Made available in DSpace on 2014-06-11T19:30:20Z (GMT). No. of bitstreams: 0 Previous issue date: 2011-08-23Bitstream added on 2014-06-13T21:00:52Z : No. of bitstreams: 1 garcia_rd_me_botfca.pdf: 728744 bytes, checksum: 4e020da519ee70d29e2ebfa30b7c2d97 (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / No Brasil observa-se que o cultivo do eucalipto contribui expressivamente na economia nacional, isso se dá em decorrência de elevadas taxas de produtividade das florestas existentes no território nacional. Analisando então, uma floresta produtiva se inicia através de mudas com boa qualidade morfológicas e resistentes às adversidades fisiológicas apresentadas pelo meio, para se definir como sendo de boa qualidade uma série de fatores são considerados essenciais no processo de produção desta muda, e o manejo hídrico é um fator primordial a ser analisado. Os estudos sobre manejo hídrico em viveiros de produção são escassos e desatualizados em decorrência das diversas descobertas no âmbito clonal de produção de mudas. Neste sentido o trabalho teve por objetivo verificar a influência da lâmina e a frequência de irrigação, aplicadas as mudas de Eucalyptus grandis vs. Eucalyptus urophylla em ambiente controlado, com o foco final na formação de uma muda de melhor qualidade para as mais diversas necessidades do mercado florestal. O experimento foi conduzido no Viveiro do Departamento de Recursos Naturais/Setor de Ciências Florestais, na Fazenda Experimental Lageado, pertencente à Faculdade de Ciências Agronômicas (FCA) da UNESP, no município de Botucatu – SP. O trabalho constituiu-se de dois experimentos sendo um com o clone 144 (mais resistente ao déficit hídrico) e o outro com o clone 103 (mais suscetível ao déficit hídrico). O substrato utilizado foi o produto comercial denominado Carolina Soil Florestal®. O delineamento estatístico foi um fatorial (3x2) inteiramente casualizado constituído dos seguintes fatores: 2 frequências de irrigação (2 e 5 vezes ao dia) e 3 lâminas brutas de irrigação (8, 12 e 16 mm). O experimento foi composto por 06 tratamentos, cada qual composto por... / In Brazil we found that each passing day the national economy depends on the cultivation of eucalyptus. Thus a productive forest comes from a good quality seedlings. This study aimed to investigate the influence of intensity and frequency of irrigation applied to seedlings of Eucalyptus grandis vs. Eucalyptus urophylla in a controlled environment with a focus on the formation of a final change to better quality, both in nursery and field workers. The experiment was conducted in the Nursery Department of Natural Resources / Division of Forest Science at the Experimental Farm Lageado belonging to the Faculdade de Ciências Agronômicas (FCA) of UNESP, Botucatu - SP. The substrate used was a commercial product called Carolina Forest Soil. The design was a factorial (3x2x2) randomized consists of the following factors: 3 intensity of irrigation (8, 12 and 16 mm), two irrigation frequencies (2 and 5 x day), two hybrid clones of Eucalyptus grandis vs. Eucalyptus urophylla. The experiment consisted of 12 treatments, each treatment with 04 replications, each consisting of 48 plants (24 plants useful per plate), totaling 96 plants per treatment. The seedling were evaluated as: aerial part height (H); diameter of collect (DC); aerial part... (Complete abstract click electronic access below)

Neste trabalho, são descritos os detalhes da montagem de um sistema de cultivo microbiano em estado sólido, cujo elemento principal é um biorreator de tambor rotativo em escala de bancada, com casco em vidro refratário com aproximadamente 6,2 litros de volume. A montagem do equipamento exigiu estudos para a definição dos seguintes aspectos: geometria do reator; controle de frequência e período de agitação; umidificação, aquecimento e controle do fluxo do ar injetado no sistema; mistura do meio de cultivo; medição e controle da temperatura do meio; retirada de amostras. O sistema desenvolvido foi avaliado, tendo como processo fermentativo modelo o cultivo de Aspergillus niger T0005/007-2, microrganismo produtor de enzimas pectinolíticas. Os testes foram feitos em um meio contendo farelo de trigo como suporte sólido, avaliando-se a influência de três parâmetros principais agitação, massa de meio de cultivo e temperatura do meio sobre o crescimento celular e a produção de endo-poligalacturonase (PG) por A. niger. Três formas de agitação do meio sólido foram comparadas: sem agitação; agitação a 1 rpm por 5 minutos a cada duas horas; agitação a 1 rpm por 1 hora e 55 minutos a cada duas horas. A segunda forma de agitação levou ao melhor crescimento celular, 81 mg.g-1, e a atividade de endo-PG da ordem de 80 U.g-1, semelhante ao estimado com o sistema estático. Com a terceira forma de agitação, aparentemente houve dano