http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
이동식 미분무수 노즐의 소화 특성에 대한 수치 시뮬레이션
배강열(K. Y. Bae),정희택(H. T. Chung),김형범(H. B. Kim),김찬희(C. H. Kim),이창효(C. H. Lee),김창(C. Kim) 한국동력기계공학회 2007 한국동력기계공학회 학술대회 논문집 Vol.- No.-
In the present study, the numerical investigation on the effects of water mist spray has been carried out for the fire suppression characteristics. The FDS are used to simulate the interaction of fire plume and water mists, and program describes the fire-driven flows using LES turbulence model, the mixture fraction combustion model, the finite volume method of radiation transport for a non-scattering gray gas, and conjugate heat transfer between wall and gas flow. The numerical model is consisted of a rectangular of L×W×H=4.0×4.0×2.5 m and a water mist nozzle that be installed 1.0 m from pool, and the whole walls are assumed the open. In the study, the parameter of simulation has two type that one is consisted of 24 hole on the primary flow surroundings and the other has 8 hole except the primary flow. Finally, the nozzle type that has 24 hole is appeared lower the characteristic of fire suppression than another type because of the flow through many holes is bad impact for the primary flow. The extinguishing time is obtained 65 sec in case of B-type and 170 sec in case of A-type, and B-type showed high effect of fire suppression about 2.6 times than A-type,
김형범,임정호,허유,Kim, Hyung-B.,Lim, Jung-H.,Huh, You 한국섬유공학회 2011 한국섬유공학회지 Vol.48 No.3
Multi-layered, prepreg or preform textile structures are often used in industrial components to achieve specific mechanical properties. Therefore a beam scanning method was developed to measure the internal structures of multi-layered preforms in a nondestructive way. The measurement system employed a low-energy X-ray beam and an X-ray detector. Specific software programs were also developed for image data processing. Trial measurements of textile specimens prepared from compound yams demonstrated the system's ability to image internal structures by detecting the layers' folded states. The method can feasibly be used for measuring the internal states of specimens with stacked textile structures.
H. S. Nam(남한승),H. B. Kim(김형범),I. Y. Park(박일용),S. H. Lee(이승하) 한국재활복지공학회 2021 재활복지공학회논문지 Vol.15 No.1
영농 폐기물의 증가로 인해, 빠르고 효율적으로 수거할 수 있는 스마트 영농 폐기물 모니터링 시스템 개발이 필요하다. 본 논문에서는 영농 폐기물 분류 시스템을 제안하고 실제 지역 농촌에서 직접 수집한 영상을 이용하여 CNN 기반의 전이학습 모델들을 구현하고 비교하였다. 영농 폐기물 영상 분류에 적합한 모델과 학습 조건을 찾기 위해, 3가지의 학습 자료군 구성 조건 (2종 분류, 6종 분류, 6종 하위분류를 가진 2종 분류)을 달리하여 미세 조정된 6개의 사전 훈련 CNN 모델들의 검증 정확도를 비교하였다. 그 결과, ResNet-50 모델의 성능이 모든 학습 조건에서 평균 90.9%의 정확도로 가장 높았고, 폐기물 영상을 6종 분류했을 때보다 2종 분류로 했을 때의 검증 정확도가 10% 더 높았다. 특히, 학습 자료군 구성 방법 중 6종 하위분류를 가진 2종 분류했을 때의 검증 정확도는 2종 분류했을 때와 유사했다. 이를 통해 영농 폐기물은 한 종류만 모여 있지 않을뿐더러 다양한 폐기물들이 한데 섞여 있어서 영농 폐기물의 특정한 세부 종류로 분류하는 것보다 폐기물인지 아닌지를 이진 분류하는 것이 더 효과적임을 확인하였다. 나아가, 제안된 시스템의 동작을 확인하기 위해, 영농 환경 모니터링 서버와 영농 폐기물 영상 분류 서버 사이에 TCP / IP 기반의 통신 환경을 구축하고, 모의실험을 통해 구현한 영농 폐기물 영상 분류 시스템이 스마트 영농 폐기물 모니터링 시스템으로 사용될 가능성을 확인하였다. 본 연구의 결과는 정형화되지 않거나 여러 병변이 혼합된 의료 영상을 분류하는 경우에도 활용될 수 있을 것이다. Due to the increase of farm waste in many countries, there’s a need to develop a smart farm waste monitoring system that can collect it promptly and efficiently. In this paper, we proposed, compared the performance of a convolutional neural network (CNN) -based transfer learning models and implement a farm waste image classification system, which is crucial component for the monitoring system. To find an appropriate model and labelling methods for farm waste image classification, we compared each validation accuracy of six different pre-trained CNN methods with three types of labelling scheme, using the waste images taken directly from the farming area. As a result, the ResNet-50 model performed best with an accuracy of 90.9% on average. Also, when classified into 2 categories, the accuracy was about 10% higher than that of the 6 categories. Furthermore, when the image was classified into 2 main categories with 6 sub-categories, the validation accuracy was similar to that of the 2 categories. Through these results, it seemed to be more effective to classify with binary labels such as ‘trash’ and ‘non-trash’, rather than with multiple labels of specific categories because farm waste is generated not only by single type of waste but also by various types of mixed waste. And a TCP / IP based communication environment between farm environment monitoring server and farm waste image classification server has been implemented. Experimental results using the system implemented for a smart farm waste monitoring showed that the proposed system can be used for a smart farm waste collection system. Also, the result of this study could be applied to classify medical images of unstructured and/or mixed lesion.