본 논문에서는 저 레이놀즈 수 영역에서 에어포일의 공기역학적 성능을 예측하기 위한 딥러닝 기반의 축소 모델을 제시하였다. 딥 러닝 기반 축소 모델에서 CFD 해석 결과의 높은 차원의 데이터를 효율적으로 다루기 위해 변이형 오토인코더를 결합한 합성곱 신경 망을 적용하였다. 부호화 거리 함수를 통해 에어포일의 형상과 유동 조건을 이미지 데이터화 하고, 이에 대해 합성곱 신경망을 매개변 수화 하였다. 또한, 전산유체역학 해석의 계산 비용으로 인한 부족한 훈련 데이터를 극복하기 위해 투영 기반의 비선형 매니폴드 데이 터 증강기법을 개발하였다. NACA 4계열 에어포일은 해석 예제로 고려하여 제안하는 프레임워크의 내삽과 외삽 정확도를 평가하였 으며 매니폴드 데이터 증강기법을 적용하여 프레임워크의 정확도 향상을 확인하였다.

Block pavements are widely used in various infrastructures, offering durability and aesthetic appeal. However, assessing their condition through manual methods is resource-intensive and subjective. This study proposes a deep learning approach using the Hybrid TransUNet model to enhance the accuracy and efficiency of detecting block pavement distresses. A dataset of over 10,000 images was used to train and test binary and multiclass segmentation models, significantly improving detection accuracy. The results show that the Hybrid TransUNet model outperforms other models, though challenges in detecting certain distress types like cracks persist.

Evaluating the performance of asphalt concrete using CT scanning has become an essential area of research due to its potential to revolutionize the way we assess road materials. Traditional methods often require destructive sampling, which can damage infrastructure and offer limited insight into the material's internal structure. In contrast, CT scanning provides a non-destructive, highly detailed analysis of asphalt's internal features, such as air voids, aggregate distribution, and binder coverage, all of which are critical to its durability and performance. Additionally, the ability to create 3D models from CT scans allows for deeper insights into factors like void connectivity and aggregate bonding, which directly affect the lifespan of pavements. By combining CT imaging with advanced data processing techniques, such as deep learning, this research offers more accurate and reliable methods for optimizing asphalt mix designs, ultimately leading to longer-lasting roads, reduced maintenance costs, and more sustainable construction practices.

최근 늘어나고 있는 이상 기상 현상으로 산사태 위험이 점차 증가하고 있다. 산사태는 막대한 인명 피해와 재산 피해를 초래할 수 있기에 이러한 위험을 사전에 평가함은 매우 중요하다. 최근 기술 발전으로 인해 능동형 원격탐사 방법을 사용하여 더 정확하고 상세한 지표 변위 및 강수 데이터를 얻을 수 있게 되었다. 그러나 이러한 데이터를 활용하여 산사태 예측 모델을 개발하는 연구는 찾기 힘들다. 따라서 본 연구에서는 합성개구레이더 간섭법(InSAR)을 사용한 지표 변위 자료와 하이브리드 고도면 강우(HSR) 추정 기법을 통한 강수 정보를 활용하여 산사태 민감도를 예측하는 기계학습 모델을 제시하고 있다. 나아가 기계학습의 블랙박스 문제를 극복할 수 있는 해석가능한 기계학습 방법인 SHAP을 이용하여 산사태 민감도의 영향 변수에 대한 중요도를 체계적으로 평가하였다. 경상북도 울진군을 대상으로 사례 연구를 수행한 결과, XGBoost가 가장 좋은 예측 성능을 보이며, 도로로부터의 거리, 지표 고도, 일 최대 강우 강도, 48시간 선행 누적 강우량, 사면 경사, 지형습윤지수, 단층으로 부터의 거리, 경사도, 지표 변위, 하천으로부터의 거리가 산사태 예측에 영향을 미치는 주요 변수로 밝혀졌다. 특히, 능동형 원격탐사를 통해 얻은 자료인 강우 강도와 지표 변위의 절댓값이 높을수록 산사태 발생 확률이 높음을 확인하였다. 본 연구는 능동형 원격탐사 자료의 산사태 민감도 연구에서의 활용 가능성을 실증적으로 보여주고 있으며, 해당 자료를 바탕으로 시공간적 으로 변하는 산사태 민감도를 도출함으로써 향후 산사태 민감도 모니터링에 효과적으로 활용될 수 있을 것으로 기대된다.

