<p dir="ltr">The research focuses on improving natural language processing (NLP) models by integrating the hierarchical structure of language, which is essential for understanding and generating human language. The main contributions of the study are:</p><ol><li><b>Hierarchical RNN Model:</b> Development of a deep Recurrent Neural Network model that captures both explicit and implicit hierarchical structures in language.</li><li><b>Hierarchical Attention Mechanism:</b> Use of a multi-level attention mechanism to help the model prioritize relevant information at different levels of the hierarchy.</li><li><b>Latent Indicators and Efficient Training:</b> Integration of latent indicators using the Expectation-Maximization algorithm and reduction of computational complexity with Bootstrap sampling and layered training strategies.</li><li><b>Sequence-to-Sequence Model for Translation:</b> Extension of the model to translation tasks, including a novel pre-training technique and a hierarchical decoding strategy to stabilize latent indicators during generation.</li></ol><p dir="ltr">The study claims enhanced performance in various NLP tasks with results comparable to larger models, with the added benefit of increased interpretability.</p>

No description available.

No description available.

Sediment, an integral part of rivers and watersheds, is eroded from, stored in, and transported through various watershed components. Rivers often receive sediment in the form of episodic, discrete pulses from a variety of natural and anthropogenic processes, this sediment can be transported downstream along the bed or suspended in the water column. Most sediment measurements are focused on the component suspended in the water column. Recent advances in data collection techniques have substantially increased both the resolution and spatial scale of data on suspended sediment dynamics, which is helpful in linking small, site-scale measurements of transport processes in the field with large-scale modeling efforts. Part of this research evaluates the accuracy of the latest laser diffraction instrument for suspended-sediment measurement in rivers, LISST-SL2 for measuring suspended sediment concentration (SSC), particle size distribution (PSD), and velocity by comparing to concurrent physical samples analyzed in a lab for SSC and PSD, and velocity measured using an acoustic Doppler current profiler (ADCP) at 11 sites in Washington and Virginia during 2018-2020. Another part of this work employs a 1-D river network, bed material transport model to investigate the magnitude, timing, and persistence of downstream changes due to the introduction of sediment pulses in a linear river network. We specifically focus on comparing bed responses between mixed and uniform grain size sediment pulses. Then the model capability is utilized to explore the control of hydrograph structure on debris flow sediment transport through a more complex river network at different time horizons. Another part of this work investigates the effect of differences in spatial distribution of debris flow sediment input to the network by analyzing corresponding tributary and mainstem characteristics. Based on an extensive dataset, our results highlight the need for a correction of the raw LISST-SL2 measurements to improve the estimation of effective density and particle size distribution with the help of a physical sample. Simulation results from the river network model show that bed response is primarily influenced by the sediment-pulse grain size and distribution. Intermediate mixed-size pulses are likely to have the largest downstream impact because finer sizes translate quickly and coarser sizes (median bed gravel size and larger) disperse slowly. Furthermore, a mixed-size pulse, with a smaller median grain size than the bed, increases bed mobility more than a uniform-size pulse. While investigating the hydrologic control on debris flow simulation, this study finds that differences between transport by a 30-year daily hydrograph and simplified hydrographs were greatest in the first few years, but errors decreased to around 10% after 10 years. Our simulation results highlight that the sequence of flows (initial high/low flow) is less important for transport of finer sediment. We show that such network-scale modeling can quantitatively identify geomorphically significant network characteristics for efficient transport from tributaries to the mainstem, and eventually to the outlet. Results suggest that watershed area and slope characteristics are important to predict aggradation hotspots in a network. However, to predict aggradation and fluvial geomorphic responses to variations in sediment supply from river network characteristics more confidently, more widespread (in several other river networks) model applications with field validation would be useful. This work has important implications for river management, as it allows us to better predict geomorphically significant tributaries and potential impact on downstream locations, which are important for river biodiversity. Model results lead the way to use of simplified flow hydrographs for different timescales, which is crucial in large-scale modeling as it is often restricted by computational capacity. Finally, given the ability for reliable quantification of a high-resolution time-series of different suspended-sediment characteristics, in-stream laser diffraction offers great potential to advance our understanding of suspended-sediment transport. / Doctor of Philosophy / Rivers receive sediment from different natural and human sources, and water moves this sediment in various ways. These ways include along the bottom of the stream or suspended in the water. Quantifying suspended sediment in streams is an important step to estimate the threat to riverine environments as suspended sediments not only carry chemicals and pollutants, but also interact with the river bottom to affect the characteristics of streams. Measurement of suspended-sediment concentration and particle-size is critical for many engineering, ecological, and river-structure issues, but obtaining an accurate measurement of sediment quantity in a river is challenging. The recent advancement of a laser diffraction instrument allows us to obtain frequent measurements of suspended-sediment concentration and particle size by volume. We applied the most recent such instrument at 11 sites in Washington and Virginia during 2018-2020, along with concurrent water samples to measure suspended-sediment concentration and particle size by mass in a laboratory. Our analysis suggests that at least one supporting physical mass measurement be obtained to improve the estimation from laser measurement. Beside this site-scale measurement, we apply a large-scale river network model to estimate how sediment moves along the bed of rivers at large spatial extents. We simulate how this added sediment results in downstream changes in the amount of sediment in the river channel. We compare observed changes in the elevation of the stream bottom and sediment accumulation rates in a downstream lake to model results. Then we investigate the magnitude, timing, and persistence of downstream changes due to the introduction of added sediment by comparing the changes against a baseline condition (without the added sediment). We find that the added sediment that is half as large as on the river bottom and with a range of sizes are likely to affect the largest downstream changes because smaller sizes move quickly and larger sizes move slowly. Furthermore, added sediment that is smaller than on the river bottom and with a range of sizes help more sediment on the river bottom move than if that sediment addition all had the same particle size. We also employ this model to explore the effect of flow variation and river characteristics on sediment movement. Comparing between a 30-year flow record and simplified flow records, we show that results from simplified flow records vary initially, but errors decrease after 10 years. That is, both flow records result in similar sediment movement in the long-term. In terms of aggradation from added sediment, results show that the characteristics of elevation change of the river bottom play a vital role along with the contributing landscape area. This work has important implications for river management, as it not only allows us to accurately measure suspended sediment with an advanced instrument, but also better understand how rivers and aquatic habitat are affected by variations in added sediment.

