• 한국안전학회지
• /
• 제29권4호
• /
• pp.116-122
• /
• 2014

This study proposes a system reliability analysis of rack storage facilities subjected to forklift colliding events. The proposed system reliability analysis consists of two steps: the first step is to identify dominant failure modes that most contribute to the failure of the whole rack facilities, and the second step is to evaluate the system failure probability. In the first step, dominant failure modes are identified by using a simulation-based selective searching technique where the contribution of a failure mode to the system failure is roughly estimated based on the distance from the origin in the space of the random variables. In the second step, the multi-scale system reliability method is used to compute the system reliability where the first-order reliability method (FORM) is initially used to evaluate the component failure probability (failure probability of one member), and then the probabilities of the identified failure modes and their statistical dependence are evaluated, which is called as the lower-scale reliability analysis. Since the system failure probability is comprised of the probabilities of the failure modes, a higher-scale reliability analysis is performed again based on the results of the lower-scale analyses, and the system failure probability is finally evaluated. The illustrative example demonstrates the results of the system reliability analysis of the rack storage facilities subjected to forklift impact loadings. The numerical efficiency and accuracy of the approach are compared with the Monte Carlo simulations. The results show that the proposed two-step approach is able to provide accurate reliability assessment as well as significant saving of computational time. The results of the identified failure modes additionally let us know the most-critical members and their failure sequence under the complicated configuration of the member connections.

• 정보화연구
• /
• 제9권4호
• /
• pp.413-422
• /
• 2012

웹 문서의 지속적인 증가로 인해 텍스트 기반, Page Rank 등의 방법으로 한 연구들이 증가하고 있다. 특히 웹 문서 내 URL 정보, HTML Tag 정보 등을 활용하는 연구들이 다시 주목을 받고 있다. 따라서 웹 문서 장르 분류를 위해 앞서 언급한 웹 문서 내 특징 요소들을 바탕으로 본 논문에서는 STW(Semantic Term Weight)를 적용하여 웹 문서 장르 분류하는 연구를 기술한다. 웹 문서 장르 분류에 사용되는 데이터 셋은 학습 문서와 테스트 문서로 구성되고, SVM 알고리즘을 사용하여 웹 문서 분류 실험을 수행한다. 학습 과정을 위해 20-Genre-collection corpus 내 1,000여개의 문서를 선정하여 SVM 알고리즘을 통해 학습하였고, 테스트 과정에서 사용된 데이터 셋은 KI-04 corpus를 사용하였다. 테스트 과정 후 STW를 사용한 실험과 STW를 사용하지 않은 실험으로 분류하여 정확도를 측정하였다. 또한 이를 바탕으로 1,212개의 테스트 문서를 분류하였다. 그 결과 STW를 사용한 실험 이 그렇지 않은 실험 보다 약 10.2% 높은 정확도를 보였다.

• 전기학회논문지
• /
• 제60권4호
• /
• pp.862-870
• /
• 2011

In this paper, we introduce a new topology of Radial Basis Function-based Polynomial Neural Networks (RPNN) that is based on a genetically optimized multi-layer perceptron with Radial Polynomial Neurons (RPNs). This study offers a comprehensive design methodology involving mechanisms of optimization algorithms, especially Fuzzy C-Means (FCM) clustering method and Particle Swarm Optimization (PSO) algorithms. In contrast to the typical architectures encountered in Polynomial Neural Networks (PNNs), our main objective is to develop a design strategy of RPNNs as follows : (a) The architecture of the proposed network consists of Radial Polynomial Neurons (RPNs). In here, the RPN is fully reflective of the structure encountered in numeric data which are granulated with the aid of Fuzzy C-Means (FCM) clustering method. The RPN dwells on the concepts of a collection of radial basis function and the function-based nonlinear (polynomial) processing. (b) The PSO-based design procedure being applied at each layer of RPNN leads to the selection of preferred nodes of the network (RPNs) whose local characteristics (such as the number of input variables, a collection of the specific subset of input variables, the order of the polynomial, and the number of clusters as well as a fuzzification coefficient in the FCM clustering) can be easily adjusted. The performance of the RPNN is quantified through the experimentation where we use a number of modeling benchmarks - NOx emission process data of gas turbine power plant and learning machine data(Automobile Miles Per Gallon Data) already experimented with in fuzzy or neurofuzzy modeling. A comparative analysis reveals that the proposed RPNN exhibits higher accuracy and superb predictive capability in comparison to some previous models available in the literature.

