Dynamic responses of nuclear power plant structure subjected to earthquake loads should be carefully investigated for safety. Because nuclear power plant structure are usually constructed by material of reinforced concrete, the aging deterioration of R.C. have no small effect on structural behavior of nuclear power plant structure. Therefore, aging deterioration of R.C. nuclear power plant structure should be considered for exact prediction of seismic responses of the structure. In this study, a machine learning model for seismic response prediction of nuclear power plant structure was developed by considering aging deterioration. The OPR-1000 was selected as an example structure for numerical simulation. The OPR-1000 was originally designated as the Korean Standard Nuclear Power Plant (KSNP), and was re-designated as the OPR-1000 in 2005 for foreign sales. 500 artificial ground motions were generated based on site characteristics of Korea. Elastic modulus, damping ratio, poisson’s ratio and density were selected to consider material property variation due to aging deterioration. Six machine learning algorithms such as, Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Artificial Neural Networks (ANN), eXtreme Gradient Boosting (XGBoost), were used t o construct seispic response prediction model. 13 intensity measures and 4 material properties were used input parameters of the training database. Performance evaluation was performed using metrics like root mean square error, mean square error, mean absolute error, and coefficient of determination. The optimization of hyperparameters was achieved through k-fold cross-validation and grid search techniques. The analysis results show that neural networks present good prediction performance considering aging deterioration.

The ocean is linked to long-term climate variability, but there are very few methods to assess the short-term performance of forecast models. This study analyzes the short-term prediction performance regarding ocean temperature and salinity of the Global Seasonal prediction system version 5 (GloSea5). GloSea5 is a historical climate re-creation (2001-2010) performed on the 1st, 9th, 17th, and 25th of each month. It comprises three ensembles. High-resolution hindcasts from the three ensembles were compared with the Array for Real-Time Geostrophic Oceanography (ARGO) float data for the period 2001-2010. The horizontal position was preprocessed to match the ARGO float data and the vertical layer to the GloSea5 data. The root mean square error (RMSE), Brier Score (BS), and Brier Skill Score (BSS) were calculated for short-term forecast periods with a lead-time of 10 days. The results show that sea surface temperature (SST) has a large RMSE in the western boundary current region in Pacific and Atlantic Oceans and Antarctic Circumpolar Current region, and sea surface salinity (SSS) has significant errors in the tropics with high precipitation, with both variables having the largest errors in the Atlantic. SST and SSS had larger errors during the fall for the NINO3.4 region and during the summer for the East Sea. Computing the BS and BSS for ocean temperature and salinity in the NINO3.4 region revealed that forecast skill decreases with increasing lead-time for SST, but not for SSS. The preprocessing of GloSea5 forecasts to match the ARGO float data applied in this study, and the evaluation methods for forecast models using the BS and BSS, could be applied to evaluate other forecast models and/or variables.

고성능 콘크리트(HPC) 압축강도는 추가적인 시멘트질 재료의 사용으로 인해 예측하기 어렵고, 개선된 예측 모델의 개발이 필수적 이다. 따라서, 본 연구의 목적은 배깅과 스태킹을 결합한 앙상블 기법을 사용하여 HPC 압축강도 예측 모델을 개발하는 것이다. 이 논 문의 핵심적 기여는 기존 앙상블 기법인 배깅과 스태킹을 통합하여 새로운 앙상블 기법을 제시하고, 단일 기계학습 모델의 문제점을 해결하여 모델 예측 성능을 높이고자 한다. 단일 기계학습법으로 비선형 회귀분석, 서포트 벡터 머신, 인공신경망, 가우시안 프로세스 회귀를 사용하고, 앙상블 기법으로 배깅, 스태킹을 이용하였다. 결과적으로 본 연구에서 제안된 모델이 단일 기계학습 모델, 배깅 및 스태킹 모델보다 높은 정확도를 보였다. 이는 대표적인 4가지 성능 지표 비교를 통해 확인하였고, 제안된 방법의 유효성을 검증하였다.

전산유체역학을 사용하는 일반적인 선박의 저항성능 평가는 많은 시간과 비용이 필요하며, 이를 줄이기 위한 다양한 방법이 연구되고 있다. 선박의 주요 치수나 단면을 이용하는 기존의 방법들은 선형에 크게 좌우되는 저항성능을 추정하는데 한계가 있다. 본 논 문에서는 선형 격자의 기하학적 정보를 입력으로 선체 표면의 저항성능을 빠르게 추정할 수 있는 심층신경망 모델을 제안한다. Perceiver IO 기반의 제안하는 심층신경망 모델은 시간 단계별로 계산이 필요한 전산유체역학 기법과 달리 바로 저항성능 추정이 가능하며, 저속비 대선의 일종인 50K 탱커 선박을 대상으로 한 데이터집합에서 평균 1% 미만의 오차로 저항성능을 추정하는 결과를 보인다.