Rapidly-increasing medium-voltage power electronics applications in emerging industry systems, including electrical ships, more electric aircraft, and microgrids, have emphasized the critical need for highly energy-efficient, reliable, and fast switching devices. As a result, Wide-Bandgap (WBG) devices have gained considerable interest over conventional silicon-based switches in recent years. For example, emerging WBG devices have unlocked new dimensions for modern motor drive systems with increased efficiency, switching frequency, and superior power density. Commercially-developed WBG devices such as Silicon Carbide (SiC) and Gallium Nitride (GaN) offer promising opportunities to meet those pressing requirements. However, the fast switching operation of WBG devices may cause substantially increased EMI emissions in medium-voltage applications, which can decrease the overall system’s performance or merits of power converters. This will be particularly an issue in a system where electric ground is unavailable, such as an electric ship, as a large Electro-Magnetic Interference current will be circulating within the system. The EMI in the WBG switch module will be emitted up to 500 MHz. This is the near radio-frequency (RF) band whose impact had not been clearly understood or properly analyzed in the power electronics field until recently. With new and critical challenges in recent years, to reliably adopt WBG devices in emerging power systems, there has been significant effort to improve electromagnetic compatibility (EMC) with new EMI mitigation techniques that comply with existing standards, including International Special Committee on Radio Interference (CISPR), Federal Communications Commission (FCC), Department of Defense (DOD), International Electro-Technical Commission (IEC), etc. This research investigates the common-mode EMI in networked power electronics-enabled power systems. Common-mode EMI phase information is a vital degree of freedom in EMI study that has not been considered in the state of the art. The EMI phase information reduces EMI without implementing any active or passive filter circuit. An effective and less complex method is introduced to reduce EMI in power electronics network. The work includes developing hybrid filter with passive and virtual filter. Including virtual filter reduces the passive common mode choke weight and volume significantly. Finally, a simplified switching node capacitance characterization technique for packaged WBG SiC has been introduced.

