56016 research outputs found
Sort by
????????? ???????????? ????????? ?????? ?????? ????????? ????????? ???????????? ???????????? ????????? ?????????
No full textDepartment of Physicsclos
???????????? ????????? ?????? ??-????????? ??????????????? C(sp3)-H ?????????
No full textDepartment of Chemistryclos
???????????? ??????????????????????????? ??????-??????????????? ?????? ??? ????????? ?????? ????????? ????????? ?????? ????????? ?????? ??? ???????????? ??????
Get PDFDepartment of Urban and Environmental Engineering (Environmental Science and Engineering)Gas hydrates are crystalline inclusion compounds that consist of well-defined hydrogen-bonded water frameworks (host cages) into which guest molecules are incorporated. As water is the main component
required for gas hydrate formation, gas hydrates are considered environmentally benign. Therefore, gas hydrates have been applied to various energy and environmental fields. This study deals with two main
topics as hydrate application technology.
In Topic 1, guest replacement and hydrate dissociation/reformation behavior during injection of CO2 and flue gas into NGH layer were comprehensively observed through an experimental method. we investigated CH4 - CO2 replacement behavior according to the experimental conditions or changes in the injected gas. Firstly, the time-dependent behavior during CH4 - CO2 replacement was quantitatively examined via a multi-methodological approach, which is a combination of PXRD analysis and in-situ Raman measurement, demonstrated that a significant guest exchange in the large 5 126 2 cages had a greater effect on the extent of replacement. Secondly, dissociation behavior of CH4 hydrate after N2 gas injection was confirmed. Time-dependent Raman spectra confirmed that the N2 molecules began to be captured in the hydrate cages immediately after the N2 gas injection at certain conditions, and the extent of N2 incorporation in the hydrate phase increased at a higher N2 gas injecting pressure, which indicates that N2 molecules revealed the dual functional roles of inhibitor for hydrate dissociation and external guest for guest replacement. Next, condition-dependent CH4 - CO2 + N2 replacement behavior was observed in terms of reaction kinetics and extent of replacement. At a higher temperature, the extent of replacement did not change, but CO2/N2 ratio in the replaced hydrates decreased slightly. An increase in the pressure led to an accelerated CO2 inclusion rate in the large (5126 2) cages at the initial stage and an enhanced N2 inclusion in the small (512) cages at the final stage. In case of injecting different composition of CO2 + N2 mixed gas, the extent of replacement was increased as N2 composition in the injecting gas increased at same partial pressure of injecting CO2, but the CO2 storage was decreased, which was due to the additional participation of N2 occurred in both large and small cages. In addition, the CO2 enclathration behavior in various mixed gas conditions are observed for CO2 sequestration. The presence of NaCl shifted the equilibrium conditions of CO2 + N2 hydrates to the higher-pressure or lower-temperature region, whereas it increased thermodynamic CO2 selectivity at a specified
temperature and pressure. In situ Raman spectroscopic measurements demonstrated that CO2 was kinetically selective at the early stage of CO2 + N2 hydrate formation and that kinetic CO2 selectivity was more noticeable in the saline water system. In case of CH4 + CO2 + N2 mixed hydrates, the CO2/CH4 ratio in the hydrate phase was not changed in different additional N2 compositions in same CO2/CH4 ratio of injecting gas.
In Topic 2, we focused on an evaluation of hydrate-based desalination efficiency from gaseous hydrate formers (propane, R134a, R152a, and R22). The thermodynamic stability, crystallographic information, dissociation enthalpy, and kinetic growth behavior of various hydrate formers were experimentally measured. The Hu-Lee-Sum (HLS) correlation was employed to predict the equilibrium shift and hydrate depression temperature. A novel approach to examine the maximum achievable salinity and maximum water yield using the HLS correlation was introduced. The theoretical HBD efficiency increased as the initial salinity decreased, the operating pressure decreased, and the initial subcooling temperature increased. In addition, theoretical HBD efficiency was employed as a quantitative standard for evaluating the kinetic performance of the HBD. At a fixed initial subcooling (2 K), R134a gave faster formation kinetics in the early stage, but R22 eventually offered highest hydrate conversion. At a fixed temperature (272 K), R152a showed fastest formation kinetics and highest HBD efficiency due to its milder hydrate equilibrium conditions. In addition, the estimation of theoretical desalianiton efficiency was also conducted on the liquid phase-hydrate former, cyclopentane (CP). The operating condition-dependent cooling demand for the process was simply examined from a thermodynamic point of view. The study provides a valuable theoretical foundation for the further development of gas hydrate-based desalination technology.ope
IC Level EMI Analysis using EMI Source Modeling
No full textDepartment of Electrical EngineeringOver the past decades, electronic applications have evolved for high speeding up the data process. Accordingly, the operating frequency of all electronic devices gets higher, and the systems have been integrated as much as possible to minimize the influence of parasitic components inside the system. Especially, in portable devices such as mobile phones, many of the chips including antennas are incorporated within a very short distance. In addition, with the development of 3D stacking packaging technology to improve the performance of each chip along with the form factor, the size of individual packages has become much larger.
However, the radiated emission caused by high-speed signals and by the large structure size of each package has raised package-to-package electromagnetic interference (EMI) problems. The distances between components are also close enough to affect each other inside the integrated system. Interference between the components causes data distortion and even causes malfunction. Particularly, interference at the antenna by the adjacent components is fatal since it reduces the sensitivity of the receiving antenna inside the portable devices. Hence, many studies have previously been conducted to analyze the EMI from the integrated circuits (IC), but these research have several constraints or do not sufficient to predict exact EMI from the real operating incorporated IC.
