• 응용통계연구
• /
• 제29권5호
• /
• pp.961-975
• /
• 2016

H-SVM은 입력변수들이 그룹화 되어 있는 경우 분류함수의 추정에서 그룹 및 그룹 내의 변수선택을 동시에 할 수 있는 방법론이다. 그러나 H-SVM은 입력변수들의 중요도에 상관없이 모든 변수들을 동일하게 축소 추정하기 때문에 추정의 효율성이 감소될 수 있다. 또한, 집단별 개체수가 상이한 불균형 자료의 분류분석에서는 분류함수가 편향되어 추정되므로 소수집단의 예측력이 하락할 수 있다. 이러한 문제점들을 보완하기 위해 본 논문에서는 적응적 조율모수를 사용하여 변수선택의 성능을 개선하고 집단별 오분류 비용을 차등적으로 부여하는 WAH-SVM을 제안하였다. 또한, 모의실험과 실제자료 분석을 통하여 제안한 모형과 기존 방법론들의 성능 비교하였으며, 제안한 모형의 유용성과 활용 가능성 확인하였다.

• Journal of Radiation Protection and Research
• /
• 제16권1호
• /
• pp.53-72
• /
• 1991

An integrated model is developed in this paper for the performance assessment of high level radioactive waste repository. This integrated model consists of two simple mathematical models. One is a multiple-barrier failure model of the repository system based on constant failure rates which provides source terms to biosphere. The other is a biosphere model which has multiple pathways for radionuclides to reach to human. For the parametric uncertainty and sensitivity analysis for the risk assessment of high level radioactive waste repository, Latin hypercube sampling and rank correlation techniques are applied to this model. The former is cost-effective for large computer programs because it gives smaller error in estimating output distribution even with smaller number of runs compared to crude Monte Carlo technique. The latter is good for generating dependence structure among samples of input parameters. It is also used to find out the most sensitive, or important, parameter groups among given input parameters. The methodology of the mathematical modelling with statistical analysis will provide useful insights to the decision-making of radioactive waste repository selection and future researches related to uncertain and sensitive input parameters.

• 대한전자공학회논문지SP
• /
• 제44권6호
• /
• pp.47-53
• /
• 2007

블록 기반 프레임 레이트 변환 (frame-rate conversion) 또는 필름 떨림 보상 (film judder compensation)을 수행하기 위해서는 참 움직임 벡터(true motion vector)를 찾아야 한다. 이를 위해서 현재 블록의 공간적 및 시간적 상관성을 최대로 하여 시각적으로 덜 부자연스럽게 느끼도록 하는 방법들이 연구되었다. 그러나 기존의 블록단위 절대값 차이의 합 (SAD)만으로는 비정형성 객체의 움직임 에러를 추정할 수 없었다. 본 논문에서는 비정형성 객체가 등속운동을 하는 경우 재귀적으로 기존의 움직임을 유지하도록 하는 방법을 제안하였다. 현재 블록의 등속움직임 추정값을 재귀평균으로 구하였으며 현재 블록 벡터의 신뢰도를 계산하여 원래의 움직임 벡터와 재귀평균 움직임 벡터중에서 가중치를 두도록 하였다. 실험결과 비정형성 등속운동 객체의 움직임을 블록기반으로 추정함을 확인할 수 있었다.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 2004년도 추계학술대회 학술발표 논문집 제14권 제2호
• /
• pp.433-436
• /
• 2004