본 연구는 한국어 학습자들이 자신들의 언어학습 신념이 그들의 전반 적인 언어학습과 새로운 환경에 적응하는 것에 어떠한 관계가 있는지를 찾기 위함이 목적이다. 여섯 명의 한국대학교에 재학 중인 외국인 유학 생들을 대상으로 그들의 언어학습 신념에 대한 인터뷰가 진행되었다. Horwitz의 BALLI 문항을 기반으로 한 반구조적 인터뷰가 진행되었으며, 그들의 답변은 외국어 능력, 언어학습의 어려움, 언어학습의 본성, 학습 과 소통전략, 동기와 기대 등의 주제를 기반으로 비슷한 패턴을 보이는 항목별로 구분되어졌다. 연구 결과는 외국인 유학생들은 그들의 한국어 실력이 향상될수록 자신감을 가졌으며, 그들은 문화에 대한 학습이 언어 학습에 중요한 부분이라는 것을 인지했다. 또한, 가장 효과적인 학습법은 실제적인 상황과 사람들과 어울리는 것을 꼽았으며, 그들은 대학 졸업 이후에도 한국에 남아 있기를 희망했다.

PVA 섬유 보강 시멘트 복합체는 매우 복잡한 미세구조를 가지고 있으며, 재료의 거동을 정확히 평가하기 위해서는 미세구조 특성 을 반영하여 실제 실험과 시너지효과를 내며 효율적인 재료 설계를 가능하게 하는 해석 모델의 개발이 중요하다. PVA 섬유 보강 시멘 트 복합체의 역학적 성능은 PVA 섬유의 방향성에 큰 영향을 받는다. 그러나 마이크로-CT 이미지로부터 얻은 PVA 섬유의 회색조 값 을 인접한 상과 구분하기 어려워, 섬유 분리 과정에 많은 시간이 소요된다. 본 연구에서는 섬유의 3차원 분포를 얻기 위하여 0.65μm3 의 복셀 크기를 가지는 마이크로-CT 이미지 촬영을 수행하였다. 학습에 사용될 학습 데이터를 생성하기 위해 히스토그램, 형상, 그리 고 구배 기반 상 분리 방법을 적용하였다. 본 연구에서 제안된 U-net 모델을 활용하여 PVA 섬유 보강 시멘트 복합체의 마이크로- CT 이미지로부터 섬유를 분리하는 학습을 수행하였다. 훈련의 정확도를 높이기 위해 데이터 증강을 적용하였으며, 총 1024개의 이미지 를 훈련 데이터로 사용하였다. 모델의 성능은 정확도, 정밀도, 재현율, F1 스코어를 평가하였으며, 학습된 모델의 섬유 분리 성능이 매 우 높고 효율적이며, 다른 시편에도 적용될 수 있음을 확인하였다.

This study was conducted to calculate the damage of Italian ryegrass (IRG) by abnormal climate using machine learning and present the damage through the map. The IRG data collected 1,384. The climate data was collected from the Korea Meteorological Administration Meteorological data open portal.The machine learning model called xDeepFM was used to detect IRG damage. The damage was calculated using climate data from the Automated Synoptic Observing System (95 sites) by machine learning. The calculation of damage was the difference between the Dry matter yield (DMY)normal and DMYabnormal. The normal climate was set as the 40-year of climate data according to the year of IRG data (1986~2020). The level of abnormal climate was set as a multiple of the standard deviation applying the World Meteorological Organization (WMO) standard. The DMYnormal was ranged from 5,678 to 15,188 kg/ha. The damage of IRG differed according to region and level of abnormal climate with abnormal temperature, precipitation, and wind speed from -1,380 to 1,176, -3 to 2,465, and -830 to 962 kg/ha, respectively. The maximum damage was 1,176 kg/ha when the abnormal temperature was -2 level (+1.04℃), 2,465 kg/ha when the abnormal precipitation was all level and 962 kg/ha when the abnormal wind speed was -2 level (+1.60 ㎧). The damage calculated through the WMO method was presented as an map using QGIS. There was some blank area because there was no climate data. In order to calculate the damage of blank area, it would be possible to use the automatic weather system (AWS), which provides data from more sites than the automated synoptic observing system (ASOS).

