56016 research outputs found
Sort by
GC-RED : Graph Convolution-based Recurrent Encoder-Decoder for anomaly diagnosis in multivariate time series
No full textDepartment of Industrial Engineeringclos
Graph Neural Networks for Multivariate Time Series Forecasting with Applications
No full textDepartment of Industrial Engineeringclos
Hydrologic and water quality modeling using SWAT and deep learning model
No full textDepartment of Urban and Environmental Engineering (Environmental Science and Engineering)clos
Improvement of Air Quality Forecasts using Aerosol Data Assimilation and Artificial Intelligence
No full textDepartment of Urban and Environmental Engineering (Environmental Science and Engineering)This thesis proposes the development of advanced aerosol data assimilation (DA) system combined with artificial intelligence (AI) to improve air quality forecasts in South Korea. The aerosol DA system could significantly improve aerosol forecasting skills by reducing the uncertainty of initial condition (IC). This study developed the aerosol data assimilation system based on the Weather Research and Forecasting model with chemistry (WRF-Chem). There are two main approaches in DA: one is the variational method (e.g. 3D-VAR, 4D-VAR) and the other is the ensemble method (e.g. EnKF). The advantages of EnKF are flow-dependent background error covariance which is important in fast-developing air quality system and considering the nonlinearity of background states through ensemble forecast. However, EnKF has a limitation of sampling problem due to inadequate ensemble size. Without enough ensemble spread, EnKF exhibits poor performance. Although previous studies employed inflation and covariance localization to alleviate the sampling problem in chemical data assimilation, it is still deficient, as the ensemble spread is smaller than the prior root mean square error. This study suggested two new effective methods for increasing model background error in the EnKF aerosol data assimilation: the Multiphysics approach and perturbation to prognostic variables. Both methods improved the quality of surface PM analysis substantially. And the EnKF experiment which incorporates both uncertainty in model physics and prognostic variables, demonstrates the best performance, resulting in a larger ensemble spread.
Unlike climate models, emission inventory is a large part of model error in chemical transport models. Bottom-up emission inventory from detailed industrial statistical data has some limitations such as slow updates and missing values, due to its diverse emission sources and complex calculation processes. However, a timely update of the emission inventories is an essential prerequisite for an acceptable air quality forecast. Therefore, optimized emissions from a top-down approach using EnKF could be timely updated and lead to improvements in the forecasting of air quality.
Furthermore, the data assimilation has some limitations regarding linear assumption, Gaussian hypothesis, and unbiased assumption. I propose a novel approach that improves the DA system by applying an artificial intelligence algorithm that does not require the above assumptions and can consider nonlinearity. By replacing the AOD observation operator in the DA system with AI algorithms, the performance of AOD data assimilation showed significant benefits. By combining data assimilation and machine learning, the effectiveness of DA using satellite AOD retrievals on air quality simulations can be maximized. Overall, WRF-Chem and the improved aerosol data assimilation system provide a framework for air quality forecasting and emissions constraint that can be used to enhance our understanding of the interactions between air quality, climate change, and human health.clos
????????? ????????? ????????? ?????? ????????? ??????????????? ?????? ??????
Get PDFSchool of Energy and Chemical Engineering (Chemical Engineering)Optoelectronic devices are essential components that exploit electromagnetic radiation based on light-matter interactions and have been developed with the advance of nanotechnology. In particular, communication between materials at interfaces is crucial because optoelectronic devices are made up of multilayer structures. Interface/surface engineering allows the regulation of physical stress, energy barrier, interfacial band alignment, and defect passivation, all of which are directly related to the performance of devices. The arrangement of energy levels affects charge transport and can be adjusted by the dipole moment on the surface of nanomaterials. Defect passivation on the interface/surface of the materials reduces trap states that induce charge accumulation as a charge injection barrier or act as nonradiative recombination centers. These effects are prominent in nanomaterials because of their high surface-area-to-volume ratio. Furthermore, since the nanomaterial layer of optoelectronics is a microscopic assembly, surface engineering can regulate the communication between nanomaterials by changing the interparticle distance and packing density. This dissertation describes surface engineering strategies for semiconductor nanomaterials to improve the efficiency of optoelectronic devices.
Chapter 2 demonstrates the interface engineering between the graphene electrode and the hole transport layer for organic solar cells. A thin layer of norepinephrine, an amphiphilic catecholamine derivative, is applied to the graphene electrode as a hydrophilic surface modifier to enable efficient surface modification without significantly decreasing the optical transmittance or the electrical conductivity. The power conversion efficiency of organic solar cells fabricated with this poly(norepinephrine)-coated graphene electrode is 7.93%, close to that of the ITO-based reference device with a power conversion efficiency of 8.73%.
