• 한국수자원학회논문집
• /
• 제53권2호
• /
• pp.97-105
• /
• 2020

최근 데이터 예측 방법으로 인공신경망(Artificial Neural Network, ANN)분야에 대한 관심이 높아졌으며, 그 중 시계열 데이터 예측에 특화된 LSTM(Long Short-Term Memory)모형은 수문 시계열자료의 예측방법으로도 활용되고 있다. 본 연구에서는 구글에서 제공하는 딥러닝 오픈소스 라이브러리인 텐서플로우(TensorFlow)를 활용하여 LSTM모형을 구축하고 금강 상류에 위치한 용담다목적댐의 유입량을 예측하였다. 분석 자료로는 WAMIS에서 제공하는 용담댐의 2006년부터 2018년까지의 시간당 유입량 자료를 사용하였으며, 예측된 유입량과 관측 유입량의 비교를 통하여 평균제곱오차(RMSE), 평균절대오차(MAE), 용적오차(VE)를 계산하고 모형의 학습변수에 따른 정확도를 평가하였다. 분석결과, 모든 모형이 고유량에서의 정확도가 낮은 것으로 나타났으며, 이와 같은 문제를 해결하기 위하여 용담댐 유역의 시간당 강수량 자료를 추가 학습 자료로 활용하여 분석한 결과, 고유량에 대한 예측의 정확도가 높아지는 것을 알 수 있었다.

• 대한토목학회논문집
• /
• 제38권2호
• /
• pp.215-226
• /
• 2018

본 연구에서는 도시지역 관망 내 수위 관측자료를 이용하여 도시지역 유출해석 및 관망해석의 정확도를 높이고자 한다. 이를 위해 도시유출해석의 주요 매개변수별 민감도 분석을 수행하고, 매개변수의 보정을 수행하였다. 매개변수의 민감도는 관의 조도계수, 불투수지역의 조도계수, 유역폭, 투수지역의 조도계수 순으로 나타났다. 민감도가 높은 4개의 매개변수를 이용하여 매개변수 고려 개수와 종류에 따라 6가지 시나리오를 구성하였으며, 자동보정기법인 PEST를 도시유출 모형인 SWMM과 연계 해석하여 분석하였다. 각 조건을 2013년 7월 21일 집중호우로 인하여 침수피해가 발생한 서초3, 4, 5, 역삼, 논현 배수분구에 적용하였다. 민감도 결과를 이용하여 시나리오별 분석을 실시하였을 때 SWMM 모형만을 이용하였을 때 보다 불확실성이 줄어든 결과를 보였다. 모의결과 RMSE는 최대 2.41cm가 감소하였으며, 상대첨두오차는 13.7%가 감소하였다. 민감도가 낮은 투수지역의 지표면 조도계수를 고려한 시나리오의 경우가 고려하지 않은 시나리오의 경우보다 정확도가 소폭 낮아졌으며 계산시간도 많이 소요되었다. 본 연구 결과 대상 유역에 대한 민감도 분석 후 민감도가 높은 매개변수만을 고려하여 시나리오를 구성하였을 때 보다 효율적인 모의가 될 수 있다는 것을 알 수 있었으며 향후 추가적인 강우사상과 침수해석을 연계할 시에는 도시지역의 침수예경보 연구에 기여할 수 있을 것으로 판단된다.

• 한국IT서비스학회지
• /
• 제14권1호
• /
• pp.195-215
• /
• 2015

Recently, as the number of smart-phone users has been rapidly increased, enterprise managers have a keen interest in business application of social media. Most previous studies have focused on perspective of the individual unit of analysis instead of enterprise level unit. The study is focused on the relationship between the enterprises' recognition and behavioral intention (and use) about social media application. The purpose of this study is to develop the model of small and medium enterprises' social media application, and to find the factors affecting their behavioral intention or use behavior. The moderating effects of four corporate characteristics on the relationship between the enterprises' recognition and behavioral intention are also examined. We surveyed 900 corporate staffs and received 203 responses. After questionnaires with unreliable responses had been excluded, 182 effective samples were used in the final analysis. The findings suggest that Performance Expectation, Social Influence, Facilitating Conditions significantly affect Behavioral Intention of social medea, and Behavioral Intention affects USE. Furthermore, some corporate characteristics have moderating effect on the relationship between recognition of social media and Behavioral Intention.

