• 한국수자원학회논문집
• /
• 제40권9호
• /
• pp.687-696
• /
• 2007

본 연구는 다목적 유전자알고리즘을 이용하여 Tank 모형의 매개변수를 추정하는데 있어서 선호적순서화(preference ordering)를 적용한 연구로써, 목적함수의 개수가 여러 개인 경우에 발생할 수 있는 파레토최적화의 단점을 해결하기 위한 것이다. 최적화를 위한 목적함수는 모두 4가지를 사용하였으며, 선호적순서화를 통해서 구한 2차 효율성(2nd order efficiency)을 가지면서 정도(degree)가 3인 4개의 해 중에서 1개의 해만을 최우선해로 선정하였다. NSGA-II로 도출된 최우선해의 적합성을 살펴보기 위해서, 자동보정방법인 Powell 방법과 SGA(simple genetic algorithm)를 매개변수 자동보정 방법으로 이용하고 하나의 단일목적함수로 사용해서 최적화한 결과와 비교해보았으며, 비교결과 다목적 유전자 알고리즘을 4개의 목적함수에 모두 적용해서 한번에 도출된 매개변수를 이용한 결과가 보정기간뿐만 아니라 검정기간에 대해서도 비교적 양호한 결과를 나타내는 것으로 나타났다.

• 전자공학회논문지C
• /
• 제35C권9호
• /
• pp.81-92
• /
• 1998

실수형 최적화 문제의 전역 최적해를 빠르고 정확하게 찾을 가능성을 높이기 위해, 서로 다른 진화특성을 가지는 여러 부집단들을 사용한 새로운 하이브리드 기법이 제안된다. 제안된 알고리듬은 세 개의 부집단을 사용하는데, 복잡한 적합도 함수를 가지는 문제에서 좋은 성능을 보이는 NPOSA 알고리듬이 두개의 부집단에 적용되고, 진화 방향과 크기가 조절되는 자기 적응 진화 알고리듬이 나머지 하나의 부집단에 적용되었다. 각 부집단들은 서로 다른 방법으로 진화하며 부집단들간의 상호교류를 통해 전역 최적해로 빠르게 도달하게 한다. 이 기법의 효율성은 몇 개의 표준 테스트 문제들을 사용하여 검증하였다. 마지막으로, 제안한 알고리듬이 실제 문제에 적용 가능함을 보이기 위해 카메라 파라메터의 최적값을 찾는 문제에 적용하였다. 보정 블럭에서 측정된 특징점들을 사용하여 오차 함수를 정의한 후, 하이브리드 방법이 그 오차 함수를 최소화하는 카메라 파라메터의 값을 찾을 수 있음을 보였다.

• 대한토목학회논문집
• /
• 제39권6호
• /
• pp.853-858
• /
• 2019

철도 네트워크를 효율적으로 사용하는 방법 중 하나는 연결선 건설을 통해 여러 노선이 하나의 선로를 공유하는 방법인데, 국내에서도 이미 사용하고 있거나 사업화 추진 중인 노선이 존재한다. 본 연구는 철도 연결선 건설시 사업화 단계가 아닌 계획 단계에서 연결선 건설의 위치 선정, 우선순위 산정 등에 필요한 연결선 설계 모형을 구축하고 이에 대한 해법을 제시하였다. 모형은 연결선건설비용, 노선운영비용, 이용자 총통행비용을 최소화하기 위한 비선형 최적화문제로 구축하였고, 결정변수로는 연결선 건설여부 및 방향, 노선의 열차 운행 빈도로 설정하였다. 본 연구에서는 문제 풀이 알고리즘과 경로선택 알고리즘을 각각 제시하였고, 예제네트워크에 적용을 통하여 모형의 실용성을 검증하였다. 향후 현실을 반영한 실제네트워크와 파라미터를 설정하여 연구의 결과를 발전시킬 여지가 있다.