No description available.

Growing environmental concerns are causing a large transformation within the energy industry. Within the gas turbine industry, there is a large drive to develop improved modern dry-low emission combustion systems. The aim is to enable gas turbines to run on green fuels like hydrogen, while still keeping emission as NOx down. To design these systems, a thorough understanding of the aerothermal and kinetic processes within the combustion system of a gas turbine is essential. The goal of the thesis was to develop a one-dimensional general network model of the combustion system of Siemens Energy SGT-700, which accurately could predict pressure losses, mass flows, key temperatures, and emissions. Three models were evaluated and a code that emulated some aspects of the control system was developed. The models and the code were evaluated and compared to each other and to test data from earlier test campaigns performed on SGT-700 and SGT-600. Simulations were also carried out with hydrogen as the fuel.  In the end, a model of the SGT-700 combustion chamber was developed and delivered to Siemens Energy. The model had been verified against test data and predictions made by other Siemens Energy thermodynamic calculation software, for a range of load conditions. The preforms of the model, when hydrogen was introduced into the fuel mixture, were also tested and compared to test data / En växande medvetenhet kring klimatfrågan, har medfört stora förändringar i energibranschen. I och med detta behöver även gasturbinindustrin förbättra de nuvarande dry-low emissions systemen och göra det möjligt för gasturbiner att förbränna gröna bränslen som väte. Samtidigt måste också utsläppen av NOx hållas nere. För att kunna utforma dessa system behövs en fullständig förståelse för de aerotermiska och kinetiska processerna i en gasturbins förbränningskammare. Målet med detta examensarbete var att utveckla en endimensionell generell nätverksmodell för förbränningssystemet i Siemens Energys SGT-700. Modellen skulle noggrant kunna förutsäga tryckförluster, massflöden, viktiga temperaturer samt utsläpp. Tre modeller utvärderades och en kod som emulerade vissa aspekter av styrsystemet utvecklades också. Modellerna och koden utvärderades och jämfördes mot varandra och även mot testdata från tidigare testserier som utfördes på SGT-700 och SGT-600. Simuleringar utfördes också med väte som bränsle. Slutligen levererades en modell av SGT-700 förbränningskammaren till Siemens Energy. Modellen har verifierats för en rad olika lastfall, mot testdata och data som genererats av andra termodynamisk beräkningsprogram som utvecklats av Siemens Energy. Hur modellen uppförde sig när väte var introducerat in i olika lastfall jämfördes också mot testdata

