In this dissertation, dimensionality reduction for hyperspectral remote sensing imagery is investigated to alleviate practical application difficulties caused by high data dimension. Band selection and band clustering are applied for this purpose. Based on availability of object prior information, supervised, semi-supervised, and unsupervised techniques are proposed. To take advantage of modern computational architecture, parallel implementations on cluster and graphics processing units (GPU) are developed. The impact of dimensionality reduction on the following data analysis is also evaluated. Specific contributions are as below. 1. A similarity-based unsupervised band selection algorithm is developed to select distinctive and informative bands, which outperforms other existing unsupervised band selection approaches in the literature. 2. An efficient supervised band selection method based on minimum estimated abundance covariance is developed, which outperforms other frequently-used metrics. This new method does not need to conduct classification during band selection process or examine original bands/band combinations as do traditional approaches. 3. An efficient semi-supervised band clustering method is proposed, which uses class signatures to conduct band partition. Compared to traditional unsupervised clustering, computational complexity is significantly reduced. 4. Parallel GPU implementations with computational cost saving strategies for the developed algorithms are designed to facilitate onboard processing. 5. As an application example, band selection results are used for urban land cover classification. With a few selected bands, classification accuracy can be greatly improved, compared to the one using all the original bands or those from other frequently-used dimensionality reduction methods.

Concerns arise in radiation therapy for head and neck cancers when dental prostheses are involved. These prostheses are high-density materials that induce image artifacts in computed tomography (CT) scans used for dose calculation. Two approaches are utilized in mitigating the impact of these artifacts on the accuracy of dose calculation. First, metal artifact reduction (MAR) algorithms or dual-energy CT scans are used to recover image quality. Second, a planning protocol is adopted whereby residual artifacts are manually contoured and assigned appropriate densities. This study evaluated the current planning process using a holistic approach. In this work, an axial section of a head phantom containing dental implants at the level of the oral cavity was constructed and scanned using various protocols on two different commercial scanners; Philips and Siemens, to assess the appearance of artifacts. An MVCT image set was merged with the corresponding kVCT image to improve visualization of the dental implants for use in density overrides. Three ion chamber measurement points in the simulated mouth facilitated the determination of measured dose which was compared to calculated dose at various single beam irradiation geometries. The influence of density override values on agreement between calculation and measurement was investigated for each geometry and imaging modality. Percent error was computed, and initial results compared to results manipulated by use of; a CT density table (Head); density overrides of walls and wax; and density overrides of walls, wax, and effective density of saturation regions. The study established that normal tissue doses are not significantly affected by metal artifact reduction (MAR) algorithms, and improvements in dose calculation compared to uncorrected CT images are small. Furthermore, the inclusion of a MVCT image set improved implant visualization reducing the treatment planning time while providing more information. Evidence led to the deduction that manual overrides of effective density for clipped OMAR CT pixels reduce dose calculation errors. When the phantom was configured with amalgam and Co-Cr-Mo alloy dental implants the effective density of these implants was found to be 4.5 g/cm3. When the phantom was configured with implants containing amalgam and gold, the effective density of amalgam in the presence of gold was 5.5 g/cm3 while gold had an effective density of 6.5 g/cm3. The median and maximum range of errors for the uncorrected images were ± 0.6 % and 7.4% respectively for the phantom configured with amalgam and Co-Cr-Mo (tray one) and ± 0.5 % and 18.1 % respectively for the phantom containing amalgam and gold (tray two). The median and maximum range of errors for the corrected images after applying overrides of effective densities were ± 0.5 % and 4.7% respectively for tray one and ± 0.3 % and 7.7 % respectively for tray two. In conclusion, introduction of density overrides of walls, wax and effective density of high-density materials can reduce the errors induced by metal artifacts and improve the accuracy of dose calculations in treatment planning systems to deliver the relevant dose to a target organ. / Thesis / Master of Science (MSc)