• 한국물환경학회지
• /
• 제25권2호
• /
• pp.311-321
• /
• 2009

Long-Term Hydrologic Impact Assessment (L-THIA) was modified to improve runoff and pollutant load prediction for Korean watersheds with changes in land use classification and event mean concentration produced from observed data in Korea. The L-THIA model was linked with SCE-UA, which is one of the global optimization techniques, to automatically calibrate direct runoff. Modified L-THIA model was applied to Gumho River Basins to analyze spatial distribution of nonpoint source pollution. The results of model calibration during 1991~2000 and validation during 1981~1990 for direct runoff represented high model efficiency of 0.76 for calibration and 0.86 for validation. As a results of spatial analysis of nonpoint source pollution, the BOD was mainly loaded from urban area but SS, TN, and TP from agricultural area which is mainly located along the stream. Modified L-THIA model improve its accuracy with minimum imput data and application efforts. From this study, we can find out the L-THIA model is very useful tool to predict direct runoff and pollutant loads from the watershed and spatial analysis of nonpoint source pollution.

• 한국자동차공학회논문집
• /
• 제19권2호
• /
• pp.98-104
• /
• 2011

The introduction of a diesel engine into the passenger car and light duty applications in the United States involves significant technical challenges for the automotive makers. This paper describes the SCR System optimization procedure for such a diesel engine application to meet Tier2 Bin5 emission regulation. A urea SCR system, a representative $NO_x$ reduction after-treatment technique, is applied to a 3.0 liter diesel engine. To achieve the maximum $NO_x$ reduction performance, the exhaust system layout was optimized using series of the computational fluid dynamics and the urea distribution uniformity test. Furthermore a comprehensive simulation model for the key factors influencing $NO_x$ reduction performance was developed and embedded in the Simulink/Matlab environment. This model was then applied to the urea SCR system and played a key role to shorten the time needed for SCR control parameter calibration. The potential of a urea SCR system for reducing diesel $NO_x$ emission is shown for FTP75 and US06 emission standard test cycle.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권10호
• /
• pp.4117-4135
• /
• 2020

Due to the angle of view, illumination and scene diversity, real-time dynamic face detection and recognition is no small difficulty in those unrestricted environments. In this study, we used the intrinsic correlation between detection and calibration, using a multi-task cascaded convolutional neural network(MTCNN) to improve the efficiency of face recognition, and the output of each core network is mapped in parallel to a compact Euclidean space, where distance represents the similarity of facial features, so that the target face can be identified as quickly as possible, without waiting for all network iteration calculations to complete the recognition results. And after the angle of the target face and the illumination change, the correlation between the recognition results can be well obtained. In the actual application scenario, we use a multi-camera real-time monitoring system to perform face matching and recognition using successive frames acquired from different angles. The effectiveness of the method was verified by several real-time monitoring experiments, and good results were obtained.

• Mass Spectrometry Letters
• /
• 제15권1호
• /
• pp.49-53
• /
• 2024

Ceramide is a lipid in which sphingoid bases and fatty acids are linked by amide bonds. As a marker of skin disease in the human stratum corneum, its disease-causing and therapeutic effects have been partially confirmed, and it is therefore an important element in commercially available cosmetic formulations. However, structural diversity caused by differences in the chain length, number, and location of hydroxyl groups makes quality control difficult. In this study, a method was established to separate different ceramide species using reversed-phase LC-MS/MS and thus enable qualitative evaluation. Separation of four standards was achieved within a short retention time, and the accuracy and sensitivity of the method were demonstrated by the low limit of detection (LOD) calculated based on the calibration curve showing linearity, with R² > 0.994. After verification of reproducibility and reliability through intra- and inter-day analyses, the efficiency of the method was confirmed through analysis of commercial cosmetic raw materials.

• Bulletin of the Korean Chemical Society
• /
• 제26권3호
• /
• pp.440-446
• /
• 2005

A method is described for the determination of dibutyltin (DBT) in sediment by isotope dilution using liquid chromatography inductively-coupled plasma/mass spectrometry (LC-ICP/MS). To achieve the highest accuracy and precision, special attentions are paid in optimization and evaluation of overall processes of the analysis including extraction of analytes, characterization of the standards used for calibration and LC-ICP/MS conditions. An approach for characterization of natural abundance DBT standard has been developed by combining inductively-coupled plasma/optical emission spectrometry (ICP/OES) and LC-ICP/MS for the total Sn assay and the analysis of Sn species present as impurities, respectively. An excellent LC condition for separation of organotin species was found, which is suitable for simultaneous DBT and tributyltin (TBT) analysis as well as impurity analysis of DBT standards. Microwave extraction condition was also optimized for high efficiency while preventing species transformation. The present method determines the amount contents of DBT in sediments with expanded uncertainty of less than 5% and its result shows high degree of equivalence with reference values of an international inter-comparison and a certified reference material (CRM) within stated uncertainties.