ao micélio fúngico e, com isso, resultados inferiores foram alcançados. A avaliação do efeito da massa de meio de cultivo sobre o processo foi feita com cargas crescentes de meio que resultaram na ocupação de 30, 45 e 60 % do volume útil do reator. Quanto ao crescimento celular, o melhor resultado foi alcançado com a menor carga de meio, enquanto que a carga intermediária resultou no mais alto título enzimático final, 107,2 U.g-1. Nos experimentos sobre a influência da temperatura sobre o cultivo de A. niger, a maior atividade enzimática (80,6 U.g-1) foi obtida numa condição de trabalho que permitiu que a temperatura do meio atingisse valores da ordem de 45ºC. Com o meio mantido a 30ºC, a atividade enzimática máxima foi substancialmente mais baixa, 46,4 U.g-1. Nos testes realizados com o processo modelo, não se verificou uma clara influência positiva da agitação sobre a produção de endo-PG. Entretanto, a partir dos experimentos com temperatura controlada, é possível sugerir que a formação de endo-PG é favorecida por uma condição de estresse para o microrganismo, no caso representado por temperaturas do meio acima de 40ºC, visto que a temperatura ideal de crescimento de A. niger encontra-se na faixa de 28 a 34ºC. Este resultado, que discorda do que é genericamente descrito na literatura especializada sobre cultivos em estado sólido, é um exemplo da importância de dispor-se de um biorreator de tambor rotativo de pequena escala como equipamento básico para a realização de estudos fundamentais a respeito deste tipo de processo. Adicionalmente, o fato de o corpo do reator ser construído em vidro permite a visualização do espaço interno do tambor rotativo e observar o efeito da agitação sobre o meio de cultivo. / In this work, details on the assembling of a solid state cultivation system, whose main component is a bench scale rotating-drum bioreactor, are described. The bioreactor was built in refractory glass and has an approximate volume of 6.2 liters. To assemble the equipment, the following aspects have been studied: geometry of the bioreactor; control of frequency and period of agitation; moistening, warming and controlling of the inlet air flux; mixing of cultivation medium; determination and control of medium temperature; sample withdrawn. The developed system was evaluated, being the cultivation of the pectinolytic enzymeproducing microorganism Aspergillus niger T0005/007-2 used as the model process. The fermentative tests were carried out in a medium containing wheat-straw as solid support and were used to evaluate the effects of three main parameters agitation, mass of cultivation medium and medium temperature on the cell growth and the production of endopolygalacturonase (endo-PG) by A. niger. For agitation of the solid medium, three modes were compared: no agitation; agitation of 1 rpm for 5 minutes each 2 hours; agitation for 1 hour and 55 minutes, each 2 hours. The second mode of agitation led to the best cell growth, 81 mg.g-1, and to endo-PG activity of approximately 80 U.g-1, similar to that obtained with the static system. With the third mode of agitation occurred, apparently, some damage to fungus mycelium and inferior results were achieved. The evaluation of the mass of cultivation medium on the process was done by loading the reactor with masses that result in the occupation of 30, 45 and 60% of the working volume of the bioreactor. With respect to the cell growth, the best result was attained with the smallest load of medium, whereas the intermediate load resulted in the highest endo-PG activity, 107.2 U.g-1. In the experiments on the influence of the temperature on the cultivation of A. niger, the largest endo-PG activity (80.6 U.g-1) was obtained in a process condition that allowed that the temperature reached values close to 45ºC. When the medium temperature was controlled at 30ºC, the endo-PG activity was substantially lower, 46.4 U.g-1. In the tests with the model process, no clearly positive influence of agitation on the production of endo-PG was observed. From the temperature-controlled experiments however, it is possible to suggest that endo-PG formation is favored by a stress condition for the microorganism, represented in this case by medium temperatures over 40ºC, since the optimal temperatures for A. niger growth is found in the range of 28 to 34ºC. This result disagrees from what is generically described in the specialized literature for the solid state cultivations, being this fact an example of the importance of having a small-scale rotating-drum bioreactor as a basic equipment for fundamental studies on that type of fermentative process. Additionally, the fact of the reactor body being built on glass allows the inspection of the internal space of the rotating drum and to observe the effects of agitation on the cultivation medium.