• 한국생산제조학회지
• /
• 제6권1호
• /
• pp.7-17
• /
• 1997

During the past decade, there were many well-established theories for the adaptive control of linear systems, but there exists relatively little general theory for the adaptive control of nonlinear systems. Adaptive control technique is essential for providing a stable and robust performance for application of industrial robot control. Neural network computing methods provide one approach to the development of adaptive and learning behavior in robotic system for manufacturing. Computational neural networks have been demonstrated which exhibit capabilities for supervised learning, matching, and generalization for problems on an experimental scale. Supervised learning could improve the efficiency of training and development of robotic systems. In this paper, a new scheme of adaptive-neuro control system to implement real-time control of robot manipulator using digital signal processors is proposed. Digital signal processors, DSPs, are micro-processors that are developed particularly for fast numerical computations involving sums and products of variables. The proposed neuro control algorithm is one of learning a model based error back-propagation scheme using Lyapunov stability analysis method. The proposed adaptive-neuro control scheme is illustrated to be an efficient control scheme for implementation of real-time control for SCARA robot with four-axes by experiment.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2002년도 Proceedings of International Symposium on Remote Sensing
• /
• pp.343-346
• /
• 2002

Toward more accurate determination of the water cycle in association with climate variability and change as well as baseline data on the impacts of this variability on water resources, the Coordinated Enhanced Observing Period (CEOP) was launched on July 1,2001. The preliminary data period, EOP-1, was implemented from July to September in 2001. The first annual enhanced observing period, EOP-3, is going to start on October 1,2002. CEOP is seeking to achieve a database of common measurements from both in situ and satellite remote sensing, model output, and four-dimensional data analyses (4DDA; including global and regional reanalyses) for a specified period. In this context a number of carefully selected reference stations are linked closely with the existing network of observing sites involved in the GEWEX Continental Scale Experiments, which are distributed across the world. The initial step of CEOP is to develop a pilot global hydro-climatological dataset with global consistency under the climate variability that can be used to help validate satellite hydrology products and evaluate, develop and eventually predict water and energy cycle processes in global and regional models. Based on the dataset, we will address the studies on the inter-comparison and inter-connectivity of the monsoon systems and regional water and energy budget, and a path to down-scaling from the global climate to local water resources, as the second step.

• 한국농공학회논문집
• /
• 제52권5호
• /
• pp.1-9
• /
• 2010

Evapotranspiration (ET) is one of the basic components of the hydrologic cycle and is essential for estimating irrigation water requirements. In this study, artificial neural network (ANN) models for reference crop evapotranspiration ($ET_0$) estimation were developed on a monthly basis (May~October). The models were trained and tested for Suwon, Korea. Four climate factors, daily maximum temperature ($T_{max}$), daily minimum temperature ($T_{min}$), rainfall (R), and solar radiation (S) were used as the input parameters of the models. The target values of the models were calculated using Food and Agriculture Organization (FAO) Penman-Monteith equation. Future climate data were generated using LARS-WG (Long Ashton Research Station-Weather Generator), stochastic weather generator, based on HadCM3 (Hadley Centre Coupled Model, ver.3) A1B scenario. The evapotranspirations were 549.7 mm/yr in baseline period (1973-2008), 558.1 mm/yr in 2011-2030, 593.0 mm/yr in 2046-2065, and 641.1 mm/yr in 2080-2099. The results showed that the ANN models achieved good performances in estimating future reference crop evapotranspiration.