A typical consumer is exposed to more than 5,000 advertisements per day (Story 2015) by exposure to around 500 advertising messages before ending breakfast (Marshall, 2015), and it is impossible for consumers to remember most of the advertisement images and messages. Thus, one consistent yet not thoroughly investigated question for advertisers is how advertisers draw consumers' attention by differentiating their brand from competitors' brands. One suggestion from academia is making more "creative" advertising (Dahlén et al., 2008; Lehnert et al., 2014; Rosengren et al., 2013; Smith et al., 2008). However, it is still questionable the exact meaning of "advertising creativity," and the effects of creative advertising on consumer evaluation have not been fully investigated the effects of creativity in advertising evaluation by considering various boundary conditions. The objective of this research is to redefine advertising creativity, to understand how advertising creativity shapes consumers' evaluation, and how these effects are moderated by the different types of boundary conditions, such as industry category, by analyzing more than 100,000 advertising images and copies using a cutting-edge transfer learning technique. The results of the transfer learning algorithm indicate that both cognitive dimensions (e.g., novelty of image) and affective dimensions (e.g., awe and coolness) simultaneously affect the consumers' perception of the advertising creativity, and the current algorithm enables to detect of creative advertising image with 92% accuracy rate.

The popularity of live streaming is driving the emergence of a new business model, known as live-streaming commerce (LSC). While there are more and more broadcasters in LSC, their behaviors and performance of them are significantly different. To have a better understanding of broadcasters, we employ different machine learning models to identify different portraits in both static and dynamic dimensions. We collect a rich live-streaming dataset from one leading platform in China. Our dataset features information for both broadcasters and viewers, including viewers’ purchasing behaviors, viewers’ records of posting words, broadcasters’ gender, the number of followers for broadcasters, and the live streaming show information, including the start and end time, and the viewers in each live streaming show. The rich textual information in broadcasters’ profile induction provides us a good opportunity to uncover different static portraits and the records in live streaming shows give us a chance to identify different dynamic behavioral portraits for broadcasters.

Social media have emerged as one of the most important tools for firms to engage customers (e.g., Chandrasekaran et al., 2022; Cheng & Edwards, 2015; Lee et al., 2018; Wedel & Kannan, 2016). Within the tourism industry, scholars have investigated the role of social media communication in various contexts, such as online travel information search (Xiang & Gretzel, 2010), sharing travel experiences (So et al., 2018; Wang et al., 2022) and establishing positive customer relationships (Jamshidi et al., 2021). Insights into which social media content makes for generating positive engagement are, however, still largely based on marketers’ intuitions or focusing on message factors of social media posts such as message appeals (e.g., Wang & Lehto, 2020). It also often neglects the importance of the visual component of social media posts, and only a few research have investigated the effects of the image in social media on the travel industry (e.g., Fusté-Forné, 2022). The objective of this research is, therefore, to understand how textual features and image features generate user engagement in social media utilizing cutting-edge transfer learning techniques and to propose how these features should be customized to maximize user engagement for online travel shopping companies. We collect and analyze more than 10,000 Instagram posts from three online travel shopping companies, including Expedia, Priceline, and Kayak. The results from transfer learning algorithms utilizing 24 features, such as the number of people in the image, emotions expressed in the people in the image, hue, and RGB value, successfully predict the level of engagement measured by the number of likes and comments.