• Journal of Nutrition and Health
• /
• 제38권10호
• /
• pp.873-879
• /
• 2005

Impact and process evaluations were performed in order to verify the effectiveness of a public health nutrition program developed for child-bearing aged women in Korea. Participants included 58 high school girls who were divided into two groups. Each group received four 50 - 60 minute nutrition education lectures regarding healthy eating, osteoporosis, constipation and nutrition labeling in every two weeks. Each session took 50- 60 minutes. Regarding nutrition knowledge, there was a significant increase of degree of perception (p = 0.0004) , but no change in degree of accuracy after implementation (p = 0.9522) . Nutrition education was also effective in attitude change, showing more participants were ready to change their eating behaviors in terms of meal regularity (p = 0.0455) and less processed food intake (p =0.0143) . After implementing nutrition education, effective behavioral changes were observed in milk consumption (p =0.0037) and meal regularity (p = 0.0882) as well as daily activity such as stair use (p = 0.0701) . However, nutrition education had no effect on body mass index and perceived health status. In process evaluation conducted by a 9 item questionnaire, grand mean score was $4.17 \pm$0.72 out of 5. Proportion of items with scores higher than 4 ranged $68-91\%$. These results suggest that the nutrition education program used in this study was effective and useful. For a wider use of this program, more research was recommend for a strategy development of program diffuse. (Korean J Nutrition 38(10): 873$\sim$879,2005)

• Nutrition Research and Practice
• /
• 제13권6호
• /
• pp.521-528
• /
• 2019

BACKGROUND/OBJECTIVES: The aim of this study was to develop Korean food image detection and recognition model for use in mobile devices for accurate estimation of dietary intake. MATERIALS/METHODS: We collected food images by taking pictures or by searching web images and built an image dataset for use in training a complex recognition model for Korean food. Augmentation techniques were performed in order to increase the dataset size. The dataset for training contained more than 92,000 images categorized into 23 groups of Korean food. All images were down-sampled to a fixed resolution of $150{\times}150$ and then randomly divided into training and testing groups at a ratio of 3:1, resulting in 69,000 training images and 23,000 test images. We used a Deep Convolutional Neural Network (DCNN) for the complex recognition model and compared the results with those of other networks: AlexNet, GoogLeNet, Very Deep Convolutional Neural Network, VGG and ResNet, for large-scale image recognition. RESULTS: Our complex food recognition model, K-foodNet, had higher test accuracy (91.3%) and faster recognition time (0.4 ms) than those of the other networks. CONCLUSION: The results showed that K-foodNet achieved better performance in detecting and recognizing Korean food compared to other state-of-the-art models.

• 지식경영연구
• /
• 제22권3호
• /
• pp.91-106
• /
• 2021

최근 딥러닝 기술이 빠르게 발전함에 따라 국가 R&D 분야의 방대한 텍스트 문서를 다양한 관점에서 분석하기 위한 수요가 급증하고 있다. 특히 대용량의 말뭉치에 대해 사전학습을 수행한 BERT(Bidirectional Encoder Representations from Transformers) 언어모델의 활용에 대한 관심이 높아지고 있다. 하지만 국가 R&D와 같이 고도로 전문화된 분야에서 높은 빈도로 사용되는 전문어는 기본 BERT에서 충분히 학습이 이루어지지 않은 경우가 많으며, 이는 BERT를 통한 전문 분야 문서 이해의 한계로 지적되고 있다. 따라서 본 연구에서는 최근 활발하게 연구되고 있는 추가 사전학습을 활용하여, 기본 BERT에 국가 R&D 분야 지식을 전이한 R&D KoBERT 언어모델을 구축하는 방안을 제시한다. 또한 제안 모델의 성능 평가를 위해 보건의료, 정보통신 분야의 과제 약 116,000건을 대상으로 분류 분석을 수행한 결과, 제안 모델이 순수한 KoBERT 모델에 비해 정확도 측면에서 더 높은 성능을 나타내는 것을 확인하였다.

