• 한국농림기상학회지
• /
• 제25권2호
• /
• pp.80-98
• /
• 2023

본 연구에서는 계절내-계절(Subseasonal to seasonal, S2S) 기후예측의 주별 예측 성능을 개선하기 위해서 딥러닝 기반의 후보정(post processing) 기술을 개발하였다. 그 첫 단계로, 일 최고, 최저기온과 일 강수를 목표 변수로, 자료의 특성과 분포에 적합한 자료 변환 및 특성 공학 기법을 규명하고자 하였다. 먼저, 6개 개별 기후모델의 S2S 예측 자료를 딥러닝 모델에 입력하기 위한 훈련자료로 변환하고, 이로부터 다중모델앙상블(Multi-Model Ensemble, MME) 기반 훈련자료를 구축하였다. 참값(label)으로는 ECMWF의 ERA5 재분석 자료를 사용하였다. 자료 변환 알고리즘은 최고 및 최저 차이를 계산하여 입력자료의 범위를 변형시키는 MinMax 및 MaxAbs 변환, 표준편차를 이용하는 Standard 변환 및 분위수를 지정하여 변형하는 Robust와 Quantile 변환으로 구성된 전처리 파이프라인을 구축하였으며, 변환된 훈련자료와 예측 변수와의 상관관계를 계산하여 순위에 따라 훈련자료의 특성을 선택하는 특성 선택 기법을 추가하였다. 본 연구는 U-Net 모델에 TimeDistributed wrapper를 모든 합성곱 층(convolutional layer)에 적용하여 활용하였다. 5개 알고리즘으로부터 변환된 6개 개별 기후모델 및 MME S2S 훈련자료(일 최고 및 최저기온, 강수)에 훈련 모델을 적용한 결과와 훈련 모델을 적용하지 않은 결과를 ERA5와의 공간상관계수(spatial Pattern Correlation Coefficient)를 계산하고 그 개선율인 기술 점수(skill score)를 평가한 결과, 일 강수의 PCC 기술 점수는 Standard 및 Robust 변환으로 처리된 것에서 전체 예측선행(1~4주)에 대해 모두 높았고, 일 최고 및 최저기온에서는 예측 선행시간 3~4주에서만 높게 나타났다. 또한, 일 강수에서 특성 선택에 따른 훈련자료의 차원 감소가 예측 성능 변화에 영향을 미치지 않는 것으로 나타났다. 일 최고 및 최저기온의 경우에는 특성 선택에 의한 훈련자료의 특성 정보 감소가 오히려 예측 성능을 저하시킬 수 있는 것으로 확인되었으며, 원시자료에서 예측성이 높은 1~2주 기온 예측 개선을 위한 적합한 전처리 변환 알고리즘이나 특성 선택을 찾을 수 없었다. 후속 연구에서는 원시 예측 성능이 강수에 비해 높으나 딥러닝 훈련 모델에 의한 후보정 효과가 미미한 예측 선행 1~2주 기온 예측의 저조 원인에 대해 탐색하고, 다양한 딥러닝 훈련 모델로의 적용 및 초매개변수 조정 등 학습 과정의 최적화를 통해 S2S 기후 예측 성능을 개선하고자 한다.

• International Journal of Naval Architecture and Ocean Engineering
• /
• 제10권3호
• /
• pp.413-420
• /
• 2018

This paper documents the result of a preliminary analysis on the influence of hull-ice friction coefficient on model resistance and power predictions and their correlation to full-scale measurements. The study is based on previous model-scale/full-scale correlations performed on the National Research Council - Ocean, Coastal, and River Engineering Research Center's (NRC/OCRE-RC) model test data. There are two objectives for the current study: (1) to validate NRC/OCRE-RC's modeling standards in regarding to its practice of specifying a CFC (Correlation Friction Coefficient) of 0.05 for all its ship models; and (2) to develop a correction methodology for its resistance and propulsion predictions when the model is prepared with an ice friction coefficient slightly deviated from the CFC of 0.05. The mean CFC of 0.056 and 0.050 for perfect correlation as computed from the resistance and power analysis, respectively, have justified NRC/OCRE-RC's selection of 0.05 for the CFC of all its models. Furthermore, a procedure for minor friction corrections is developed.