본 연구는 딥러닝을 위한 비선형 변환 접근법을 사용하여 Single-lap joint의 접착 영역을 조사하기 위한 진동 응답 기반 탐지 시스템 을 제시한다. 산업 혹은 공학 분야에서 분해가 쉽지 않은 구조 내에 보이지 않는 부분의 상태와 접착된 구조의 접착 부위 상태를 알기 어려운 문제가 있다. 이러한 문제를 해결하기 위해 본 연구는 비선형 변환을 이용하여 기준 시편의 진동 응답으로 다양한 시편의 접착 면적을 조사하는 탐지 방법을 제안한다. 이 연구에서는 CNN 기반 딥러닝으로 진동 특성을 파악하기 위해 비선형 변환을 적용한 주파 수 응답 함수를 사용했고 분류를 위해 가상의 스펙트로그램을 사용했다. 또한, 제시된 방법을 검증하기 위해 알루미늄, 탄소섬유복합 재 그리고 초고분자량 폴리에틸렌 시편에 대한 진동 실험, 분석적 해, 유한요소해석을 수행했다.

Since radon was detected in mattresses of famous bed furniture brands in 2018, the nuclear safety and security commission (NSSC) announced the radiation safety management act in April 2021 to protect the public health and environment. This act stipulates the safety management of radiation that can be encountered in the natural environment such as the notification of radioactivity concentration of source materials, process by-products, the installation and operation of radioactive monitors. In this study, a model was established to predict radioactive exposure dose from radioactive materials such as radon and uranium detected in consumer products such as bed mattresses, pillows, shower, bracelets and masks in order to identify major radioactive substances that largely affect the exposure dose. A period of seven years from 2014 to 2020 was investigated for the source materials and exposure doses of consumer products containing naturally occurring radioactive materials (NORMs). We analyzed these using machine learning models such as classification and regression tree (CART), Random Forest and TreeNet. Index development and verification were performed to evaluate the predictive performance of the models. Overall, predictive performance was highest when Random Forest or TreeNet was used for each consumer product. Thoron had a great influence on the internal exposure dose of bedding, clothing and mats. Uranium had a great influence on the internal exposure dose of other consumer products except whetstones. When the number of data is very small or the missing value rate is high, it is difficult to expect accurate predictive performance even with machine learning techniques. If we significantly reduce the missing value rate of data or use the limit of detection value instead of missing values, we can build a model with more accurate predictive performance.

Radioactivity of radiostrontiums, Sr-89 and Sr-90, which are both pure beta-emitters, are generally measured via Cherenkov counting. However, the determination of Cherenkov counting efficiencies of radiostrontiums requires a complicated procedure due to the presence of Y-90 (also a pure betaemitter) which is the daughter nuclide of Sr-90. In this study, we have developed a machine learning approach using a linear regression model which allows an easier and simultaneous determination of the Cherenkov counting efficiencies of the radiostrontiums. The linear regression model was employed because total net Cherenkov count (Ct) from the three beta-emitters at time t after the separation of Y- 90, can be expressed as a linear combination of their respective time-varying radioactivities with their respective coefficients (parameters) being their counting efficiencies: Ct = εSr-90[ASr-90·exp(–λSr-90·t)] + εSr-89[ASr-89·exp(–λSr-89·t)] + εY-90[ASr-90·exp(1–λSr-90·t)], where ε is a counting efficiency, A is an initial activity, λ is a decay constant and t is time after the separation of Y-90, Thus, if we train the model with multiple Cherenkov counts measured from the three beta emitters, then we can obtain their estimates for counting efficiencies (so-called parameters) straightforward. For this, the model has been trained by two methods: Ordinary Least Squares (OLS) and Bayesian linear regression (BLR), for which two software packages, PyMC3 and Stan were employed to compare their performances. The results showed that the accuracy of the OLS was worse than that of the BLR. Particularly, the counting efficiency of Sr-90 was estimated to be smaller than 0, which is an unrealistic value. On the other hand, the estimates of the BLR gave realistic values which are close to the true values. Additionally, the BLR was able to provide a distribution for each counting efficiency (so-called “posterior”) from which various types of inference can be made including median and credible interval in the Bayesian statistics which is analogous to, but different from confidence interval in the Frequentist statistics. In the results of the BLR, the Stan package gave more accurate estimates than the PyMC3 package. Therefore, it is expected that counting efficiencies of the radiostrontiums including radioyttrium can be determined at the same time, more easily and accurately, by using the BLR with the Stan package and that the activities of radiostrontium also can be determined more easily by using the BLR if we know their counting efficiencies in advance. It is worth noting that the usage of the linear regression model in this study was different from the usual one where the trained model is used to predict a response value (count) from a set of unseen regressor values (activities).