In this paper, we introduce a new Fuzzy Polynomial Neural Networks (FPNNS)-like structure whose neuron is based on the Fuzzy Set-based Fuzzy Inference System (FS-FIS) and is different from that of FPNNS based on the Fuzzy relation-based Fuzzy Inference System (FR-FIS) and discuss the ability of the new FPNNS-like structure named Fuzzy Set-based Polynomial Neural Networks (FSPNN). The premise parts of their fuzzy rules are not identical, while the consequent parts of the both Networks (such as FPNN and FSPNN) are identical. This difference results from the angle of a viewpoint of partition of input space of system. In other word, from a point of view of FS-FIS, the input variables are mutually independent under input space of system, while from a viewpoint of FR-FIS they are related each other. The proposed design procedure for networks architecture involves the selection of appropriate nodes with specific local characteristics such as the number of input variables, the order of the polynomial that is constant, linear, quadratic, or modified quadratic functions being viewed as the consequent part of fuzzy rules, and a collection of the specific subset of input variables. On the parameter optimization phase, we adopt Information Granulation (IC) based on HCM clustering algorithm and a standard least square method-based learning. Through the consecutive process of such structural and parametric optimization, an optimized and flexible fuzzy neural network is generated in a dynamic fashion. To evaluate the performance of the genetically optimized FSPNN (gFSPNN), the model is experimented with using the time series dataset of gas furnace process.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2004년도 학술대회 논문집 정보 및 제어부문
• /
• pp.346-348
• /
• 2004

In this rarer, we introduce a new Fuzzy Polynomial Neural Networks (FPNNs)-like structure whose neuron is based on the Fuzzy Set-based Fuzzy Inference System (FS-FIS) and is different from that of FPNNs based on the Fuzzy relation-based Fuzzy Inference System (FR-FIS) and discuss the ability of the new FPNNs-like structurenamed Fuzzy Set-based Polynomial Neural Networks (FSPNN). The premise parts of their fuzzy rules are not identical, while the consequent parts of the both Networks (such as FPNN and FSPNN) are identical. This difference results from the angle of a viewpoint of partition of input space of system. In other word, from a point of view of FS-FIS, the input variables are mutually independent under input space of system, while from a viewpoint of FR-FIS they are related each other. In considering the structures of FPNN-like networks such as FPNN and FSPNN, they are almost similar. Therefore they have the same shortcomings as well as the same virtues on structural side. The proposed design procedure for networks' architecture involves the selection of appropriate nodes with specific local characteristics such as the number of input variables, the order of the polynomial that is constant, linear, quadratic, or modified quadratic functions being viewed as the consequent part of fuzzy rules, and a collection of the specific subset of input variables. On the parameter optimization phase, we adopt Information Granulation (IG) based on HCM clustering algorithm and a standard least square method-based learning. Through the consecutive process of such structural and parametric optimization, an optimized and flexible fuzzy neural network is generated in a dynamic fashion. To evaluate the performance of the genetically optimized FSPNN (gFSPNN), the model is experimented with using gas furnace process dataset.

• 한국콘텐츠학회:학술대회논문집
• /
• 한국콘텐츠학회 2008년도 춘계 종합학술대회 논문집
• /
• pp.788-791
• /
• 2008

본 연구에서는 한강수계의 소유역을 24개로 분할하였고 강우의 공간 분포를 작성하기 위해 151개의 강우관측소를 이용하여 강우자료를 정리하였다. 한강수계의 주요 제어지점으로 소양강댐, 충주댐, 충주조정지댐, 횡성댐, 화천댐, 춘천댐, 의암댐, 청평댐, 팔당댐을 선정하였다. SSARR(Streamflow Synthesis and Reservoir Regulation) 모형을 기반모형으로 선정하여 모형의 입력자료를 작성하고 2002년의 수문자료를 이용하여 매개변수의 민감도분석을 수행하였다. 민감도 분석 결과, 유역유출과 관련된 매개변수 중 토양습윤상태별 유출율, 침투량별 지하수유입률 및 지표수와 복류수를 분리하는 매개변수가 비교적 큰 민감도를 나타내었다.

