http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
A Study on High Performance Converter Topology for Hydrogen Gas Generation Electrolysis System
Taewon Kang(강태원),Yuran Go(고유란),Yongsug Suh(서용석),Junik Jeong(정준익),Dohawn Rho(노도환) 전력전자학회 2010 전력전자학술대회 논문집 Vol.2010 No.7
This paper investigates a high performance converter topology for hydrogen gas generation electrolysis system. The proposed converter topology consists of full-bridge inverter, medium frequency transformer, and diode rectifier. Hydrogen gas generation electrolysis process considered in the paper is analyzed and characterized by its equivalent circuit. The electrolysis cell is modeled as effective resistance, capacitance, inductance, and internal emf voltage source. The proposed converter topology provides enhanced efficiency of hydrogen gas generation process under the operating condition of dc output voltage with high frequency ripple on it. The high performance operation of proposed converter is confirmed through the simulation with the electrolysis cell considered in the equivalent circuit model.
Taewon Kang(강태원),Changwoo Kim(김창우),Yongsug Suh(서용석),Hyeoncheol Park(박현철),Byungil Kang(강병익),Daegyun Kim(김대균) 전력전자학회 2011 전력전자학술대회 논문집 Vol.2011 No.7
This paper presents a simple and cost-effective stand-alone rapid battery charging system of 30kW for electric vehicles. The proposed system mainly consists of active front-end rectifier of neutral point clamped 3-level type and non-isolated bi-directional dc-dc converter of multi-phase interleaved half-bridge topology with coupled inductors. The charging system is designed to operate for both lithium-polymer and lithium-ion batteries. The complete charging sequence is made up of three sub-interval operating modes; pre-charging mode, constant-current mode, and constant-voltage mode. The pre-charging mode employs the staircase shaped current profile to accomplish shorter charging time while maintaining the reliable operation of the battery. The proposed system is able to reach the full-charge state within less than 16min for the battery capacity of 8kWh by supplying the charging current of 67A. The optimal discharging algorithm for Vehicle to the Grid (V2G) operation has been adopted to maintain the discharging current of 1C. Owing to the simple and compact power conversion scheme, the proposed solution has superior module-friendly mechanical structure which is absolutely required to realize flexible power expansion capability in a very highcurrent rapid charging system.
강태원(Taewon Kang) 한국생산제조학회 2013 한국생산제조학회지 Vol.22 No.1
It is important to have a well developed strain energy function in order to understand the mechanical behavior of biomaterial like the blood vessel of artery. However, since it is not possible to have a complete form of strain energy function of artery, theoretical framework describing the behaviour of Rubber-like material which is similar to blood vessel is applied to infer useful forms of strain energy function of biomaterial. Based on Chuong-Fung model and Mooney-Rivlin model, material parameters are estimated based on experimental data. From the results, it can be inferred that the estimated parameters can be used to explain the difference of mechanical characteristics between normal vessel and vessel with stent.
생체재료를 설명하는 스트레인 에너지 함수에 대한 이론적 고찰
강태원(Taewon Kang) 한국생산제조학회 2013 한국생산제조학회지 Vol.22 No.1
In order to understand the biomaterial like the blood vessel of artery, there is a need to quantify the biomechanical behavior of the vessel. However, theoretical framework to describe and quantify the behaviour of blood vessel was not well established so far. For studying the biomechanical behavior of artery, Rubber-liked material which is similar to passive artery is selected since conventional theoretical interpretation is very limited to understand and predict the behavior of biomaterial. Rubber-like material is assumed to be very similar to artery and has properties of isotropy, homogeneity and is undergoing large deformation. Based on this assumption, stress developed on Rubber-like material is described by strain energy function and strain invariants which are required to understand the nonlinear elastic behavior of biomaterial. The descriptor which would be used for understanding the biomechanical behavior of artery is studied in this work.
강태원(Taewon Kang),최원식(Won Sik Choi),김태우(Tae Woo Kim),이기성(Kee Sung Lee) 한국생산제조학회 2016 한국생산제조학회지 Vol.25 No.4
Digital printing has been used in various industrial areas, including semiconductor manufacturing and textile printing. However, implications on ceramic textile have not been well established so far. Printing high-viscosity materials requires an understanding of their behavior. An inorganic high viscous material with a viscosity range of 20-30 cps is analyzed using a viscometer and through X-ray diffraction. In this study, a digital printer is designed and assembled using a high-viscosity material with software for PC control, resulting in reduced processing at a fast area velocity of 20 m²/hr. The present study demonstrated that the printer is capable of controlling the shape of the drop mass to smear ink smoothly onto the ceramic surface under an economic budget. In addition, to avoid any difficulty in color management, the ceramic printer is equipped with an independent color management system designed to cope with images on a highly viscous material.
강태원(Taewon Kang) 한국정보기술학회 2020 한국정보기술학회논문지 Vol.18 No.2
This study is about the CGANN algorithm, which uses compact genetic algorithms to optimize multi-layer neural network weights. CGA is a variation of the genetic algorithm that does not construct a population sets and genetic operations like crossover and mutation but uses a prototype vector to generate chromosomes. In CGANN, the prototype vector is a vector storing the probability that the genes that make up a chromosome i.e. weights constituting the neural network take a specific value, and the CGA learns this to optimize the neural network. Using Kaggles Fruits360 image data set, CGANN was applied to neural networks for image classification problems, and it was found that, according to learning coefficient, it could be trained with more accuracy than GANN combined with common genetic algorithms.