In Chapter 3, we demonstrate the surface engineering of InP@ZnSeS quantum dots for a color filter. Phenylethyl mercaptan ligands are used as surface modifiers, substituting the original oleic acid ligands. In the propylene glycol monomethyl ether acetate solvent, more than 75% of photoluminescence quantum yields are retained without a peak shift in the emission spectrum during the ligand exchange procedure. Surface-engineered quantum dots show high colloidal stability in propylene glycol monomethyl ether acetate for more than one month under ambient conditions.
Chapter 4 demonstrates the interface engineering between CsPbBr3 and Cs4PbBr6 perovskite nanocrystals for blue light-emitting diodes. Highly emissive blue CsPbBr3 quantum dots are developed by controlling the reactivity of metal precursors. In situ generated Cs4PbBr6 nanocrystals eliminate the surface defects of CsPbBr3 quantum dots, resulting in high colloidal and thermal stability. The fabricated blue light-emitting diode shows an external quantum efficiency of 4.65% at 480 nm and excellent spectral stability under operational conditions.
As previously discussed, we have developed a catechol-based nanocoating method, a thiol-based ligand exchange method, and a surface reconstruction strategy for perovskite materials. Our strategies can accelerate the development of a wide range of optoelectronic applications thanks to their versatility.ope
Interlayer Exchange Coupling in Pt/Co Bilayers with Perpendicular Magnetic Anisotropy
No full textDepartment of Materials Science and EngineeringIn this thesis, we studied interlayer exchange coupling and domain wall dynamics in the perpendicularly magnetized Pt/Co bilayers separated by Pt spacer. Before studying the interlayer exchange coupling of Pt/Co bilayer structures, it is necessary to understand the magnetic properties of the Pt/Co/Pt structures. The first part of the thesis focuses on the magnetic properties of the Pt/Co/Pt structures for the Co and Pt thicknesses. The magnetic properties, including coercivity and perpendicular anisotropy, are affected by the thickness of the Co and Pt layers. We offer a method to control the coercivity by tailoring the Pt and Co thicknesses in the Pt/Co/Pt structures.
The interlayer exchange coupling and magnetization reversal processes were investigated using a polar MOKE system and a magneto-optical Kerr microscope. The interlayer coupling in Pt/Co bilayers consisting of two Co layers with different thicknesses is determined by the Pt spacer thickness. The Pt/Co bilayer with tPt=3 nm exhibits ferromagnetic coupling, and at tPt > 3 nm the two Co layers switch individually. The interlayer coupling varies for the tPt, and minor loops of the Pt/Co bilayers with tPt in the range of 3.1???4.5 nm show the presence of antiferromagnetic interlayer coupling. These interlayer couplings attribute to orange-peel and RKKY-like couplings. Interlayer coupling can be interpreted as a minor loop shift of the Co soft layer, and we demonstrate RKKY-like coupling as the oscillatory behavior of the Hmls. Also, the interlayer coupling can be revealed by direct observation of the domain behavior in the Co soft layer. The presence of interlayer coupling gives rise to a difference in domain wall speed and asymmetry magnetization reversal at the equal driving force. We found that the Co soft layer and the hard layer favor the anti-parallel state by directly observing the magnetization reversal of domains with different magnetization orientations.
We demonstrate that the strength of the stray field emanating from the CoS and CoH layers depends on the magnetization configuration adjacent to the DW of the CoH layer. The DW is pinned by stray fields emanating from the CoS and CoH layers across the Pt spacer. The DW pinning was evidenced by Kerr imaging. This thesis provides a comprehensive understanding of the interlayer exchange coupling on the Pt/Co bilayers with PMA and offers a potential implementation into magnetoresistive devices and magnetic memories.clos
Multiscale simulation study on the improvement of conventional electrolytes for rechargeable batteries
Get PDFSchool of Energy and Chemical Engineering (Chemical Engineering)ope
?????? ?????? ?????? ????????? ?????? ????????? ????????? ?????? ?????? ?????? ???????????? ???????????? ?????? ??????????????? ??? ????????? ?????? ??????