To overcome the limitations of previous research that for predicting the real IC EMI, the clarified EMI source information at the IC level is much desired. Thus, in this thesis, the novel modeling method of chip EMI source is proposed, and it is validated by comparing the simulation result and measurement result. In detail, an analysis of EMI from the package re-distribution layer (RDL) which is widely used in the newest mobile dynamic random-access memory (DRAM), LPDDR5, or others is discussed. In this research, the RDL is simply alternated into the simplified RDL structure using top metal in consideration of the wafer-level package process. After that, chip level EMI is analyzed for several conditions using a clarified EMI source model resulting in the proposed solution to reduce the chip EMI.
Finally, the main contribution of this paper is addressed by dividing the chip EMI analysis into the chip functional element and the structural element, the chip EMI can be managed and predicted using the simple EMI source circuit model in the field simulator before manufacturing the chip package.clos
Incremental Few Shot Learning across Datasets
No full textGraduate School of Artificial IntelligenceTo mimic the human being's intellectual characteristics, modern deep learning algorithms focus on two goals: adapting the new knowledge with few samples and recognizing both old and new knowledge as much as possible. The incremental Few Shot Learning problem tackles the above goals by targeting a fast adaptation and preventing catastrophic forgetting. Here, I present a new Incremental Few Shot Learning benchmark across multiple datasets. To address it, I utilize Meta Dataset, which composes of 10 diverse datasets. I also conduct experiments with prior incremental few-shot learning algorithms along with my novel algorithm to compare.clos
Inferring the 3D shapes of early galaxies
No full textDepartment of PhysicsGalaxies are shaped by a wide range of different physical processes, including processes related to the dark matter, formation and explosions of stars, evolution of supermassive black holes, and accretion and cooling of gas. Today, we still lack an understanding of how these different processes shape galaxies and lead to the diverse galaxy population observed. An observational challenge is that we only observe the 2D projections of galaxies on the sky, while we are interested in actually inferring the 3D shapes of galaxies. Recently, there have been several studies on how to use 2D galaxy images to constrain 3D shapes [1].
In this thesis, we develop a new Bayesian Hierarchical model to infer 3D galaxy information by deprojecting 2D data. We assume that the 3D shape of galaxies can be described as ellipsoids. We test how different assumptions regarding noise properties and ergodicity in the galaxy population affects our ability to recover the 3D shapes. We find that we can recover 3D shapes accurately for large samples of galaxies (N>100) that follow a Gaussian distribution in their 3D properties, given realistic observational uncertainties.
In the second part of the thesis, we contributed to the development of a novel photometry tool called ForcePho to measure fluxes and the 2D shapes for a large set of galaxies observed with the James Webb Space Telescope (JWST). The main assumption of this code is that the galaxies can be described with S??rsic profiles. Using ForcePho on the first JWST images, we measured 2D projections for thousands of high-redshift galaxies. Together with our deprojection code, we highlight the potential of how JWST together with sophisticated Bayesian Hierarchical model techniques can be used to constrain the shape of early galaxies and thereby learn more about their formation scenario.ope
?????? ????????? ????????? ????????? ????????? ????????? ?????? ?????? ?????? ??????
Get PDFGraduate School of Artificial IntelligenceTransfer learning has been successful in improving the performance of models and state-of-the-art models that work well for many benchmarks allow public access. But transfer learning is not always effective if the domain of the model is different from the domain of given data. Also, this strategy is not available for black box models. This type of model includes from a pre-compiled model to any type of model whose internal structure is not disclosed. Our problem is called "problem translation". Here, our goal is to improve a target model???s performance on a task of interest given other models and their databases.We start from the fact that information about the model is not the only information we have. We explore the framework to exploit the information about the data and the model together to solve our challenges. Our two main settings are when the model given information is black box problem and white box. We use the image-to-image translation model to change the domain of the data into a domain that the model can understand well, put this data in the target model and the reference model, and comprehensively utilize the final output to improve the performance of the target task of interest.ope
FiFo: Fishbone Forwarding in Massive IoT Networks
No full textMassive Internet of Things (IoT) networks have a wide range of applications, including but not limited to the rapid delivery of emergency and disaster messages. Although various benchmark algorithms have been developed to date for message delivery in such applications, they pose several practical challenges, such as insufficient network coverage and/or highly redundant transmissions to expand the coverage area, resulting in considerable energy consumption for each IoT device. To overcome this problem, we first characterize a new performance metric, forwarding efficiency, which is defined as the ratio of the coverage probability to the average number of transmissions per device, to evaluate the data dissemination performance more appropriately. Then, we propose a novel and effective forwarding method, fishbone forwarding (FiFo), which aims to improve the forwarding efficiency with acceptable computational complexity. Our FiFo method completes two tasks: 1) it clusters devices based on the unweighed pair group method with the arithmetic average and 2) it creates the main axis and subaxes of each cluster using both the expectation-maximization algorithm for the Gaussian mixture model and principal component analysis. We demonstrate the superiority of FiFo by using a real-world data set. Through intensive and comprehensive simulations, we show that the proposed FiFo method outperforms benchmark algorithms in terms of the forwarding efficiency