The success of machine learning approach to identify key correlation in large database is critically controlled by the reliability and accuracy of the data. Here, we demonstrate that rigorous material properties of radioactive nuclear fuels can be obtained by integrated approach of first principles calculations and the machine learning approach. The reliable database is established by density functional theory and molecular dynamics simulations, which is the input of the machine learning to analyze any correlation among the database. The outcomes are applied to evaluate thermodynamic, kinetic and electrochemical properties, which plays a key role for safe management of spent nuclear fuels.

The rapid development of computer vision and deep learning has enabled these technologies to be applied to the automated classification and counting of microscope images, thereby relieving of some burden from pathologists in terms of performing tedious microscopic examination for analysis of a large number of slides for pathological lesions. Recently, the use of these digital methods has expanded into the field of medical image analysis. In this study, the Inception-v3 deep learning model was used for classification of chondrocytes from knee joints of rats. Knee joints were extracted, fixed in neutral buffered formalin, decalcified, processed and embedded in paraffin, and hematoxylin and eosin (H&E) stained. The H&E stained slides were converted into whole slide imaging (WSI), and the images were cropped to 79 × 79 pixels. The images were divided into training (60.42%) and test (39.58%) sets (46,349 and 30,360 images, respectively). Then, images containing chondrocytes were classified by Inception-v3 and accuracy was calculated. We visualized the images containing chondrocytes in WSIs by adding colored dots to patches. When images of chondrocytes in knee joints were evaluated, the accuracy was within the range of 91.20 ± 8.43%. Therefore, it is considered that the Inception-v3 deep learning model was able to distinguish chondrocytes from non-chondrocytes in knee joints of rats with a relatively high accuracy. The above results taken together confirmed that this deep learning model could classify the chondrocytes and this promising approach will provide pathologists a fast and accurate analysis of diverse tissue structures.

밸브의 내부 누설 현상은 밸브의 내부 부품의 손상에 의해 발생하며 배관 시스템의 사고와 운전정지를 일으키는 주요 요인이 다. 본 연구는 버터플라이형 밸브의 내부 누설에 따라 배관계에서 발생하는 음향방출 신호를 이용하여 배관 가동 중 실시간 누설 진단의 가능성을 검토하였다. 이를 위해 밸브의 작동 모드별로 측정한 시간영역의 AE 원시신호를 취득하였으며 이로부터 구축한 데이터셋은 데 이터 기반의 인공지능 알고리즘에 적용하여 밸브의 내부 누설 유무를 진단하는 모델을 생성하였다. 누설 유무진단을 분류의 문제로 정의 하여 SVM 기반의 머신러닝과 CNN 기반의 딥러닝 분류 알고리즘을 적용하였다. 데이터의 특징 추출에 기반한 SVM 분류 모델의 경우, 이 진분류 모델에서 구축된 모델에 따라 83~90%의 정확도를 나타냈으며, 다중 클래스인 경우 분류 정확도가 66%로 감소하였다. 반면, CNN 기반의 다중 클래스 분류 모델의 경우 99.85%의 분류 정확도를 얻을 수 있었다. 결론적으로 밸브 내부 누설 진단을 위한 SVM 분류모델은 다중 클래스의 정확도 향상을 위해 적절한 특징 추출이 필요하며, CNN 기반의 분류모델은 프로세서의 성능 저하만 없다면 누설진단과 밸브 개도 분류에 효율적인 접근방법임을 확인하였다.