /
• pp.31-39
• /
• 2011

The flowfield in a solid rocket motor was simulated at the wall-injection test model, which has a fin-slot grain and submerged nozzle, and visualized by a smoke-wire. The high speed CCD camera captured the visualized images around the nozzle inlet through the grain center port. The vortical tube structure and circumferential flow patterns at the nozzle throat were visualized. The radial momentum transfer caused by the shear-interactions of slot-outlet flow, fin-base flow and grain port flow from upstream worked as the source of these phenomena.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.130-130
• /
• 2023

제트류는 다양한 크기와 운동량의 에디가 복잡하게 혼합되어 이루어져 있으며, 이를 정확하게 모델링하고 이해하기 위해서는 제트류의 다양한 특성들을 잘 반영하여 연구를 수행해야 한다. 다양한 연구 수행 방법 중 수치해석 방법은 상대적으로 공간 및 시간적 비용이 적게 들어서 널리 사용되고 있다. 이러한 수치해석 방법에는 DNS(Direct Numerical Simulation), LES(Large Eddy Simulation), RANS(Reynolds Averaged Navier Stokes) 등이 있으며, 그중 LES는 난류 모델링을 사용하는 RANS 방법에 비해 더욱 정확한 흐름 모델링을 제공하는 장점이 있다. 이러한 LES는 대규모 에디는 직접 해석하면서, 일정 크기 이하의 에디는 모델링을 사용해 해석하는 것이 특징이다. 하지만, LES를 사용하기 위해서는 적절한 그리드 크기를 결정하는 것이 중요하며, 이는 모델의 정확성과 연산 비용에 큰 영향을 미친다. 하지만, 여전히 적절한 그리드 크기를 결정하는 것은 어려운 문제이다. 이러한 LES 모델링을 사용할 때 적절한 그리드 크기를 결정하기 위해서는 정확한 시간 평균 속도 변동을 연구하는 것이 앞서 선행되어야 한다. 따라서, 본 연구에서는 기계학습 기반 접근 방식을 사용하여 난류 제트 내 시간 평균 속도 변동을 예측하는 연구를 진행하였다. 즉, 난류 제트 역학을 이해하는 데 중요한 파라미터인 시간 평균 유속을 이용하여 시간 평균 속도 변동을 예측하는 데 초점을 맞추었다. 모델의 성능은 평균 제곱 오차와 R-제곱 등 다양한 지표를 사용하여 평가되었다.

• Journal of the Society of Naval Architects of Korea
• /
• v.52 no.3
• /
• pp.206-221
• /
• 2015

Supercavitation is a modern technique which can be used to surround an underwater vehicle with a bubble in order to reduce the resistance of the vehicle. When the vehicle is at low speed in the deep sea, the cavitation number is relatively big and it is difficult to generate a cavity large enough to envelope the vehicle. In this condition, the artificial cavity, called ventilated cavity, can be used to solve this problem by supplying gas into the cavity and can maintain supercavitating condition. In this paper, a relationship between the ventilation gas supply rate and the cavity shape is determined. Based on the relationship a ventilation rate control is developed to maintain the supercavitating state. The performance of the ventilation control is verified with a depth change control. In addition, dynamics modeling for the supercavitating vehicle is performed by defining forces and moments acting on the vehicle body in contact with water. Simulation results show that the ventilation control can maintain the supercavity of an underwater vehicle at low speed in the deep sea.

• Fire Science and Engineering
• /
• v.13 no.4
• /
• pp.30-37
• /
• 1999

In this study, reduced-scale experiments as the alternative to a real-scale fire test were conducted to understand fire properties in atrium space. The scaling laws were derived from $\pi$-parameters which were deduced by dimensional analysis of governing equations (continuity, conservation of momentum and conservation energy). The 1/50 scale experiment simulated the real-scale fire test in SIVANS atrium at Japan were conducted under the scaling laws. And this results were compared with real-scale experiment results. Furthermore these results were visualized by video recording system using laser light sheet.

• 한국신재생에너지학회:학술대회논문집
• /
• 2007.06a
• /
• pp.374-377
• /
• 2007

This paper deals with a pitch control for reducing load of the wind turbine system. To make a model of the wind turbine system, the Momentum Theory and Blade Element Theory are used. Considering wind shear, wind model was also built. Due to a difference of the wind speed between upper parts and lower parts of the sweep area, overturning moment of the wind turbine is generated. So, in this paper through analyzing of the system model of the wind turbine, a control algorithm which was able to achieve both maintaining power and reducing overturning moment was proposed. Using matlab simulink, controller performance was verified.