A abordagem de modos normais de baixa frequência na descrição das flutuações conformacionais dos estados nativos das proteínas globulares tem ajudado na caracterização das suas funções biológicas. Vários métodos teóricos e experimentais têm sido empregados para a determinação destas flutuações internas. Estes movimentos podem ser caracterizados pelo fator Debye-Waller (fator-B), correspondente à mobilidade local do resíduo em nível atômico. A análise de modos normais utilizando os modelos de rede elástica (ENM) demonstra ser uma técnica robusta. Fatores-B experimentais são reproduzidos teoricamente por meio desta técnica em tempos computacionais relativamente curtos, mostrando-se competitiva com as técnicas mais sofisticadas. O modelo de rede elástica é uma abordagem t ipo coarse-grain na qual a proteína no seu estado enovelado é representada por uma rede elástica tridimensional de carbonos conectados por molas. As molas representam as interações ligantes e não ligantes entre os carbonos . Neste trabalho, inicialmente, estudamos os modelos de rede elástica já conhecidos na literatura. Em seguida, realizamos um estudo comparativo entre eles. Neste estudo, comprovamos que os modelos pfGNM e pfANM apresentam melhor correlação com os fatores-B experimentais que os os modelos GNM e ANM tradicionais. Desenvolvemos também uma nova abordagem, a qual intitulamos número de contatos ponderados anisotrópica (AWCN). Mostramos que a abordagem AWCN apresenta um desempenho significativamente melhor que o modelo de rede elástica anisotrópica tradicional. Por fim, realizamos um estudo de caráter investigativo do comportamento do peso das interações entre resíduos. Este estudo re velou que, para os modelos WCN e AWCN, a correlação exibe o seu valor máximo para interações ponderadas $1/R^p$, entre resíduos $i$ e $j$j, para valores de $p$ em torno de 2. Nos modelos pfGNM e pfANM a correlação é maximizada para dois valores de $p$, o primeiro em torno de 2 e o segundo em torno de 4,75, indicando que a ponderação pelo recíproco do quadrado da distância, usualmente empregada na literatura, pode não ser adequada para obter a melhor correlação. / Low frequency normal mode approach to describe conformational fluctuations of globular proteins has helped to characterize their biological functions. Various theoretical and experimental methods have been employed to det ermine the magnitudes of those internal motions. Those motions can be characterized by the Debye-Waller factor (B-factor), co rresponding to the local mobility of the residue at the atomic level. Normal mode analysis using elastic network models (ENM) has demonstrated to be a robust technique. Experimental B-factors has been reproduced theoretically by means of this techniq ue in a short computational time and it has been shown to be competitive with more sophisticated techniques. The ENM is a coarse-grained approach in which the protein is represented by a three-dimensional elastic network of alpha-carbon atoms connect ed by springs. Springs represent bonded and non-bonded interactions between the alpha-carbon atoms. In this work, we study th e elastic network models known in the literature. Next, we perform a comparative study between them. We show that the pfGNM a nd pfANM models present better correlation with experimental B-factors than the traditional GNM and ANM models. We also devel op a new approach, which we entitled anisotropic weighted contact number (AWCN). We show that it presents results significantly better than the traditional anisotropic elastic network model. Finally, we perform a study of investigative character of the behavior for the weight of the interactions between residues. This study revealed that, for the WCN and AWCN models, the correlation exhibits its maximum value for weighted interactions $1/R^p$, between residues $i$ and $j$, for values of $p$ around 2. In the pfGNM and pfANM models the correlation is max imized for two values of $p$, the first one around 2 and the second one around 4.75. This indicates that the weighting by the reciprocal of the square of the distance, usually employed in the literature, may not be appropriate to obtain the best correlation.

The understanding of multiphase flow properties is essential for the exploitation of hydrocarbon reserves in a reservoir; these properties in turn are dependent on the geometric properties and connectivity of the pore space. The determination of the pore size distribution in carbonate reservoirs remains challenging; carbonates exhibit complex pore structures comprising length scales from nanometers to several centimeters. A major challenge to the accurate evaluation of these reservoirs is accounting for pore scale heterogeneity on multiple scales. This is the topic of this thesis. Conventionally, this micron scale information is achieved either by building stochastic models using 2D images or by combining log and laboratory data to classify pore types and their behaviour. None of these capture the true 3D connectivity vital for flow characterisation. We present here an approach to build realistic 3D network models across a range of scales to improve property estimation through employment of X-ray micro-Computed Tomography (μCT) and Focussed Ion Beam Tomography (FIBT). The submicron, or microporous, regions are delineated through a differential imaging technique undertaken on x-ray CT providing a qualitative description of microporosity. Various 3-Phase segmentation methods are then applied for quantitative characterisation of those regions utilising the attenuation coefficient values from the 3D tomographic images. X-ray micro-CT is resolution limited and can not resolve the detailed geometrical features of the submicron pores. FIB tomography is used to image the 3D pore structure of submicron pores down to a scale of tens of nanometers. We describe the experimental development and subsequent image processing including issues and difficulties resolved at various stages. The developed methodology is implemented on cores from producing wackstone and grainstone reservoirs. Pore network models are generated to characterise the 3D interconnectivity of pores. We perform the simulations of petrophysical properties (permeability and formation resistivity) directly on the submicron scale image data. Simulated drainage capillary pressure curves are matched with the experimental data. We also present some preliminary results for the integration of multiscale pore information to build dual-scale network models. The integration of multiscale data allows one to select appropriate effective medium theories to incorporate sub-micron structure into property calculations at macro scale giving a more realistic estimation of properties.