• 한국의류산업학회지
• /
• 제23권2호
• /
• pp.247-260
• /
• 2021

While it is an important element of clothing construction, research has so far been very limited on the similarities between virtual and real clothing in terms of the type of neckline. The purpose of this study is to verify the similarity, accuracy of virtualization, and actuality of neckline, which all play an important role in individual impressions and image formation, and require considerable modification when fitting real samples. A total of 5 neckline models were selected through the analysis of dress composition textbooks. The selected designs were then planned and manufactured in muslin. The specimen clothes were then tested on a female model in her 20s. 2 kinds of virtual bodies were created in order to compare the real and the virtual dressing. The first virtual body was made through an Artec 3D Eva scan of the model, and the other was made by entering the model's measurements in a CLO 3D program. A visual image of the front, side, and back image of both the real and virtual dressing were subsequently collected. The collected images were then evaluated by 20 professional fashion workers who checked the similarity between the real and the virtual versions. The current study found that the similarity between the actual and virtual wearing of the five neckline designs with reality appeared higher with the virtual wearing image using the 3D-scanned body. The results of this study could provide further information on the selection of appropriate avatars to clothing companies that check the fit of clothing by utilizing 3D virtualized programs.

• 한국융합학회논문지
• /
• 제12권7호
• /
• pp.53-60
• /
• 2021

LOD-기반 추천 시스템은 보통 DBpedia와 같은 LOD 데이터세트 내에서 사용가능한 데이터를 활용하여 최종 사용자에게 영화, 책, 음악과 같은 아이템을 추천한다. 이러한 시스템은 링크드 데이터 리소스 쌍 간의 일치 정도를 측정하는 의미 유사도 알고리즘을 사용한다. 이 논문에서는 LOD 그래프의 링크에 사용자 평가 등급을 변환한 가중치를 할당하여 LOD-기반 추천 시스템에서 의미 거리를 측정하는 새로운 접근방식을 제안했다. 이 논문에서 제안된 의미 거리 측정 모델은 가중치 계산을 통해 그래프가 사용자에게 개인화되는 처리 단계와 이러한 가중치를 LDSD에 적용하는 방법을 기반으로 한다. 실험 결과는 다른 유사한 방법들과 비교하여 제안된 방법이 더 높은 정확도를 보였으며, 추천 시스템의 의미 거리 측정의 범위를 넓혀서 유사도 향상에 기여하였다. 향후 연구로는 다른 방법의 LOD-기반 유사도 측정을 사용하여 모델에 미치는 영향을 분석하는 것을 목표로 한다.

• Structural Engineering and Mechanics
• /
• 제70권6호
• /
• pp.765-779
• /
• 2019

Most of the existing evaluation criteria of vibration serviceability rely on the peak acceleration of the structure rather than that of the people keeping their own body unmoved on the structure who is the real receiver of structural vibrations. In order to accurately assess the vibration serviceability, therefore, a full path assessment approach of vibration serviceability based on vibration source, path and receiver is not only tentatively proposed in this paper, taking the peak acceleration of receiver into account, but also introduce a probability procedure to provide more instructive information instead of a single value. In fact, semi-rigid supported on both sides of the structure is more consistent with the actual situation than simply supported or clamped due to the application of the prefabricated footbridge structures. So, the footbridge is regarded as a beam with semi-rigid supported on both sides in this paper. The differential quadrature-integral quadrature coupled method is not only to handle different type of boundary conditions, but also after being further modified via the introduction of an approximation procedure in this work, the time-varying system problem caused by human-structure interaction can be solved well. The analytical results of numerical simulations demonstrate that the modified differential quadrature-integral quadrature coupled method has higher reliability and accuracy compared with the mode superposition method. What's more, both of the two different passive control measures, the tuned mass damper and semi-rigid supported, have good performance for reducing vibrations. Most importantly, semi-rigid supported is easier to achieve the objective of reducing vibration compared with tuned mass damper in design stage of structure.

• 한국농공학회논문집
• /
• 제61권4호
• /
• pp.1-10
• /
• 2019

Unmanned aerial vehicle(UAV) can acquire images with lower cost than conventional manned aircraft and commercial satellites. It has the advantage of acquiring high-resolution aerial images covering in the field area more than 50 ha. The purposes of this study is to develop the rice grain yield distribution using UAV. In order to develop a technology for estimating the rice yield using UAV images, time series UAV aerial images were taken at the paddy fields and the data were compared with the rice yield of the harvesting area for two rice varieties(Singdongjin, Dongjinchal). Correlations between the vegetation indices and rice yield were ranged from 0.8 to 0.95 in booting period. Accordingly, rice yield was estimated using UAV-derived vegetation indices($R^2=0.70$ in Sindongjin, $R^2=0.92$ in Donjinchal). It means that the rice yield estimation using UAV imagery can provide less cost and higher accuracy than other methods using combine with yield monitoring system and satellite imagery. In the future, it will be necessary to study a variety of information convergence and integration systems such as image, weather, and soil for efficient use of these information, along with research on preparing management practice work standards such as pest control and nutrient use based on UAV image information.