• 한국산학기술학회논문지
• /
• 제1권2호
• /
• pp.69-74
• /
• 2000

중소기업의 정보화 영역은 생산정보화, 경영관리자동화, 네트워크화의 3가지 요소로 구분된다. 본 논문은 경영정보시스템을 이용하는 경영관리자동화를 주로 다룬다. 기능적으로 보면 생산, 판매, 인사, 회계 등 4개의 분야로 크게 나누어지나 대부분의 중소기업은 생산과 판매에 더 많은 주안점을 둔다. 따라서 이 두 개의 핵심 기능을 중심으로 객체지향방법론에 기반하여 통합된 정보시스템이 구축된다. 본 논문이 제안하는 단계별모델의 중요한 하나의 단계인 이 통합시스템은 단순화와 집중화의 원리를 수용했을 뿐만 아니라 객체지향패러다임을 이용하여 모듈화 및 친숙화를 구현하였다.

• Journal of Multimedia Information System
• /
• 제4권4호
• /
• pp.233-238
• /
• 2017

Automatic identification of disease in plants from their leaves is one of the most challenging task to researchers. Diseases among plants degrade their performance and results into a huge reduction of agricultural products. Therefore, early and accurate diagnosis of such disease is of the utmost importance. The advancement in deep Convolutional Neural Network (CNN) has change the way of processing images as compared to traditional image processing techniques. Deep learning architectures are composed of multiple processing layers that learn the representations of data with multiple levels of abstraction. Therefore, proved highly effective in comparison to many state-of-the-art works. In this paper, we present a plant disease identification methodology from their leaves using deep CNNs. For this, we have adopted GoogLeNet that is considered a powerful architecture of deep learning to identify the disease types. Transfer learning has been used to fine tune the pre-trained model. An accuracy of 85.04% has been recorded in the identification of four disease class in Apple plant leaves. Finally, a comparison with other models has been performed to show the effectiveness of the approach.

• STI Policy Review
• /
• 제6권1호
• /
• pp.24-35
• /
• 2015

This research constructs a data set regarding competition policy through a comprehensive review of previous studies, and performs a meta-analysis to quantitatively assess the price effects of deregulation. A structural econometric model is used to eliminate possible biases from heterogeneity of the studies,such as in publication types and measurement methods. Four types of regulations that deter competition are characterized and three groups of industries are made for drawing practical implications. We fnd that deregulation to promote competition reduces prices by 0.23% and that these estimated price effects are more stable when we control for the publication types and measurement ways. Easing regulations that restrict consumers' choice is shown to be most effcient in promoting competition, lowering prices by 0.7%. This is followed by eliminating the limitation in the number of frms in the industry, with 0.2% price reduction. Overall, the network and service industries are shown to be more responsive to deregulation than the R&D industry. These results could shed light on policy implementation when a pro-competition policy is called for due to restrictive regulations in the corresponding industries.

• JSTS:Journal of Semiconductor Technology and Science
• /
• 제14권2호
• /
• pp.252-261
• /
• 2014

A Bayesian network (BN) based fault diagnosis framework for semiconductor etching equipment is presented. Suggested framework contains data preprocessing, data synchronization, time series modeling, and BN inference, and the established BNs show the cause and effect relationship in the equipment module level. Statistically significant state variable identification (SVID) data of etch equipment are preselected using principal component analysis (PCA) and derivative dynamic time warping (DDTW) is employed for data synchronization. Elman's recurrent neural networks (ERNNs) for individual SVID parameters are constructed, and the predicted errors of ERNNs are then used for assigning prior conditional probability in BN inference of the fault diagnosis. For the demonstration of the proposed methodology, 300 mm etch equipment model is reconstructed in subsystem levels, and several fault diagnosis scenarios are considered. BNs for the equipment fault diagnosis consists of three layers of nodes, such as root cause (RC), module (M), and data parameter (DP), and the constructed BN illustrates how the observed fault is related with possible root causes. Four out of five different types of fault scenarios are successfully diagnosed with the proposed inference methodology.