With increasing complexity of microwave integrated circuits and tendency towards building integrated modules, real estate in printed circuit boards becomes more at premium. On the other hand, building MIC's on a single semiconductor substrate such as GaAs has other drawbacks as substrate requirements for different components are sometimes contradictory. This has motivated researchers to consider multi-layer and stacked designs. Multi-layer planar circuits offer advantages that cannot be equaled by traditional single layer designs. In this respect, a new class of planar structures, based upon a multi-layered stack of dual-mode stripline or microstrip patches is becoming increasingly popular. In the new stacked design coupling between planar circuits separated by a ground plane is accomplished through coupling apertures in the common ground plane. <br ><br /> This thesis is about developing a new approximate multiport network model for fast analysis of multi-layered planar structures with ground plane slots. To extend applicability of multiport network model (MNM) to the class of planar structures containing ground plane slots, a generalized network formulation for aperture problems is combined with traditional MNM to account for the presence of the slot. To this end, the slot is replaced by an unknown equivalent surface magnetic current. Slot ports are defined in terms of electric and magnetic fields over the slot in accordance with the generalized network formulation for aperture problems. While traditional MNM for planar circuits is based on generalized impedance matrices, we adopt a hybrid matrix approach for multi-layer structures. The hybrid matrix consists of four sub-matrices that relate terminal voltages and currents of edge and slot ports. The same generalized impedance matrix in the absence of the slot can be used to relate terminal voltages and currents of edge ports when the slot ports are short-circuited. Open circuit voltage at edge ports due to terminal voltages at slot ports and terminal currents at slot ports due to input currents at edge ports are represented by two transfer matrices. Both these transfer matrices can be calculated from 2D analysis which only considers <em>TM^z</em> modes. <br ><br /> Interaction among slot ports, represented by a generalized admittance matrix, however, requires considering both <em>TM^z</em> and <em>TE^z</em> modes. This generalized admittance matrix is obtained from tangential component of the magnetic field over the slot due to the equivalent surface magnetic current and relates terminal voltages and currents of slot ports. Full modal expansion consisting of both <em>TM^z</em> and <em>TE^z</em> modes is used to compute the generalized admittance matrix of a slot in a regularly shaped planar cavity. For irregularly shaped patches, modal expansion is not available. Instead, a new contour integral equation for magnetic field, derived for the first time in this thesis, is combined with complex images method for calculation of generalized admittance matrix of a slot radiating in a planar cavity of arbitrary shape. <br ><br /> Once the hybrid matrix representation of a planar circuit on a ground plane containing a slot is derived, it can be connected to the hybrid matrix of any other planar circuit on the other side of the ground plane. This can be done by enforcing network equivalent of continuity of tangential fields across the slot. This leads to a generalized impedance matrix for the multi-layer structure relating terminal voltages and currents of edge ports of both planar circuits. <br ><br /> To show the accuracy of the proposed method of analysis, several proof-of-concept structures have been analyzed by both this method and ANSOFT HFSS full-wave simulator as a reference. In most cases excellent agreement is achieved in predicting the return loss and radiation patterns of these multi-layer structures which proves the validity of the proposed approach for fast analysis and design of multi-layer planar structures.