Studies into sources of alternative liquid transport fuel energy have identified agro-industrial wastes, which are lignocellulosic in nature, as a potential feedstock for biofuel production against the background of depleting nonrenewable fossil fuels. In South Africa, large quantities of apple and other fruit wastes, called pomace, are generated from fruit and juice industries. Apple pomace is a rich source of cellulose, pectin and hemicellulose, making it a potential target for utilisation as a lignocellulosic feedstock for biofuel and biorefinery chemical production. Lignocellulosic biomass is recalcitrant in nature and therefore its degradation requires the synergistic action of a number of enzymes such as cellulases, hemicellulases, pectinases and ligninases. Commercial enzyme cocktails, containing some of these enzymes, are available and can be used for apple pomace degradation. In this study, the degradation of apple pomace using commercial enzyme cocktails was investigated. The main focus was the optimisation of the release of sugar monomers that could potentially be used for biofuel and biorefinery chemical production. There is no or little information reported in literature on the enzymatic degradation of fruit waste using commercial enzyme mixtures. This study first focused on the characterisation of the substrate (apple pomace) and the commercial enzyme cocktails. Apple pomace was found to contain mainly glucose, galacturonic acid, arabinose, galactose, lignin and low amounts of xylose and fructose. Three commercial enzyme cocktails were initially selected: Biocip Membrane, Viscozyme L (from Aspergillus aculeatus) and Celluclast 1.5L (a Trichoderma reesei ATCC 26921 cellulase preparation). The selection of the enzymes was based on activities declared by the manufacturers, cost and local availability. The enzymes were screened based on their synergistic cooperation in the degradation of apple pomace and the main enzymes present in each cocktail. Viscozyme L and Celluclast 1.5L, in a 50:50 ratio, resulted in the best degree of synergy (1.6) compared to any other combination. The enzyme ratios were determined on Viscozyme L and Celluclast 1.5L based on the protein ratio. Enzyme activity was determined as glucose equivalents using the dinitrosalicylic acid (DNS) method. Sugar monomers were determined using Megazyme assay kits. There is limited information available on the enzymes present in the commercial enzyme cocktails. Therefore, the main enzymes present in Viscozyme L and Celluclast 1.5L were identified using different substrates, each targeted for a specific enzyme and activity. Characterisation of the enzyme mixtures revealed a large number of enzymes required for apple pomace degradation and these included cellulases, pectinases, xylanases, arabinases and mannanases in different proportions. Viscozyme L contained mainly pectinases and hemicellulases, while Celluclast 1.5L displayed largely cellulase and xylanase activity, hence the high degree of synergy reported. The temperature optimum was 50ºC for both enzyme mixtures and pH optima were observed at pH 5.0 and pH 3.0 for Viscozyme L and Celluclast 1.5L, respectively. At 37ºC and pH 5.0, the enzymes retained more that 90% activity after 15 days of incubation, allowing the enzymes to be used together with less energy input. The enzymes were further characterised by determining the effect of various compounds, such as alcohols, sugars, phenolic compounds and metal ions at various concentrations on the activity of the enzymes during apple pomace hydrolysis. Apart from lignin, which had almost no effect on enzyme activity, all the compounds caused inhibition of the enzymes to varying degrees. The most inhibitory compounds were some organic acids and metal ions, as well as cellobiose and xylobiose. Using the best ratio for Viscozyme L and Celluclast 1.5L (50:50) for the hydrolysis of apple pomace, it was observed that synergy was highest at the initial stages of hydrolysis and decreased over time, though the sugar concentration increased. The type of synergy for optimal apple pomace hydrolysis was found to be simultaneous. There was no synergy observed between Viscozyme L and Celluclast 1.5L with ligninases - laccase, lignin peroxidase and manganese peroxidase. Hydrolysing apple pomace with ligninases prior to addition of Viscozyme L and Celluclast 1.5L did not improve degradation of the substrate. Immobilisation of the enzyme mixtures on different supports was performed with the aim of increasing stability and enabling reuse of the enzymes. Immobilisation methods were selected based on the chemical properties of the supports, availability, cost and applicability on heterogeneous and insoluble substrate like apple pomace. These methods included crosslinked enzyme aggregates (CLEAs), immobilisation on various supports such as nylon mesh, nylon beads, sodium alginate beads, chitin and silica gel beads. The immobilisation strategies were unsuccessful, mainly due to the low percentage of immobilisation of the enzyme on the matrix and loss of activity of the immobilised enzyme. Free enzymes were therefore used for the remainder of the study. Hydrolysis conditions for apple pomace degradation were optimised