• Journal of Korea Water Resources Association
• /
• v.34 no.4
• /
• pp.303-316
• /
• 2001

In this study, Radial Basis Function(RBF) Neural Networks Model, a kind of Hybrid Neural Networks was applied to hydrological forecasting in a small watershed. RBF Neural Networks Model has four kinds of parameters in it and consists of unsupervised and supervised training patterns. And Gaussian Kernel Function(GKF) was used among many kinds of Radial Basis Functions(RBFs). K-Means clustering algorithm was applied to optimize centers and widths which ate the parameters of GKF. The parameters of RBF Neural Networks Model such as centers, widths weights and biases were determined by the training procedures of RBF Neural Networks Model. And, with these parameters the validation procedures of RBF Neural Networks Model were carried out. RBF Neural Networks Model was applied to Wi-Stream basin which is one of the IHP Representative basins in South Korea. 10 rainfall events were selected for training and validation of RBF Neural Networks Model. The results of RBF Neural Networks Model were compared with those of Elman Neural Networks(ENN) Model. ENN Model is composed of One Step Secant BackPropagation(OSSBP) and Resilient BackPropagation(RBP) algorithms. RBF Neural Networks shows better results than ENN Model. RBF Neural Networks Model spent less time for the training of model and can be easily used by the hydrologists with little background knowledge of RBF Neural Networks Model.

• Geomechanics and Engineering
• /
• v.18 no.5
• /
• pp.493-502
• /
• 2019

Internal stability is an important safety issue for levees, embankments, and other earthen structures. Since a large part of the world's population lives near oceans, lakes and rivers, floods resulting from breaching of dams can lead to devastating disasters with tremendous loss of life and property, especially in densely populated areas. There are some main factors that affect the internal stability of dams, levees and other earthen structures, such as the erodibility of the soil, the water velocity inside the soil mass and the geometry of the earthen structure, etc. Thus, the mechanism of internal erosion and stability of soils is very complicated and it is vital to investigate the assessment methods of internal stability of soils in embankment dams and their foundations. This paper presents an improved support vector machine (SVM) model to predict the internal stability of soils. The grid search algorithm (GSA) is employed to find the optimal parameters of SVM firstly, and then the cross - validation (CV) method is employed to estimate the classification accuracy of the GSA-SVM model. Two examples of internal stability of soils are presented to validate the predictive capability of the proposed GSA-SVM model. In addition to verify the effectiveness of the proposed GSA-SVM model, the predictions from the proposed GSA-SVM model were compared with those from the traditional back propagation neural network (BPNN) model. The results showed that the proposed GSA-SVM model is a feasible and efficient tool for assessing the internal stability of soils with high accuracy.

• Geomechanics and Engineering
• /
• v.31 no.5
• /
• pp.529-543
• /
• 2022

The lattice-spring-based synthetic rock mass model (LS-SRM) technique has been extensively employed in large open-pit mining and underground projects in the last decade. Since the LS-SRM requires a complex and time-consuming calibration process, a robust approach was developed using the Response Surface Methodology (RSM) to optimize the calibration procedure. For this purpose, numerical models were designed using the Box-Behnken Design technique, and numerical simulations were performed under uniaxial and triaxial stress states. The model input parameters represented the models' micro-mechanical (lattice) properties and the macro-scale properties, including uniaxial compressive strength (UCS), elastic modulus, cohesion, and friction angle constitute the output parameters of the model. The results from RSM models indicate that the lattice UCS and lattice friction angle are the most influential parameters on the macro-scale UCS of the specimen. Moreover, lattice UCS and elastic modulus mainly control macro-scale cohesion. Lattice friction angle (flat joint fiction angle) and lattice elastic modulus affect the macro-scale friction angle. Model validation was performed using physical laboratory experiment results, ranging from weak to hard rock. The results indicated that the RSM model could be employed to calibrate LS-SRM numerical models without a trial-and-error process.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.137-137
• /
• 2021

In this study we aimed to enhance streamflow prediction skill of a land-surface hydrological model, WRF-Hydro, over one of the snow dominated catchments lies in Himalayan mountainous range, Astore. To assess the response of the Himalayan river flows to climate change is complex due to multiple contributors: precipitation, snow, and glacier melt. WRF-Hydro model with default glacier module lacks generating streamflow in summer period but recently developed WRF-Hydro-CROCUS model overcomes this issue by melting snow/ice from the glaciers. We showed that by implementing WRF-Hydro-CROCUS model over Astore the results were significantly improved in comparison to WRF-Hydro with default glacier module. To constraint the model with the observed streamflow we chose 17 sensitive parameters of WRF-Hydro, which include groundwater parameters, surface runoff parameters, channel parameters, soil parameters, vegetation parameters and snowmelt parameters. We used Dynamically Dimensioned Search (DDS) method to calibrate the daily streamflow with the Nash-Sutcliffe efficiency (NSE) being greater than 0.7 both in calibration (2009-2010) and validation (2011-2013) period. Based on the number of iterations per parameter, we found that the parameters related to channel and runoff process are most sensitive to streamflow. The attempts to address the responses of the streamflows to climate change are still very weak and vague especially northwest Himalayan Part of Pakistan and this study is one of a few successful applications of process-based land-surface hydrologic model over this mountainous region of UIB that can be utilized to have an in-depth understanding of hydrological responses of climate change.

