• Proceedings of the Korea Water Resources Association Conference
• /
• 2018.05a
• /
• pp.18-18
• /
• 2018

상수관망 시스템(Water Distribution System, WDS)은 이용자에게 양질의 상수도를 공급하기 위해 구축된 사회기반시설물로써, 정수된 물이 사용처에 도달하기까지 송수과정에서 발생 가능한 수질저하를 고려해야 한다. 일반적으로 정수장에서 염소처리를 한 후, 도달시간을 고려한 시스템 내 잔류 염소농도를 유지함으로써 수질저하를 예방한다. 여기서 상수도 내 잔류 염소농도는 미생물 번식 및 관내 부식물 등 다양한 생물 화학적 오염을 효과적으로 예방하는 반면, 과다할 경우 이용자의 음용성을 저해할 수 있어 시스템 전반에 걸쳐 염소농도의 적절한 관리가 요구된다. 특히, 상수관망에서는 공급경로 및 공급량에 따라 각 수요처의 도달 염소농도가 다르게 분포할 수 있으므로, 시설운영자는 균등하고 적절한 염소농도를 유지하기 위해 추가적인 염소 재투입시설을 설치하여 함께 관리하고 있다. 이 때, 염소투입 시설의 운영계획은 EPANET과 같은 상수관망 해석모형의 수질모의를 바탕으로 수립된다. 그러나 일반적으로 수질모의는 수리해석과는 달리 긴 시간이 소요되는 단점이 존재한다. 본 연구에서는 이러한 단점을 개선하기 위해, 특정 네트워크의 수질모의 결과를 학습시킨 인공신경망(ANN) 모형을 구축하고 이를 이용하여 상수관망 수질모의 계산시간을 단축하고자 하였다. 여기서 ANN모형의 학습은 EPANET을 통해 미리 선정된 다양한 염소 투입지점의 염소 투입농도와 용수 공급량 자료, 그리고 주요 관측지점에서 측정된 염소농도자료를 이용하였다. 학습된 ANN모형을 EPANET 수질모의 결과와 비교 및 검증을 실시한 결과, 사전에 소요된 학습시간을 제외하면 수질모의 소요시간 측면에서 큰 개선효과를 보였으며, 대표지점에서의 수질모의 결과가 유사하였다. 추가적으로, 본 연구에서는 학습된 ANN모형과 최적화 알고리즘인 GA(Genitic Algorithm)를 연계하여 상수관망에서의 염소 재투입 스케줄링을 최적화하는 프로그램을 개발함으로써, 안전하고 경제적인 상수관망의 수질운영에 기여하고자 하였다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.121-121
• /
• 2019

특정 지역에 집중적으로 비가 내리는 현상인 국지성호우가 빈번히 발생함에 따라 하천 주변 사회기반시설의 침수 위험성이 증가하고 있다. 침수 위험성 판단 여부는 주로 수위정보를 이용하며 수위 예측은 대부분 수치모형을 이용한다. 본 연구에서는 빅데이터 기반의 RNN(Recurrent Neural Networks)기법 알고리즘을 활용하여 수위를 예측하였다. 연구대상지는 조위의 영향을 많이 받는 한강 전역을 대상으로 하였다. 2008년~2018년(10개년)의 실제 침수 피해 실적을 조사한 결과 잠수교, 한강대교, 청담대교 등에서 침수 피해 발생률이 높게 나타났고 SNS(Social Network Services)와 같은 비정형화 자료에서는 청담대교가 가장 많이 태그(Tag)되어 청담대교를 연구범위로 설정하였다. 본 연구에서는 Python에서 제공하는 Tensor flow Library를 이용하여 수위예측 알고리즘을 적용하였다. 데이터는 정형화 데이터와 비정형 데이터를 사용하였으며 정형화 데이터는 한강홍수 통제소나 기상청에서 제공하는 최근 10년간의 (2008~2018) 수위 및 강우량 자료를 수집하였다. 비정형화 데이터는 SNS를 이용하여 민간 정보를 수집하여 정형화된 자료와 함께 전체자료를 구축하였다. 민감도 분석을 통하여 모델의 은닉층(5), 학습률(0.02) 및 반복횟수(100)의 최적값을 설정하였고, 24시간 동안의 데이터를 이용하여 3시간 후의 수위를 예측하였다. 2008년~ 2017년 까지의 데이터는 학습 데이터로 사용하였으며 2018년의 수위를 예측 및 평가하였다. 2018년의 관측수위 자료와 비교한 결과 90% 이상의 데이터가 10% 이내의 오차를 나타내었으며, 첨두수위도 비교적 정확하게 예측되는 것을 확인하였다. 향후 수위와 강우량뿐만 아니라 다양한 인자들도 고려한다면 보다 신속하고 정확한 예측 정보를 얻을 수 있을 것으로 기대된다.

• Journal of Broadcast Engineering
• /
• v.24 no.5
• /
• pp.870-878
• /
• 2019

In this paper, we propose an Attention-based BLSTM for predicting the video compression standard of a video. Recently, in NLP, many researches have been studied to predict the next word of sentences, classify and translate sentences by their semantics using the structure of RNN, and they were commercialized as chatbots, AI speakers and translator applications, etc. LSTM is designed to solve the gradient vanishing problem in RNN, and is used in NLP. The proposed algorithm makes video compression standard prediction possible by applying BLSTM and Attention algorithm which focuses on the most important word in a sentence to a bitstream of a video, not an sentence of a natural language.