• 유체기계공업학회:학술대회논문집
• /
• 유체기계공업학회 2002년도 유체기계 연구개발 발표회 논문집
• /
• pp.111-115
• /
• 2002

The objective of this work is to develop improved slip factor model and correction method to predict flow through impeller in forward-curved centrifugal fan by investigating the validity of various slip factor models. Both steady and unsteady three-dimensional CFD analyses were performed with a commercial code tn validate the slip factor model and the correction method. The results show that the improved slip factor model presented in this paper could provide more accurate predictions for forward-curved centrifugal impeller than the other slip factor models since the presented model takes into account the effect of blade curvature. The comparison with CFD results also shows that the improved slip factor model coupled with the present correction method provides accurate predictions for mass-averaged absolute circumferential velocity at the exit of impeller near and above the flow rate of peaktotal pressure coefficient.

• Journal of Mechanical Science and Technology
• /
• 제18권2호
• /
• pp.302-312
• /
• 2004

This work developed improved slip factor model and correction method to predict flow through impeller in forward-curved centrifugal fan. Both steady and unsteady three-dimensional CFD analyses were performed to validate the slip factor model and the correction method. The results show that the improved slip factor model presented in this paper could provide more accurate predictions for forward-curved centrifugal impeller than the other slip factor models since the present model takes into account the effect of blade curvature. The correction method is provided to predict mass-averaged absolute circumferential velocity at the exit of impeller by taking account of blockage effects induced by the large-scale backflow near the front plate and flow separation within blade passage. The comparison with CFD results also shows that the improved slip factor model coupled with the present correction method provides accurate predictions for mass-averaged absolute circumferential velocity at the exit of impeller near and above the flow rate of peak total pressure coefficient.

• 한국정보과학회논문지:컴퓨팅의 실제 및 레터
• /
• 제9권6호
• /
• pp.693-698
• /
• 2003

야구는 매 경기마다 다양한 기록을 생성하며, 이러한 기록을 기반으로 다음 경기에 대한 승패예측이 이루어진다. 프로야구 승패 예측에 대한 연구는 많은 사람들에 의해 행해져 왔으나 아직 이렇다할 결과를 얻지 못하고 있는 상태이다. 이처럼 승패 예측이 어려운 이유는 많은 경기 기록들 중 승패 예측에 영향을 주는 요소의 선별이 어렵고, 예측에 사용된 자료들 간의 중복 요인으로 인해 학습 모델의 복잡도만 증가시킬 뿐 좋은 성능을 보이지 못하고 있다. 이에 본 논문에서는 전문가들의 의견을 바탕으로 학습 요소들을 선택하고, 선택된 자료들을 이용하여 휴리스틱 함수를 구성하였다. 요소들 간의 조합을 통해 예측에 영향을 줄 수 있는 새로운 값을 산출함과 동시에 학습 알고리즘에 사용될 입력 값의 차원을 줄일 수 있는 혼합형 모델을 제안하였다. 그 결과, 학습 알고리즘으로 사용된 역전파 알고리즘의 복잡도를 감소시키고, 프로야구 경기 승패 예측에 있어서도 정확성이 향상되었다.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2023년도 학술발표회
• /
• pp.29-29
• /
• 2023

To mitigate the damaging impacts of floods, accurate prediction of runoff, streamflow and flood inundation is needed. Conventional approach of simulating hydrology and hydraulics using loosely coupled models cannot capture the complex dynamics of surface and sub-surface processes. Additionally, the scarcity of data in ungauged basins and quality of data in gauged basins add uncertainty to model predictions, which need to be quantified. In this presentation, first the role of integrated modeling on creating accurate flood simulations and inundation maps will be presented with specific focus on urban environments. Next, the use of machine learning in producing streamflow predictions will be presented with specific focus on incorporating covariate shift and the application of theory guided machine learning. Finally, a framework to quantify the uncertainty in flood models using Hierarchical Bayesian Modeling Averaging will be presented. Overall, this presentation will highlight that creating accurate information on flood magnitude and extent requires innovation and advancement in different aspects related to hydrologic predictions.