Screw jet equipment has been developed based on the existing accumulated experimental indicators in the semiconductor industry, and for specific performance development, it is necessary to visually check a process in which a high viscosity solution is discharged to a nozzle through a screw. Since the transparency of the exterior is not guaranteed after design and production due to the characteristics of the equipment, simulation must be performed to confirm the performance data according to the internal shape. Therefore, in this study, the screw jet equipment was simulated using the moving particle system, and through this, all processes of the screw jet internal solution flow were visually checked and computerized data capable of predicting the performance of the equipment was secured.

PURPOSES : In this study, surface distress (SD), rutting depth (RD), and international roughness index (IRI) prediction models are developed based on the zones of Incheon and road classes using regression analysis. Regression analysis is conducted based on a correlation analysis between the pavement performance and influencing factors. METHODS : First, Incheon was categorized by zone such as industrial, port, and residential areas, and the roads were categorized into major and sub-major roads. A weather station triangle network for Incheon was developed using the Delaunay triangulation based on the position of the weather station to match the road sections in Incheon and environmental factors. The influencing factors of the road sections were matched Based on the developed triangular network. Meanwhile, based on the matched influencing factors, a model of the current performance of the road pavement in Incheon was developed by performing multiple regression analysis. Sensitivity analysis was conducted using the developed model to determine the influencing factor that affected each performance factor the most significantly. RESULTS : For the SD model, frost days, daily temperature range, rainy days, tropical nights, and minimum temperatures are used as independent variables. Meanwhile, the truck ratio, freeze–thaw days, precipitation days, annual temperature range, and average temperatures are used for the RD model. For the IRI model, the maximum temperature, freeze–thaw days, average temperature, annual precipitation, and wet days are used. Results from the sensitivity analysis show that frost days for the SD model, precipitation days and freeze–thaw days for the RD model, and wet days for the IRI model impose the most significant effects. CONCLUSIONS : We developed a road pavement performance prediction model using multiple regression analysis based on zones in Incheon and road classes. The developed model allows the influencing factors and circumstances to be predicted, thus facilitating road management.

국내 주요 사회기반시설의 70% 이상이 철근콘크리트 구조물로 구성되어 있다. 최근 다양한 사회적ㆍ환경적 변화로 인한 내하력 저하 및 노후화 진행이 발생됨에 따라 섬유강화 복합소재(FRP)를 활용한 유지보수 수요 및 비용이 급격히 증가되 고 있다. 이에 따라 보다 경제적이고 효율적으로 FRP 보강재를 활용함에 있어서 성능을 예측할 수 있는 방법이 요구된다. 본 연구에서는 CFRPㆍBFRP 복합재료를 실험 대상으로 선정하고 성능을 결정하는 주요 인자인 섬유/수지 함침률을 54.3%, 43.9%, 39% 3가지로 분류하여 성능을 평가하고 이를 활용하여 FRP의 성능을 예측할 수 있는 모델식을 개발하고자 하였다. 매개변수에 따른 성능평가 결과, 두 섬유 모두 함침률이 낮아질수록 재료성능 또한 감소되는 것이 확인되었으며, 특히 BFRP의 경우 39%의 함침률에서 감소폭이 CFRP 대비 더 큰 것으로 나타났다. 실험 결과와 기존의 예측 모델식과의 성능 비교를 통해 약 15%의 오 차가 나타나는 것을 확인하였으며, 이에 따른 보정계수를 산정하여 예측 모델식을 재정립하였다.

During decommissioning of a nuclear power plant, a large amount of radioactive waste is produced, and it is known to cost more than 300 billion won to dispose the waste. To reduce the disposal cost, it is essential to minimize the number of radioactive waste drums, which can be achieved by detecting and removing hotspot contaminations in the radioactive waste drums. Therefore, a Compton CT system for radioactive waste monitoring is under development, which provides the images of both the internal structure of the drum and the radioactive hotspot(s) in the drum. Based on the acquired information, the activity of hotspots can be estimated. The performance of the system is affected by various geometry factors. Therefore, it is essential to determine optimal configuration by evaluating the effects of the factors on the performance of the system. In the present study, we determined the optimum value of the factors and then predicted the performance of the optimized system by using a simulator based on the Geant4 Monte Carlo simulation. For optimization, the factors were evaluated in terms of structural similarity index measure (SSIM) and measurement time. The considered factors were the activity of the CT source, source to object distance (SOD), object to detector distance (ODD), and projection angle. The simulation result showed that the activities of the CT sources were determined as 23 mCi for 137Cs and 9.6 mCi for 60Co. The optimal SOD and ODD were 180 cm and 40 cm, respectively. The optimal projection angle was evaluated as 4° since it achieves the SSIM of 0.95 faster than other projection angles. With the optimized parameters, the performance of the system was evaluated using the IAEA gamma CT standard phantom containing a hotspot of 137Cs (7.02 μCi). The Compton image was reconstructed using the back-projection algorithm, and the CT image was reconstructed using the filtered back-projection algorithm. The result showed that the location of the hotspot in the Compton image was well identified at the true position. The acquired CT image also well represented the internal structure of the phantom, and the estimated mean linear attenuation coefficient value (μ= 0.0789 cm−1) of the phantom was close to the true value (μ= 0.0752 cm−1). In addition, the hotspot activity estimated by combining the information of the Compton image and CT image was 8.06 μCi. Hence, it was found that the Compton CT system provides essential information for radioactive waste drums.