using different temperatures and buffer systems in 1 L volumes mixed with compressed air. Hydrolysis at room temperature, using an unbuffered system, gave a better performance as compared to a buffered system. Reactors operated in batch mode performed better (4.2 g/L (75% yield) glucose and 16.8 g/L (75%) reducing sugar) than fed-batch reactors (3.2 g/L (66%) glucose and 14.6 g/L (72.7% yield) reducing sugar) over 100 h using Viscozyme L and Celluclast 1.5L. Supplementation of β- glucosidase activity in Viscozyme L and Celluclast 1.5L with Novozyme 188 resulted in a doubling of the amount of glucose released. The main products released from apple pomace hydrolysis were galacturonic acid, glucose and arabinose and low amounts of galactose and xylose. These products are potential raw materials for biofuel and biorefinery chemical production. An artificial neural network (ANN) model was successfully developed and used for predicting the optimum conditions for apple pomace hydrolysis using Celluclast 1.5L, Viscozyme L and Novozyme 188. Four main conditions that affect apple pomace hydrolysis were selected, namely temperature, initial pH, enzyme loading and substrate loading, which were taken as inputs. The glucose and reducing sugars released as a result of each treatment and their combinations were taken as outputs for 1–100 h. An ANN with 20, 20 and 6 neurons in the first, second and third hidden layers, respectively, was constructed. The performance and predictive ability of the ANN was good, with a R² of 0.99 and a small mean square error (MSE). New data was successfully predicted and simulated. Optimal hydrolysis conditions predicted by ANN for apple pomace hydrolysis were at 30% substrate (wet w/v) and an enzyme loading of 0.5 mg/g and 0.2 mg/mL of substrate for glucose and reducing sugar, respectively, giving sugar concentrations of 6.5 mg/mL and 28.9 mg/mL for glucose and reducing sugar, respectively. ANN showed that enzyme and substrate loadings were the most important factors for the hydrolysis of apple pomace.

Le développement des polymères conducteurs, surtout en termes de stabilité, a permis de les intégrer dans les dispositifs électroniques pour des applications à haute valeur ajoutée. C'est la raison pour laquelle les condensateurs tantale initialement basés sur la technologie MnO2, en tant que cathode, ont été améliorés avec le remplacement de cette dernière par un polymère organique conducteur. Ces nouveaux condensateurs tantale-polymère sont constitués d'une anode en tantale frittée, d'un film diélectrique en oxyde de tantale, et d'une cathode en polymère conducteur, typiquement le poly(3,4-éthylènedioxythiophène) (PEDOT). Le fonctionnement des condensateurs a d'abord été optimisé uniquement pour de faibles capacités par polymérisation in situ. L'étape suivante consiste à atteindre de plus grandes capacités par imprégnation du polymère conducteur pré-synthétisé. Notre travail a été axé sur la caractérisation fine et la fabrication de prototypes de condensateur. Pour mener à bien cette étude, chaque partie du condensateur tantale-polymère a été caractérisée par différentes techniques physico-chimiques. Nous avons, entre autres, étudié la microstructure du réseau de tantale et les propriétés de la solution commerciale de polymère conducteur pour déterminer les paramètres d'imprégnation des condensateurs. Les caractérisations effectuées au laboratoire sont complétées par une évaluation des performances électriques des prototypes fabriqués dans l'entreprise. Tout ce travail a contribué à la mise sur le marché d'une nouvelle gamme de condensateurs tantale-polymère par l'entreprise Exxelia Tantalum. En parallèle, une étude a été consacrée à la synthèse d'un nouveau couple de polymère plus performant dans le but de remplacer le polymère commercial. / The development of conducting polymers, especially in terms of environmental stability, has allowed them to be used in electronic devices for high value applications. That's why tantalum capacitors initially based on MnO2 cathode technology have been improved by the replacement of it with a conducting polymer. Tantalum-polymer capacitors consist of a sintered tantalum anode, an anodic tantalum oxide film as a dielectric, and a conductive polymer cathode made of poly (3,4-ethylenedioxythiophene) (PEDOT). Until recently, those capacitors have been optimized only for low capacities by in situ polymerization. The next step is to reach higher capacities using an impregnated conductive polymer. Our work focused on the characterization and fabrication of capacitors. The main study focused on the characterization of each part of the tantalum-polymer capacitor via physico-chemical investigations. We studied the microstructure of the tantalum network and the properties of the commercial polymer solution to determine parameters for the dip-coating of tantalum anodes. This laboratory characterization is complemented by an assessment of the electrical performances of samples within the company. All this work has contributed to a new range of tantalum-polymer capacitors by Exxelia Tantalum Company. At the same time, a study has been performed in the synthesis of a new pair of polymers in order to replace the commercial polymer.