• 한국수자원학회논문집
• /
• 제41권4호
• /
• pp.413-422
• /
• 2008

기상자료가 부족하거나 결측 지역의 기준 증발산량 산정을 위하여 Penman-Monteith (PM) 공식을 이용한 Hargreaves 공식의 매개변수 추정을 수행할 필요가 있다. 본 연구에서는 경기만 유역에 위치한 강화, 인천, 수원, 서산, 천안의 1997년$\sim$2004년 기상자료를 바탕으로 PM 공식을 이용하여 계산한 기준 증발산량(이하 ETo)을 이용하여 Hargreaves 공식의 매개변수를 추정하였으며, 추정된 매개변수를 이용하여 2005년$\sim$2006년의 PM 공식을 이용한 ETo 결과와 비교하여 검정을 수행하였다. 그 결과, 매개변수 조정 전 RMS 오차는 $14.63{\sim}23.30$ 정도로 파악되었으며, 모형의 검정에서도 $14.43{\sim}26.81$ 정도로 유사한 범위를 보이고 있다. 한편, Nash-Sutcliffe 일치계수는 $0.68{\sim}0.77$이며, 검정과정에서는 $0.43{\sim}0.85$로 대부분의 지역이 추정효율이 아주 떨어지는 것으로 나타났다. 반면, Hargreaves 계수를 조정한 경우, RMS 오차는 $5.23{\sim}11.75$ 정도로 파악되었으며, 모형의 검정에서도 $6.48{\sim}9.09$정도로 매개변수 조정전에 비하여 크게 감소하고 있음을 알 수 있으며, 한편, NSC는 $0.92{\sim}0.98$이며, 검정과정에서는 $0.93{\sim}0.97$로 대부분의 지역에서 추정효율이 크게 향상되는 것으로 나타났다.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2015년도 학술발표회
• /
• pp.217-217
• /
• 2015

The objective of this study was mainly to evaluate the water resources potential of Lake Tana Basin (LTB) by using Soil and Water Assessment Tool (SWAT). From SWAT simulation of LTB, about 5236 km2 area of LTB is gauged watershed and the remaining 9878 km2 area is ungauged watershed. For calibration of model parameters, four gauged stations were considered namely: Gilgel Abay, Gummera, Rib, and Megech. The SWAT-CUP built-in techniques, particle swarm optimization (PSO) and generalized likelihood uncertainty estimation (GLUE) method was used for calibration of model parameters and PSO method were selected for the study based on its performance results in four gauging stations. However the level of sensitivity of flow parameters differ from catchment to catchment, the curve number (CN2) has been found the most sensitive parameters in all gauged catchments. To facilitate the transfer of data from gauged catchments to ungauged catchments, clustering of hydrologic response units (HRUs) were done based on physical similarity measured between gauged and ungauged catchment attributes. From SWAT land use/ soil use/slope reclassification of LTB, a total of 142 HRUs were identified and these HRUs are clustered in to 39 similar hydrologic groups. In order to transfer the optimized model parameters from gauged to ungauged catchments based on these clustered hydrologic groups, this study evaluates three parameter transfer schemes: parameters transfer based on homogeneous regions (PT-I), parameter transfer based on global averaging (PT-II), and parameter transfer by considering Gilgel Abay catchment as a representative catchment (PT-III) since its model performance values are better than the other three gauged catchments. The performance of these parameter transfer approach was evaluated based on values of Nash-Sutcliffe efficiency (NSE) and coefficient of determination (R2). The computed NSE values was found to be 0.71, 0.58, and 0.31 for PT-I, PT-II and PT-III respectively and the computed R2 values was found to be 0.93, 0.82, and 0.95 for PT-I, PT-II, and PT-III respectively. Based on the performance evaluation criteria, PT-I were selected for modelling ungauged catchments by transferring optimized model parameters from gauged catchment. From the model result, yearly average stream flow for all homogeneous regions was found 29.54 m3/s, 112.92 m3/s, and 130.10 m3/s for time period (1989 - 2005) for region-I, region-II, and region-III respectively.