• Nuclear Engineering and Technology
• /
• 제54권2호
• /
• pp.493-500
• /
• 2022

The objective of this study was to derive derived concentration guideline levels (DCGLs) reflecting the site-specific characteristics of KRR-1&2. A total of 7 nuclides (H-3, C-14, Co-60, Sr-90, Cs-137, Eu-152, and Eu-154) were selected for DCGLs derivation. Radiation dose at the sites was evaluated with RESRAD-ONSITE program. The dose contribution due to direct external exposure was the highest during the entire evaluation period. Ingestion had the second effect. The DCGLs of Co-60 was derived to be 0.051 Bq/g, and DCGLs of Cs-137 was 0.193 Bq/g. The DCGLs of H-3 showed the highest value of 129 Bq/g. The ratio of DCGLs derived by applying site-specific values and default values ranged from 0.27 to 19.6. For six nuclides excluding H-3, KRR-1&2 sites and the overseas NPP sites showed similar DCGLs. H-3 showed large differences in DCGLs from this study and overseas NPPs. The large difference resulted from input parameter values applied to the sites. In conclusion, it is critical to apply site-specific parameter values reflecting the site characteristics to derive DCGLs for decommissioned site clearance. The result of this study can be used as a reference for nuclide selection and DCGLs derivation reflecting the site characteristics when decommissioning nuclear facilities, including nuclear power plants in Korea.

• Smart Structures and Systems
• /
• 제15권1호
• /
• pp.227-244
• /
• 2015

This study develops a two stage procedure to identify the structural damage based on the optimized artificial neural networks. Initially, the modal strain energy index (MSEI) is established to extract the damaged elements and to reduce the computational time. Then the genetic algorithm (GA) and artificial neural networks (ANNs) are combined to detect the damage severity. The input of the network is modal strain energy index and the output is the flexural stiffness of the beam elements. The principal component analysis (PCA) is utilized to reduce the input variants of the neural network. By using the genetic algorithm to optimize the parameters, the ANNs can significantly improve the accuracy and convergence of the damage identification. The influence of noise on damage identification results is also studied. The simulation and experiment on beam structures shows that the adaptive parameter selection neural network can identify the damage location and severity of beam structures with high accuracy.

• Journal of Electrical Engineering and Technology
• /
• 제11권1호
• /
• pp.143-149
• /
• 2016

The two-stage converter is widely used in traditional DC/AC inverter. It has several disadvantages such as complex topology, large volume and high loss. In order to overcome these shortcomings, a novel half load-cycle worked dual SEPIC single-stage inverter, which is based on the analysis of the relationship between input and output voltages of SEPIC converters operating in the discontinuous conduction mode (DCM), is presented in this paper. The traditional single-stage inverter has remarkable advantages in small and medium power applications, but it can’t realize boost DC/AC output directly. Besides one pre-boost DC/DC converter is needed between the DC source and the traditional single-stage inverter. A novel DC/AC inverter without pre-boost DC/DC converter, which is comprised of two SEPIC converters, is studied. The output of dual SEPIC converters is connected with anti-parallel and half load-cycle control is used to realize boost and buck DC/AC output directly and work properly, whatever the DC input voltage is higher or lower than the AC output voltage. The working principle, parameter selection and the control strategy of the inverters are analyzed in this paper. Simulation and experiment results verify the feasibility of the new inverter.

• Computers and Concrete
• /
• 제12권2호
• /
• pp.151-167
• /
• 2013

Prediction of concrete properties is an important issue for structural engineers and different methods are developed for this purpose. Most of these methods are based on experimental data and use measured data for parameter estimation. Three typical methods of output estimation are Categorized Linear Regression (CLR), Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN). In this paper a statistical cleansing method based on CLR is introduced. Afterwards, MLR and ANN approaches are also employed to predict the compressive strength of structural lightweight aggregate concrete. The valid input domain is briefly discussed. Finally the results of three prediction methods are compared to determine the most efficient method. The results indicate that despite higher accuracy of ANN, there are some limitations for the method. These limitations include high sensitivity of method to its valid input domain and selection criteria for determining the most efficient network.