Aluminum (Al) and its alloys are widely used in various technological applications, mainly due to the excellent thermal conductivity, non-magnetic, ecofriendly, easy formability and good recyclability. However due to the inferior corrosion resistance its applications are hampered in various engineering sectors. Besides, the corrosion related failures such as leakage of gas from pipeline, catastrophic breakdown of bridges and fire accidents in processing plants further puts the human life in jeopardy. Within the United States over $ 400 billion dollars per year are spent over research to understand and prevent the corrosion related failures. Recently, the development of transition metal(TM) aluminides (AlxTMy, where, TM = Mo, W, Ta, Nb, Cr, Zr and V) has received the global attention mainly due to high strength at elevated temperatures, light-weight, excellent corrosion and wear resistance. In light of this, surface modification via laser surface alloying (LSA) is a promising engineering approach to mitigate the corrosion and wear problems. In the present study the attempts are made to study the Al-Mo, Al-W, Al-Nb, and Al-Ta systems as a potential corrosion resistant coatings on aluminum. The refractory metal (Mo, W, Nb, Ta) precursor deposit was spray coated separately on aluminum substrate and was subsequently surface alloyed using a continuous wave diode-pumped ytterbium laser at varying laser energy densities. Microstructural analysis was conducted using scanning electron microscopy and further X-ray diffractometry was carried out to evaluate the various phases evolved during laser surface alloying. Corrosion resistance of laser alloyed coatings were evaluated using open circuit potential, cyclic potentiodynamic polarization, electrochemical impedance spectroscopy measurements were performed in 0.6 M NaCl solution (pH:6.9±0.2, 23˚C). Open circuit potential measurements indicate the more stable (steady state) potential values over long periods after laser surface alloying. Cyclic polarization results indicated reduction in the corrosion current density, enhancement in the polarization resistance, and increase in coating/protective efficiency with increase in laser energy density compared to untreated aluminum. Electrochemical impedance spectroscopy measurements also indicated an increase in charge transfer resistance after laser surface alloying of refractory metals on aluminum. Additionally, first principle calculations of thermodynamic, electronic and elastic properties of intermetallics evolved during LSA were also thoroughly investigated to correlate the corrosion performance of intermetallic coatings with these properties. The present study indicates that novel Al-Mo, Al-W, Al-Nb, and Al-Ta intermetallics has a great potential for light weight structural applications with enhanced corrosion resistance.

The current study focuses on phase stability and evolution in the titanium-zinc titanium-copper and titanium-antimony systems. The study utilized the Laser Engineering Net Shaping (LENS™) processing technique to deposit compositionally graded samples of three binary system in order to allow the assessment of phase stability and evolution as a function of composition and temperature the material is subjected to. Through LENS™ processing it was possible to create graded samples from Ti-xSb (up to 13wt%) and Ti-xCu (up to 16wt%). The LENS™ deposited gradient were solutionized, and step quenched to specific aging temperature, and the resulting microstructures and phase were characterized utilizing XRD, EDS, SEM, FIB and TEM. The Ti-Zn system proved incapable of being LENS™ deposited due to the low vaporization temperature of Zn; however, a novel processing approach was developed to drip liquid Zn onto Ti powder at temperatures above β transus temperature of Ti (882 ◦C) and below the vaporization temperature of Zn (907 ◦C). The product of this processing technique was characterized in a similar way as the graded LENS™ depositions. From measurements performed on Ti-Sb it seems that Sb could be a potential α stabilizer in Ti due to the presence of a mostly homogeneous α grains throughout the gradient; however, from XRD it can be understood that a titanium antimonide phase is present. From results obtained from the Ti-Zn samples, it can be surmised that the eutectoid reaction seems to be active, i.e. The eutectoid reaction is kinetically fast, as concluded by the presence of pearlitic structures. Finally, for the Ti-Cu system this work has been attempted to prove or disprove the existence of the Ti3Cu through the use of XRD and TEM SAD patterns. From XRD spectra collected there are peaks belonging to the Ti3Cu orthorhombic phase along with Ti2Cu and α-Ti phase. In addition to the Ti-Cu system displayed structures associated with divorced eutectoid decomposition mechanism, and at low undercooling seems to be prone to forming solid state dendrites.

