• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.340-340
• /
• 2019

지구온난화로 유발되는 기후변동성이 증가함에 따라서 정확한 수문인자의 예측은 전 세계적으로 주요 관심사항이 되고 있다. 최근에는 고성능 컴퓨터 자원의 증가로 수문기상학 연구에서 동일한 학습량에 비하여 정확도의 향상이 뚜렷한 기계학습 구조를 활용하여 위성영상 기반의 대기예측, 태풍위치 추적 및 강수량 예측 등의 연구가 활발하게 진행되고 있다. 본 연구에는 기계학습 중 시계열 분석에 널리 활용되고 있는 순환신경망(Recurrent Neural Network, RNN) 기법의 대표적인 LSTM(Long Short-Term Memory) 네트워크를 이용하여 수문인자를 예측하였다. LSTM 네트워크는 가중치 및 메모리 요소에 대한 추가정보를 셀 상태에 저장하고 시계열의 길이 조정하여 모형의 탄력적 활용이 가능하다. LSTM 네트워크를 이용한 다양한 수문인자 예측결과 RMSE의 개선을 확인하였다. 따라서 본 연구를 통하여 개발된 기계학습을 통한 수문인자 예측기술은 권역별 수계별 홍수 및 가뭄대응 계획을 능동적으로 수립하는데 활용될 것으로 판단된다. 향후 연구에서는 LSTM의 입력영역을 Bayesian 추론기법을 활용하여 구성함으로 학습과정의 불확실성을 정량적으로 제어하고자 한다.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.33 no.6
• /
• pp.238-245
• /
• 2021

In this study, we carried out case study to predict dissolved oxygen (DO) concentration of Nakdong river estuary with LSTM model. we aimed to figure out a optimal model condition and appropriate predictor for prediction in dissolved oxygen concentration with model parameter and predictor as cases. Model parameter case study results showed that Epoch = 300 and Sequence length = 1 showed higher accuracy than other conditions. In predictor case study, it was highest accuracy where DO and Temperature were used as a predictor, it was caused by high correlation between DO concentration and Temperature. From above results, we figured out an appropriate model condition and predictor for prediction in DO concentration of Nakdong river estuary.

• ETRI Journal
• /
• v.44 no.4
• /
• pp.672-685
• /
• 2022

This article proposes a subject-independent application of brain-computer interfacing (BCI). A 32-channel Electroencephalography (EEG) device is used to measure imagined speech (SI) of four words (sos, stop, medicine, washroom) and one phrase (come-here) across 13 subjects. A deep long short-term memory (LSTM) network has been adopted to recognize the above signals in seven EEG frequency bands individually in nine major regions of the brain. The results show a maximum accuracy of 73.56% and a network prediction time (NPT) of 0.14 s which are superior to other state-of-the-art techniques in the literature. Our analysis reveals that the alpha band can recognize SI better than other EEG frequencies. To reinforce our findings, the above work has been compared by models based on the gated recurrent unit (GRU), convolutional neural network (CNN), and six conventional classifiers. The results show that the LSTM model has 46.86% more average accuracy in the alpha band and 74.54% less average NPT than CNN. The maximum accuracy of GRU was 8.34% less than the LSTM network. Deep networks performed better than traditional classifiers.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.373-373
• /
• 2021

Rainfall forecasting is an important issue that is applied in many areas, such as agriculture, flood warning, and water resources management. In this context, this study proposed a statistical and machine learning-based forecasting model for monthly rainfall. The Bayesian Gaussian process was chosen to optimize the hyperparameters of the Stacked Long Short-term memory (SLSTM) model. The proposed SLSTM model was applied for predicting monthly precipitation of Seoul station, South Korea. Data were retrieved from the Korea Meteorological Administration (KMA) in the period between 1960 and 2019. Four schemes were examined in this study: (i) prediction with only rainfall; (ii) with deseasonalized rainfall; (iii) with rainfall and minimum temperature; (iv) with deseasonalized rainfall and minimum temperature. The error of predicted rainfall based on the root mean squared error (RMSE), 16-17 mm, is relatively small compared with the average monthly rainfall at Seoul station is 117mm. The results showed scheme (iv) gives the best prediction result. Therefore, this approach is more straightforward than the hydrological and hydraulic models, which request much more input data. The result indicated that a deep learning network could be applied successfully in the hydrology field. Overall, the proposed method is promising, given a good solution for rainfall prediction.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.10a
• /
• pp.391-393
• /
• 2021

In this paper, we introduce a long short-term memory (LSTM)-based fall detection system using TensorFlow that can detect falls occurring in the elderly in daily living. 3-axis accelerometer data are aggregated for fall detection, and then three types of parameter are calculated. 4 types of activity of daily living (ADL) and 3 types of fall situation patterns are classified. The parameterized data applied to LSTM. Learning proceeds until the Loss value becomes 0.5 or less. The results are calculated for each parameter θ, SVM, and GSVM. The best result was GSVM, which showed Sensitivity 98.75%, Specificity 99.68%, and Accuracy 99.28%.