Recently, the importance of preventive maintenance has been emerging since failures in a complex system are automatically detected due to the development of artificial intelligence techniques and sensor technology. Therefore, prognostic and health management (PHM) is being actively studied, and prediction of the remaining useful life (RUL) of the system is being one of the most important tasks. A lot of researches has been conducted to predict the RUL. Deep learning models have been developed to improve prediction performance, but studies on identifying the importance of features are not carried out. It is very meaningful to extract and interpret features that affect failures while improving the predictive accuracy of RUL is important. In this paper, a total of six popular deep learning models were employed to predict the RUL, and identified important variables for each model through SHAP (Shapley Additive explanations) that one of the explainable artificial intelligence (XAI). Moreover, the fluctuations and trends of prediction performance according to the number of variables were identified. This paper can suggest the possibility of explainability of various deep learning models, and the application of XAI can be demonstrated. Also, through this proposed method, it is expected that the possibility of utilizing SHAP as a feature selection method.

Algal bloom is an ongoing issue in the management of freshwater systems for drinking water supply, and the chlorophyll-a concentration is commonly used to represent the status of algal bloom. Thus, the prediction of chlorophyll-a concentration is essential for the proper management of water quality. However, the chlorophyll-a concentration is affected by various water quality and environmental factors, so the prediction of its concentration is not an easy task. In recent years, many advanced machine learning algorithms have increasingly been used for the development of surrogate models to prediction the chlorophyll-a concentration in freshwater systems such as rivers or reservoirs. This study used a light gradient boosting machine(LightGBM), a gradient boosting decision tree algorithm, to develop an ensemble machine learning model to predict chlorophyll-a concentration. The field water quality data observed at Daecheong Lake, obtained from the real-time water information system in Korea, were used for the development of the model. The data include temperature, pH, electric conductivity, dissolved oxygen, total organic carbon, total nitrogen, total phosphorus, and chlorophyll-a. First, a LightGBM model was developed to predict the chlorophyll-a concentration by using the other seven items as independent input variables. Second, the time-lagged values of all the input variables were added as input variables to understand the effect of time lag of input variables on model performance. The time lag (i) ranges from 1 to 50 days. The model performance was evaluated using three indices, root mean squared error-observation standard deviation ration (RSR), Nash-Sutcliffe coefficient of efficiency (NSE) and mean absolute error (MAE). The model showed the best performance by adding a dataset with a one-day time lag (i=1) where RSR, NSE, and MAE were 0.359, 0.871 and 1.510, respectively. The improvement of model performance was observed when a dataset with a time lag up of about 15 days (i=15) was added.

본 논문은 FRCM 공법으로 보강된 철근콘크리트 보의 휨 성능을 예측하기 위한 해석적 연구결과를 제시한다. FRCM 공법으로 보강된 철근콘크리트 보의 휨 성능 예측을 위해 상용구조해석 프로그램인 LS-DYNA를 이용하여 유한요소해석이 수행되었다. 유한요소해석시 콘크리트와 모르타르는 Solid 요소로 모델링 되었으며, 철근과 FRP 그리드는 각각 Beam 요소 및 Shell 요소로 모델링되었다. 또한, 콘크리트와 철근은 완전부착하는 것으로 가정되었으며, 콘크리트와 모르타르 경계면의 부착파괴를 모사하기 위하여 Contact_Tiebreak_Surface_to_Surface 요소가 사용되었다. 이후, 국내⋅외 여러 연구자들에 의해 수행된 실험의 재현해석을 통해 제안된 유한요소해석 모델의 신뢰성이 검증되었다. 실험결과와 해석결과의 파괴양상을 분석하였을 때, 본 연구에서 제안된 유한요소해석 모델은 실험체의 부착파괴를 적절히 모사할 수 있는 것으로 나타났다. 또한, 해석을 통해 예측된 극한 강도에 대한 실험결과의 비는 평균 1.04, 표준편차 0.064로 제안된 해석모델은 FRCM 공법으로 보강된 철근콘크리트 보의 휨 성능을 비교적 잘 예측할 수 있는 것으로 나타났다.