The objective of this thesis is to investigate the effects of adding natural fiber (kenaf fiber, retted kenaf fiber, and sugarcane fiber) into polymer materials. The effects are obtained by considering three main parts. 1. Performance in thermoplastic composites. The effect of fiber retting on polymer composite crystallization and mechanical performance was investigated. PHBV/PBAT in 80/20 blend ratio was modified using 5% by weight kenaf fiber. Dynamic mechanical analysis of the composites was done to investigate the glass transition and the modulus at sub-ambient and ambient temperatures. ESEM was conducted to analyze fiber topography which revealed smoother surfaces on the pectinase retted fibers. 2. Performance in thermoset composites. The effect of the incorporation of natural fibers of kenaf and of sugarcane combined with the polyester resin matrix is investigated. A comparison of mechanical properties of kenaf polyester composite, sugarcane polyester composite and pure polyester in tensile, bending, dynamic mechanical thermal analysis (DMA) and moisture test on performance is measured.. 3. Performance in sandwich composites. The comparison of the performance characteristics and mechanical properties of natural fiber composites panels with soft and rigid foam cores are evaluated. A thorough test of the mechanical behavior of composites sandwich materials in tensile, bending and DCB is presented here.

Background: Statistical and machine learning techniques are now being incorporated into high-dimensional mixture research to overcome issues with traditional methods. Though some methods perform well on specific tasks, no method consistently outperforms all others in complex mixture analyses, largely because different methods were developed to answer different research questions. The research presented here concentrates on answering a single mixtures question: Are there exposure patterns within a mixture corresponding with sources or behaviors that give rise to exposure? Objective: This dissertation details work to design, adapt, and apply pattern recognition methods to environmental mixtures and introduces two methods adapted to specific challenges of environmental health data, (1) Principal Component Pursuit (PCP) and (2) Bayesian non-parametric non-negative matrix factorization (BN²MF). We build on this work to characterize the relationship between identified patterns of in utero endocrine disrupting chemical (EDC) exposure and child neurodevelopment. Methods: PCP---a dimensionality reduction technique in computer vision---decomposes the exposure mixture into a low-rank matrix of consistent patterns and a sparse matrix of unique or extreme exposure events. We incorporated two existing PCP extensions that suit environmental data, (1) a non-convex rank penalty, and (2) a formulation that removes the need for parameter tuning. We further adapted PCP to accommodate environmental mixtures by including (1) a non-negativity constraint, (2) a modified algorithm to allow for missing values, and (3) a separate penalty for measurements below the limit of detection (PCP-LOD). BN²MF decomposes the exposure mixture into three parts, (1) a matrix of chemical loadings on identified patterns, (2) a matrix of individual scores on identified patterns, and (3) and diagonal matrix of pattern weights. It places non-negative continuous priors on pattern loadings, weights, and individual scores and uses a non-parametric sparse prior on the pattern weights to estimate the optimal number. We extended BN²MF to explicitly account for uncertainty in identified patterns by estimating the full distribution of scores and loadings. To test both methods, we simulated data to represent environmental mixtures with various structures, altering the level of complexity in the patterns, the noise level, the number of patterns, the size of the mixture, and the sample size. We evaluated PCP-LOD's performance against principal component analysis (PCA), and we evaluated BN²MF's performance against PCA, factor analysis, and frequentist nonnegative matrix factorization (NMF). For all methods, we compared their solutions with true simulated values to measure performance. We further assessed BN²MF's coverage of true simulated scores. We applied PCP-LOD to an exposure mixture of 21 persistent organic pollutants (POPs) measured in 1,000 U.S. adults from the 2001--2002 National Health and Nutrition Examination Survey (NHANES). We applied BN²MF to an exposure mixture of 17 EDCs measured in 343 pregnant women in the Columbia Center for Children’s Environmental Health's Mothers and Newborns Cohort. Finally, we designed a two-stage Bayesian hierarchical model to estimate health effects of environmental exposure patterns while incorporating the uncertainty of pattern identification. In the first stage, we identified EDC exposure patterns using BN²MF. In the second stage, we included individual pattern scores and their distributions as exposures of interest in a hierarchical regression model, with child IQ as the outcome, adjusting for potential confounders. We present sex-specific results. Results: PCP-LOD recovered the true number of patterns through cross-validation for all simulations; based on an a priori specified criterion, PCA recovered the true number of patterns in 32% of simulations. PCP-LOD achieved lower relative predictive error than PCA for all simulated datasets with up to 50% of the data < LOD. When 75% of values were < LOD, PCP-LOD