Get PDFDepartment of Electrical EngineeringBlood Pressure is the most important physiological signal because it is highly associated with various cardiovascular diseases, and it is the basic index to monitor these kinds of diseases. Among various blood pressure measurement methods, Vascular Transit Time (VTT)-based blood pressure estimation method utilizes Photoplethysmogram (PPG), which is the physiological signal from blood volume changes using light, and Phonocardiogram (PCG), which is the cardiac signal corresponding to heart beat sound. In order to continuous blood pressure monitoring in our daily life, real-time VTT monitoring system with PPG and PCG integrated circuit would be excellent solution. The requirements of this system include low-power consumption for long-term operation with a small-sized battery, low-noise characteristic for accurate signal acquisition, and unobstructed measurement position in order to avoid cumbersome. Also, VTT is defined as the time interval between the PCG peak and the PPG peak, peaks of the two signals are needed in VTT calculation. This master???s thesis proposes a readout integrated circuit (ROIC) which achieves low-power consumption, low-noise performance, and easy signal processing. PPG ROIC consists of an LED driver that can control the intensity of the LED and a Light-to-Digital Converter (LDC) operating with a dual-slope mechanism, which converts the photodiode current from a reflected light to digital data. Since the LDC directly converts light into digital, it does not require an additional ADC and suitable for low-power applications. During dual-slope operation, the noise performance is limited by the main integrator and comparator. For low-noise operation, the noise of the core amplifier of the main integrator is reduced by the chopper-stabilization technique, and the quantization noise of the comparator is shaped through a noise-shaping loop. In the noise-shaping loop, the practical noise-shaping performance is reduced due to the charge sharing problem that occurs during the operation of residue voltage save. To solve this problem, an improved noise-shaping loop is proposed. To calculate the VTT, the PCG peak and PPG peak are determined through the digital signal processing in the microcontroller unit (MCU). However, the process of determining the peak only through the shape of the signal waveform alone requires significant processing burden and high-power consumption in the MCU. Especially, for PCG signals, it is very challenging to distinguish different peaks of similar amplitude. Thus, the PCG S1 peak and PCG S2 peak are distinguished respectively using the proposed dual peak detector consisting of a parallel envelope detector and a sort algorithm block. The system for monitoring blood pressure on the chest is implemented with the proposed PPG and PCG integrated circuits. The prototype operates using a battery, and can measure a small chest PPG signal and obtain a PCG signal without using a stethoscope. Through this, it achieves a miniaturized size and convenient usability that can be implemented in wearable devices. The measurement results can be checked in PC MATLAB program using Bluetooth. Furthermore, the proposed system supports not only VTT-based blood pressure monitoring mode but also digital stethoscope mode and heart rate monitoring mode. For verification, a comparison of the blood pressure measurement results with a commercial blood pressure gauge is conducted, and it is confirmed that the proposed system is excellent in blood pressure estimation.ope
????????? ???????????? ???????????????????????? ??????????????????
Get PDFSchool of Energy and Chemical Engineering (Chemical Engineering)The ability of the human body to regulate its temperature is crucial for adjusting to environmental changes in temperature. Smart textiles which respond to stimuli can effectively shield the human body???s temperature from an environment that is always changing. For adaptive modulation of IR and water transmission on human skin, a smart textile based on shape memory polymer (SMP) fibers is proposed. To improve thermal insulation performance, an SMP textile is made with hierarchical micro/nanoporous structures. Also, silver nanowires are coated on one side to produce asymmetric IR reflectance and hydrophilicity. The porous SMP textile exhibits excellent tenability over its asymmetric wettability and thermal insulation through the deformation and recovery of shape and structure in response to stimuli. Due to adaptive IR reflectivity, the surface temperature of SMP textile on a hot plate is successfully managed in the IR images, and the degree of thermal insulation is regulated by 65.7% of the original value. In addition, the shape of the SMP textiles, which can be applied to remove sweat from human skin, can be used to switch on or off the directional transportation of water droplets. This water- and IR-gating smart textile can suggest a feasible approach of protecting the human skin from environmental changes.ope
Atad5??? ??? ??????????????? ????????? ?????? ??????
Get PDFDepartment of Biomedical EngineeringDuring brain development, there are a lot of endogenous replication stresses. Therefore, proper DNA repair pathways are essential to counteract various endogenous DNA damage that constantly challenges genomic integrity. Atad5 is important for maintaining genomic integrity by regulating PCNA unloading from chromatin using its ATPase activity during replication. In a previous study, Atad5 homozygous mutant mice showed embryonic lethality. Therefore, to understand the role of the Atad5 gene in brain development, we generated Atad5 brain-specific knockout mice by using the Nestin-Cre recombinase system. Atad5 brain-specific knockout mice were viable, but body weights and brain size were significantly in adult mice. In particular, the cell numbers in the hippocampus and cerebellum were decreased. Furthermore, Atad5 brain-specific knockout adult mice showed eye defects and distinct skull structure. Consistently, p53, which is a major regulator of DNA damage response, was upregulated in the Atad5 brain-specific knockout mouse. Taken together, our data suggest that Atad5, which has a PCNA unloading function, have a role in brain development.ope