• Steel and Composite Structures
• /
• 제29권3호
• /
• pp.309-318
• /
• 2018

Steel girders are the structural members often used for passing long spans. Mostly being subjected to patch loading, or concentrated loading, steel girders are likely to face sudden deformation or damage e.g., web breathing. Horizontal or vertical stiffeners are employed to overcome this phenomenon. This study aims at assessing the feasibility of a machine learning method, namely the support vector machines (SVM) in predicting the patch loading resistance of longitudinally stiffened webs. A database consisting of 162 test data is utilized to develop SVM models and the model with best performance is selected for further inspection. Existing formulations proposed by other researchers are also investigated for comparison. BS5400 and other existing models (model I, model II and model III) appear to yield underestimated predictions with a large scatter; i.e., mean experimental-to-predicted ratios of 1.517, 1.092, 1.155 and 1.256, respectively; whereas the selected SVM model has high prediction accuracy with significantly less scatter. Robust nature and accurate predictions of SVM confirms its feasibility of potential use in solving complex engineering problems.

• 한국원자력학회:학술대회논문집
• /
• 한국원자력학회 1995년도 추계학술발표회논문집(1)
• /
• pp.456-461
• /
• 1995

The concept of hyperbolicity breaking is applied to predict the flow regime transition from inverted annular flow (IAF) to agitated inverted annular flow (AIAF). The resultant correlation has the similar form to Takenaka's empirical one. To validate the proposed model, it is applied to predict Takenaka's experimental results using R-113 refrigerant with four different tube diameters of 3, 5, 7 and 10 mm. The proposed model gives accurate predictions for the tube diameters of 7 and 10 min. As the tube diameter decreases, the differences between the predictions and the experimental results increase slightly. The flow regime transition from AIAF to dispersed flow (DF) is described by the drift flux model.

• 한국전산유체공학회:학술대회논문집
• /
• 한국전산유체공학회 2010년 춘계학술대회논문집
• /
• pp.174-180
• /
• 2010

Conductive and radiative thermal model configurations of an imager of a geostationary satellite are presented. A two-plane method is introduced for three dimensional conductive coupling which is not able to be treated by thin shell plate thermal modeling technique. Especially the two-plane method is applied to massive matters and PIP(Payload Interface Plate) in the imager model. Some massive matters in the thermal model are modified by adequate correction factors or equivalent thickness in order to obtain the numerical results of thermal modeling to be consistent with the analytic model. More detailed nodal breakdown is specially employed to the object which has the rapid temperature gradient expected by a rule of thumb. This detailed thermal model of the imager is supposed to be used for detailed analyses and test predictions, and be correlated with the thermal vacuum test results before final in-flight predictions.

• Journal of Mechanical Science and Technology
• /
• 제15권7호
• /
• pp.889-894
• /
• 2001

In this study, the performances of various turbulence closure models are evaluated in the calculation of a transonic flow over axisymmetric bump. k-$\varepsilon$, explicit algebraic stress, and two Reynolds stress models, i.e., GL model proposed by Gibson & Launder and SSG model proposed by Speziale, Sarkar and Gatski, are chosen as turbulence closure models. SSG Reynolds stress model gives best predictions for pressure coefficients and the location of shock. The results with GL model also show quite accurate prediction of pressure coefficients down-stream of shock wave. However, in the predictions of mean velocities and turbulent stresses, the results are not so satisfactory as in the prediction of pressure coefficients.