• Smart Structures and Systems
• /
• v.31 no.1
• /
• pp.45-56
• /
• 2023

Vibration-based structural health monitoring is used to ensure the safety of structures by installing sensors in structures. The peak picking method, one of the applications of vibration-based structural health monitoring, is a method that analyze the dynamic characteristics of a structure using the peaks of the frequency response function. However, the results may vary depending on the person predicting the peak point; further, the method does not predict the exact peak point in the presence of noise. To overcome the limitations of the existing peak picking methods, this study proposes a new method to automate the modal analysis process by utilizing long short-term memory, a type of recurrent neural network. The method proposed in this study uses the time series data of the frequency response function directly as the input of the LSTM network. In addition, the proposed method improved the accuracy by using the phase as well as amplitude information of the frequency response function. Simulation experiments and lab-scale model experiments are performed to verify the performance of the LSTM network developed in this study. The result reported a modal assurance criterion of 0.8107, and it is expected that the dynamic characteristics of a civil structure can be predicted with high accuracy using data without experts.

• Smart Structures and Systems
• /
• v.32 no.6
• /
• pp.359-369
• /
• 2023

Servo-motor driven uniaxial shake tables have been widely used for education and research purposes in earthquake engineering. These shake tables are mostly displacement-controlled by a digital proportional-integral-derivative (PID) controller; however, accurate reproduction of acceleration time histories is not guaranteed. In this study, a control strategy is proposed and verified for uniaxial shake tables driven by a servo-motor. This strategy incorporates a deep-learning algorithm named Long Short-Term Memory (LSTM) network into a displacement PID feedback controller. The LSTM controller is trained by using a large number of experimental data of a self-made servo-motor driven uniaxial shake table. After the training is completed, the LSTM controller is implemented for directly generating the command voltage for the servo motor to drive the shake table. Meanwhile, a displacement PID controller is tuned and implemented close to the LSTM controller to prevent the shake table from permanent drift. The control strategy is named the LSTM-PID control scheme. Experimental results demonstrate that the proposed LSTM-PID improves the acceleration tracking performance of the uniaxial shake table for both bare condition and loaded condition with a slender specimen.

• Journal of the Korean Society of Systems Engineering
• /
• v.19 no.2
• /
• pp.18-31
• /
• 2023

Exploring artificial intelligence and machine learning for nuclear safety has witnessed increased interest in recent years. To contribute to this area of research, a machine learning model capable of accurately predicting nuclear power plant response with minimal computational cost is proposed. To develop a robust machine learning model, the Best Estimate Plus Uncertainty (BEPU) approach was used to generate a database to train three models and select the best of the three. The BEPU analysis was performed by coupling Dakota platform with the best estimate thermal hydraulics code RELAP/SCDAPSIM/MOD 3.4. The Code Scaling Applicability and Uncertainty approach was adopted, along with Wilks' theorem to obtain a statistically representative sample that satisfies the USNRC 95/95 rule with 95% probability and 95% confidence level. The generated database was used to train three models based on Recurrent Neural Networks; specifically, Long Short-Term Memory, Gated Recurrent Unit, and a hybrid model with Long Short-Term Memory coupled to Convolutional Neural Network. In this paper, the System Engineering approach was utilized to identify requirements, stakeholders, and functional and physical architecture to develop this project and ensure success in verification and validation activities necessary to ensure the efficient development of ML meta-models capable of predicting of the nuclear power plant response.

• Journal of the Korean Solar Energy Society
• /
• v.39 no.5
• /
• pp.41-52
• /
• 2019

The purpose of the work is to develop a simple solar irradiance prediction model using a deep learning method, the LSTM (long term short term memory). Other than existing prediction models, the proposed one uses only the cloudiness among the information forecasted from the national meterological forecast center. The future cloudiness is generally announced with four categories and for three-hour intervals. In this work, a daily irradiance pattern is used as an input vector to the LSTM together with that cloudiness information. The proposed model showed an error of 5% for learning and 30% for prediction. This level of error has lower influence on the load prediction in typical building cases.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.8
• /
• pp.307-324
• /
• 2018

Recently, recurrent neural networks have been attracting attention in solving prediction problem of sequential data through structure considering time dependency. However, as the time step of sequential data increases, the problem of the gradient vanishing is occurred. Long short-term memory models have been proposed to solve this problem, but there is a limit to storing a lot of data and preserving it for a long time. Therefore, research on memory-augmented neural network (MANN), which is a learning model using recurrent neural networks and memory elements, has been actively conducted. In this paper, we describe the structure and characteristics of MANN models that emerged as a hot topic in deep learning field and present the latest techniques and future research that utilize MANN.