outperformed PCA only when noise was low. In the POP mixture, PCP-LOD identified a rank three underlying structure. One pattern represented comprehensive exposure to all POPs. The other two patterns grouped chemicals based on known properties such as structure and toxicity. PCP-LOD also separated 6% of values as extreme events. Most participants had no extreme exposures (44%) or only extremely low exposures (18%). BN²MF estimated the true number of patterns for 99% of simulated datasets. BN²MF's variational confidence intervals achieved 95% coverage across all levels of structural complexity with up to 40% added noise. BN²MF performed comparably with frequentist methods in terms of overall prediction and estimation of underlying loadings and scores. We identified two patterns of EDC exposure in pregnant women, corresponding with diet and personal care product use as potentially separate sources or behaviors leading to exposure. The diet pattern expressed exposure to phthalates and BPA. One standard deviation increase in this pattern was associated with a decrease of 3.5 IQ points (95% credible interval: -6.7, -0.3), on average, in female children but not in males. The personal care product pattern represented exposure to phenols, including parabens, and diethyl phthalate. We found no associations between this pattern and child cognition. Conclusion: PCP-LOD and BN^2MF address limitations of existing pattern recognition methods employed in this field such as user-specified pattern number, lack of interpretability of patterns in terms of human understanding, influence of outlying values, and lack of uncertainty quantification. Both methods identified patterns that grouped chemicals based on known sources (e.g., diet), behaviors (e.g., personal care product use), or properties (e.g., structure and toxicity). Phthalates and BPA found in food packaging and can linings formed a BN²MF-identified pattern of EDC exposure negatively associated with female child intelligence in the Mothers and Newborns cohort. Results may be used to inform interventions designed to target modifiable behavior or regulations to act on dietary exposure sources.

Speech separation has long been an active research topic in the signal processing community with its importance in a wide range of applications such as hearable devices and telecommunication systems. It not only serves as a fundamental problem for all higher-level speech processing tasks such as automatic speech recognition, natural language understanding, and smart personal assistants, but also plays an important role in smart earphones and augmented and virtual reality devices. With the recent progress in deep neural networks, the separation performance has been significantly advanced by various new problem definitions and model architectures. The most widely-used approach in the past years performs separation in time-frequency domain, where a spectrogram or a time-frequency representation is first calculated from the mixture signal and multiple time-frequency masks are then estimated for the target sources. The masks are applied on the mixture's time-frequency representation to extract the target representations, and then operations such as inverse short-time Fourier transform is utilized to convert them back to waveforms. However, such frequency-domain methods may have difficulties in modeling the phase spectrogram as the conventional time-frequency masks often only consider the magnitude spectrogram. Moreover, the training objectives for the frequency-domain methods are typically also in frequency-domain, which may not be inline with widely-used time-domain evaluation metrics such as signal-to-noise ratio and signal-to-distortion ratio. The problem formulation of time-domain, end-to-end speech separation naturally arises to tackle the disadvantages in the frequency-domain systems. The end-to-end speech separation networks take the mixture waveform as input and directly estimate the waveforms of the target sources. Following the general pipeline of conventional frequency-domain systems which contains a waveform encoder, a separator, and a waveform decoder, time-domain systems can be design in a similar way while significantly improves the separation performance. In this dissertation, I focus on multiple aspects in the general problem formulation of end-to-end separation networks including the system designs, model architectures, and training objectives. I start with a single-channel pipeline, which we refer to as the time-domain audio separation network (TasNet), to validate the advantage of end-to-end separation comparing with the conventional time-frequency domain pipelines. I then move to the multi-channel scenario and introduce the filter-and-sum network (FaSNet) for both fixed-geometry and ad-hoc geometry microphone arrays. Next I introduce methods for lightweight network architecture design that allows the models to maintain the separation performance while using only as small as 2.5% model size and 17.6% model complexity. After that, I look into the training objective functions for end-to-end speech separation and describe two training objectives for separating varying numbers of sources and improving the robustness under reverberant environments, respectively. Finally I take a step back and revisit several problem formulations in end-to-end separation pipeline and raise more questions in this framework to be further analyzed and investigated in future works.