• Journal of the Korean Society of Propulsion Engineers
• /
• v.17 no.2
• /
• pp.71-83
• /
• 2013

Recently, the advanced condition monitoring methods such as the model-based method and the artificial intelligent method have been applied to maximize the availability as well as to minimize the maintenance cost of the aircraft gas turbines. Among them the non-linear GPA(Gas Path Analysis) method and the GA(Genetic Algorithms) have lots of advantages to diagnose the engines compared to other advanced condition monitoring methods such as the linear GPA, fuzzy logic and neural networks. Therefore this work applies both the non-linear GPA and the GA to diagnose AE3007 turbofan engine for an aircraft, and in case of having sensor noise and bias it is confirmed that the GA is better than the GPA through the comparison of two methods.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.466-472
• /
• 2021

Speaker Change Detection (SCD) refers to finding the moment when the main speaker changes from one person to the next in a speech conversation. In speaker change detection, difficulties arise due to overlapping speakers, inaccuracy in the information labeling, and data imbalance. To solve these problems, TIMIT corpus widely used in speech recognition have been concatenated artificially to obtain a sufficient amount of training data, and the detection of changing speaker has performed after identifying overlapping speakers. In this paper, we propose an speaker change detection system that considers the speaker overlapping. We evaluated and verified the performance using various approaches. As a result, a detection system similar to the X-Vector structure was proposed to remove the speaker overlapping region, while the Bi-LSTM method was selected to model the speaker change system. The experimental results show a relative performance improvement of 4.6 % and 13.8 % respectively, compared to the baseline system. Additionally, we determined that a robust speaker change detection system can be built by conducting related studies based on the experimental results, taking into consideration text and speaker information.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.12
• /
• pp.45-52
• /
• 2021

In this paper, we propose a prediction model for osteopenia and osteoporosis based on a convolutional neural network(CNN) using computed tomography(CT) images. In a single CT image, CNN had a limitation in utilizing important local features for diagnosis. So we propose a compound model which has two identical structures. As an input, two different texture images are used, which are converted from a single normalized CT image. The two networks train different information by using dissimilarity loss function. As a result, our model trains various features in a single CT image which includes important local features, then we ensemble them to improve the accuracy of predicting osteopenia and osteoporosis. In experiment results, our method shows an accuracy of 77.11% and the feature visualize of this model is confirmed by using Grad-CAM.

• Journal of Internet Computing and Services
• /
• v.20 no.3
• /
• pp.33-41
• /
• 2019

In this paper, we propose a method to learn the redesigned class with background and object for semantic segmentation of indoor scene image. Semantic image segmentation is a technique that divides meaningful parts of an image, such as walls and beds, into pixels. Previous work of semantic image segmentation has proposed methods of learning various object classes of images through neural networks, and it has been pointed out that there is insufficient accuracy compared to long learning time. However, in the problem of separating objects and backgrounds, there is no need to learn various object classes. So we concentrate on separating objects and backgrounds, and propose method to learn after class simplification. The accuracy of the proposed learning method is about 5 ~ 12% higher than the existing methods. In addition, the learning time is reduced by about 14 ~ 60 minutes when the class is configured differently In the same environment, and it shows that it is possible to efficiently learn about the problem of separating the object and the background.

• Journal of the Korean Society for information Management
• /
• v.36 no.3
• /
• pp.131-148
• /
• 2019

This study constructed a cognitive model of information processing to understand the topic of a sound material and its characteristics. It then proposed methods to generate sound summaries, by incorporating anterior-posterior N400/P600 components of event-related potential (ERP) response, into the language representation of the cognitive model of information processing. For this end, research hypotheses were established and verified them through ERP experiments, finding that P600 is crucial in screening topic-relevant shots from topic-irrelevant shots. The results of this study can be applied to the design of classification algorithm, which can then be used to generate the content-based metadata, such as generic or personalized sound summaries and video skims.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.19 no.1
• /
• pp.29-35
• /
• 2019

As the fourth industrial revolution is emerging, many companies are increasingly interested in smart factories and the importance of sensors is being emphasized. In the case that sensors for collecting sensing data fail, the plant could not be optimized and further it could not be operated properly, which may incur a financial loss. For this purpose, it is necessary to diagnose the status of sensors to prevent sensor' fault. In the paper, we propose a scheme to diagnose digital-sensor' fault by analyzing the rising time and falling time of digital sensors through the LSTM(Long Short Term Memory) of Deep Learning RNN algorithm. Experimental results of the proposed scheme are compared with those of rule-based fault diagnosis algorithm in terms of AUC(Area Under the Curve) of accuracy and ROC(Receiver Operating Characteristic) curve. Experimental results show that the proposed system has better and more stable performance than the rule-based fault diagnosis algorithm.

• Korean Journal of Agricultural Science
• /
• v.46 no.2
• /
• pp.381-393
• /
• 2019

A smart farm is a system that combines information and communication technology (ICT), internet of things (IoT), and agricultural technology that enable a farm to operate with minimal labor and to automatically control of a greenhouse environment. Machine learning based on recently data-driven techniques has emerged with big data technologies and high-performance computing to create opportunities to quantify data intensive processes in agricultural operational environments. This paper presents research on the application of machine learning technology to diagnose the growth status of crops and predicting the harvest time of strawberries in a greenhouse according to image processing techniques. To classify the growth stages of the strawberries, we used object inference and detection with machine learning model based on deep learning neural networks and TensorFlow. The classification accuracy was compared based on the training data volume and training epoch. As a result, it was able to classify with an accuracy of over 90% with 200 training images and 8,000 training steps. The detection and classification of the strawberry maturities could be identified with an accuracy of over 90% at the mature and over mature stages of the strawberries. Concurrently, the experimental results are promising, and they show that this approach can be applied to develop a machine learning model for predicting the strawberry harvesting time and can be used to provide key decision support information to both farmers and policy makers about optimal harvest times and harvest planning.