• Journal of the Korean Data and Information Science Society
• /
• v.28 no.5
• /
• pp.971-980
• /
• 2017

The varying coefficient regression model has gained lots of attention since it is capable to model dynamic changes of regression coefficients in many regression problems of science. In this paper we propose a varying coefficient regression model that effectively considers the errors on both input and response variables, which utilizes the kernel method in estimating the varying coefficient which is the unknown nonlinear function of smoothing variables. We provide a generalized cross validation method for choosing the hyper-parameters which affect the performance of the proposed model. The proposed method is evaluated through numerical studies.

• Proceedings of the KSRS Conference
• /
• 2004.10a
• /
• pp.205-207
• /
• 2004

The simple method of the geometric reconstruction of satellite linear pushbroom images is investigated. The model of the sensor used is based on the SPOT model that is developed by Kraiky. The satellite trajectory is a Keplerian trajectory in the approximation. Four orbital parameters, longitude of the ascending $node(\omega),$ inclination of the orbit plan(I), latitude argument of the satellite(W) and distance between earth center and satellite, are used for the camera modeling. We suppose that four orbital parameters and satellite attitude angles are exactly acquired. Then, in order to refine model, the given attitude angles and orbital parameters is not changed, but time-independent four parameters associated with LOS(Line Of Sight) vector is updated. A pair of SPOT-5 images has been used for validation of proposed method. Two GCPs acquired by GPS survey is used to controlling the LOS vector. The results are that the RMSE of 16 checking points are about 4.5m. Because the ground resolution of SPOT-5 is 2.5m, the result obtained in this study has a good accuracy. It demonstrates that the sensor model developed by this study can be used to reconstruct the geometry of satellite image taken by pushbroom camera.

• Journal of Korean Society on Water Environment
• /
• v.29 no.3
• /
• pp.365-376
• /
• 2013

The purpose of this study was to evaluate effects of spatio-temporal changes in land uses and rainfall magnitude using the Soil and Water Assessment Tool (SWAT). Prior of application of the model to real-world problem, the model should be calibrated and validated properly. In most modeling approaches, the validation process is done assuming no significant changes occurring at the study watershed between calibration and validation periods, which is not proper assumption for agricultural watersheds. If simulated results obtained with calibrated parameters match observed data with higher accuracy for validation period, this does not always mean the simulated result represents rainfall-runoff, pollutant generation and transport mechanism for validation period because temporal and spatial variables and rainfall magnitude are often not the same. In this study SWAT was applied to Mandae study watershed in Korea to evaluate effects of spatio-temporal changes in landuses using 2009 and 2010 crop data for each field at the watershed. The Nash-Sutcliffe model efficiency (NSE) values for calibration and validation with either 2009 or 2010 was evaluated and the NSE value for calibration with 2009 and calibration with 2010 were compared. It was found that if there is substantial change in land use and rainfall, model calibration period should be determined to reflect those changes. Through these approaches, inherent limitation of the SWAT, which does not consider changes in land uses over the simulation period, was investigated. Also, Effects of changes in rainfall magnitude during calibration process were analyzed.

• Journal of the Korean Data and Information Science Society
• /
• v.28 no.1
• /
• pp.227-235
• /
• 2017

When the spatial information of each location is given specifically as coordinates it is popular to use the geographically weighted regression to incorporate the spatial information by assuming that the regression parameters vary spatially across locations. In this paper, we relax the linearity assumption of geographically weighted regression and propose a geographically weighted least squares-support vector machine for estimating geographically weighted mean by using the basic concept of kernel machines. Generalized cross validation function is induced for the model selection. Numerical studies with real datasets have been conducted to compare the performance of proposed method with other methods for predicting geographically weighted mean.

• Journal of the Korean Data and Information Science Society
• /
• v.21 no.6
• /
• pp.1319-1325
• /
• 2010

Quantile regression provides a more complete statistical analysis of the stochastic relationships among random variables. Sometimes quantile functions estimated at different orders can cross each other. We propose a new non-crossing quantile regression method applying support vector median regression to restricted regression quantile, restricted support vector quantile regression. The proposed method provides a satisfying solution to estimating non-crossing quantile functions when multiple quantiles for high dimensional data are needed. We also present the model selection method that employs cross validation techniques for choosing the parameters which aect the performance of the proposed method. One real example and a simulated example are provided to show the usefulness of the proposed method.

• 한국신재생에너지학회:학술대회논문집
• /
• 2010.06a
• /
• pp.126.1-126.1
• /
• 2010

In direct methanol fuel cells(DMFCs), it is well known that methanol crossover severely reduces the cell performance and the cell efficiency. There are a number of design and operating parameters that influence the methanol crossover. This indicates that a DMFC demands a high degree of optimization. For the successful design and operation of a DMFC system, a better understanding of methanol crossover phenomena is essential. The main objective of this study is to examine methanol-crossover phenomena in DMFCs. In this study, 1D DMFC model previously developed by Ko et al. is used. The simulation results were compared with methanol-crossover data that were measured by Eccarius et al. The numerical predictions agree well with the methanol crossover data and the model successfully captures key experimental trends.