Advances in data-driven applications, particularly artificial intelligence and deep learning, are driving the explosive growth of computation and communication in today’s data centers and high-performance computing (HPC) systems. Increasingly, system performance is not constrained by the compute speed at individual nodes, but by the data movement between them. This calls for innovative architectures, smart connectivity, and extreme bandwidth densities in interconnect designs. Silicon photonics technology leverages mature complementary metal-oxide-semiconductor (CMOS) manufacturing infrastructure and is promising for low cost, high-bandwidth, and reconfigurable interconnects. Flexible and high-performance photonic switched architectures are capable of improving the system performance. The work in this dissertation explores various photonic interconnected systems and the associated optical switching functionalities, hardware platforms, and novel architectures. It demonstrates the capabilities of silicon photonics to enable efficient deep learning training. We first present field programmable gate array (FPGA) based open-loop and closed-loop control for optical spectral-and-spatial switching of silicon photonic cascaded micro-ring resonator (MRR) switches. Our control achieves wavelength locking at the user-defined resonance of the MRR for optical unicast, multicast, and multiwavelength-select functionalities. Digital-to-analog converters (DACs) and analog-to-digital converters (ADCs) are necessary for the control of the switch. We experimentally demonstrate the optical switching functionalities using an FPGA-based switch controller through both traditional multi-bit DAC/ADC and novel single-wired DAC/ADC circuits. For system-level integration, interfaces to the switch controller in a network control plane are developed. The successful control and the switching functionalitiesachieved are essential for system-level architectural innovations as presented in the following sections. Next, this thesis presents two novel photonic switched architectures using the MRR-based switches. First, a photonic switched memory system architecture was designed to address memory challenges in deep learning. The reconfigurable photonic interconnects provide scalable solutions and enable efficient use of disaggregated memory resources for deep learning training. An experimental testbed was built with a processing system and two remote memory nodes using silicon photonic switch fabrics and system performance improvements were demonstrated. The collective results and existing high-bandwidth optical I/Os show the potential of integrating the photonic switched memory to state-of-the-art processing systems. Second, the scaling trends of deep learning models and distributed training workloads are challenging network capacities in today’s data centers and HPCs. A system architecture that leverages SiP switch-enabled server regrouping is proposed to tackle the challenges and accelerate distributed deep learning training. An experimental testbed with a SiP switch-enabled reconfigurable fat tree topology was built to evaluate the network performance of distributed ring all-reduce and parameter server workloads. We also present system-scale simulations. Server regrouping and bandwidth steering were performed on a large-scale tapered fat tree with 1024 compute nodes to show the benefits of using photonic switched architectures in systems at scale. Finally, this dissertation explores high-bandwidth photonic interconnect designs for disaggregated systems. We first introduce and discuss two disaggregated architectures leveraging extreme high bandwidth interconnects with optically interconnected computing resources. We present the concept of rack-scale graphics processing unit (GPU) disaggregation with optical circuit switches and electrical aggregator switches. The architecture can leverage the flexibility of high bandwidth optical switches to increase hardware utilization and reduce application runtimes. A testbed was built to demonstrate resource disaggregation and defragmentation. In addition, we also present an extreme high-bandwidth optical interconnect accelerated low-latency communication architecture for deep learning training. The disaggregated architecture utilizes comb laser sources and MRR-based cross-bar switching fabrics to enable an all-to-all high bandwidth communication with a constant latency cost for distributed deep learning training. We discuss emerging technologies in the silicon photonics platform, including light source, transceivers, and switch architectures, to accommodate extreme high bandwidth requirements in HPC and data center environments. A prototype hardware innovation - Optical Network Interface Cards (comprised of FPGA, photonic integrated circuits (PIC), electronic integrated circuits (EIC), interposer, and high-speed printed circuit board (PCB)) is presented to show the path toward fast lanes for expedited execution at 10 terabits. Taken together, the work in this dissertation demonstrates the capabilities of high-bandwidth silicon photonic interconnects and innovative architectural designs to accelerate deep learning training in optically connected data center and HPC systems.