Iminium salt dimethyliminium hexafluorophosphate (5), a precursor of the target compound BPDO-TTF, was synthesized in five steps with significant modifications in the previously worked out scheme. 2,3-Dihydrobenzo-1, 4-dioxin(1) was synthesized from catechol in a strongly alkaline medium. Compound (1) was dichlorinated with SO2Cl2 to give 2,3-dichloro-2,3-dihydrobenzo-1,4-dioxin (2). Subsequent reaction of compound (2) with sodium dimethyldithiocarbamate followed by dehydrochlorination in triethyl amine gave 1,4-Benzodioxin-2-N,N-dimethyldithiocarbamate (3). Bromination of compound (3) followed by treatment with hexafluorophosphoric acid yielded bromo-dimethyliminium hexafluorophosphate (4). Dehydrobromination of compound (4) by solid phase thermolysis reaction resulted in dimethyliminium hexafluorophosphate (5). In an effort to prepare BPDO-TTF from a coupling reaction of dithioselenone (7), the dimethyl iminium salt (5) was reacted with NaHSe; however, no pure dithioselenone was isolable from the reaction mixture. A photochemical coupling reaction was also attempted using dimethyl iminium salt (5) and trimethyl phosphite but the desired product did not form in a detectable amount.

Advancements in DNA microarray data sequencing have created the need for sophisticated machine learning algorithms and feature selection methods. Probabilistic graphical models, in particular, have been used to identify whether microarrays or genes cluster together in groups of individuals having a similar diagnosis. These clusters of genes are informative, but can be misleading when every gene is used in the calculation. First feature reduction techniques are explored, however the size and nature of the data prevents traditional techniques from working efficiently. Our method is to use the partial correlations between the features to create a precision matrix and predict which associations between genes are most important to predicting Leukemia diagnosis. This technique reduces the number of genes to a fraction of the original. In this approach, partial correlations are then extended into a spectral clustering approach. In particular, a variety of different Laplacian matrices are generated from the network of connections between features, and each implies a graphical network model of gene interconnectivity. Various edge and vertex weighted Laplacians are considered and compared against each other in a probabilistic graphical modeling approach. The resulting multivariate Gaussian distributed clusters are subsequently analyzed to determine which genes are activated in a patient with Leukemia. Finally, the results of this are compared against other feature engineering approaches to assess its accuracy on the Leukemia data set. The initial results show the partial correlation approach of feature selection predicts the diagnosis of a Leukemia patient with almost the same accuracy as using a machine learning algorithm on the full set of genes. More calculations of the precision matrix are needed to ensure the set of most important genes is correct. Additionally more machine learning algorithms will be implemented using the full and reduced data sets to further validate the current prediction accuracy of the partial correlation method.

Manifold learning algorithms have been shown to be useful for many applications of numerical analysis. Unfortunately, existing algorithms often produce noisy results, do not scale well, and are unable to benefit from prior knowledge about the expected results. We propose a new algorithm that iteratively discovers manifolds by preserving the local structure among neighboring data points while scaling down the values in unwanted dimensions. This algorithm produces less noisy results than existing algorithms, and it scales better when the number of data points is much larger than the number of dimensions. Additionally, this algorithm is able to benefit from existing knowledge by operating in a semi-supervised manner.

Resonance frequency detuning (RFD) is a piezoelectric-based vibration reduction approach that applies to systems experiencing transient excitation through the system*s resonance—for example, turbomachinery experiencing changes in rotation speed, such as on spool-up and spool-down. This technique relies on the inclusion of piezoelectric material and manipulation of its electrical boundary conditions, which control the stiffness of the piezoelectric material. Resonance frequency detuning exploits this effect by intelligently switching between the open-circuit (high stiffness) and short-circuit (low stiffness) conditions as the excitation approaches resonance, subsequently shifting the natural frequency to avoid this resonance crossing and limit the response. The peak response dynamics are then determined by the system*s sweep rate, modal damping ratio, electromechanical coupling coefficient, and, most importantly, the trigger (represented here in terms of excitation frequency) that initiates the stiffness state switch. This thesis identifies the optimal frequency-based switch trigger over a range of sweep rates, damping ratios, and electromechanical coupling coefficients. With perfect knowledge of the system, the optimal frequency-based switch trigger decreases approximately linearly with the square of the coupling coefficient. Furthermore, phase of vibration at the time of the switch has a very small effect; switching on peak strain energy is marginally optimal. In practice, perfect knowledge is unrealistic and an alternate switch trigger based on an easily measurable parameter is necessary. As such, this thesis also investigates potential methods using the open-circuit piezoelectric voltage response envelope and its derivatives. The optimal switch triggers collapse to a near linear trend when measured against the response envelope derivatives and, subsequently, an empirical control law is extracted. This control law agrees well with and produces a comparable response to that of the optimal control determined using perfect and complete knowledge of the system.

Pollution from hydrofluorocarbons (HFC) poses a serious challenge to the environmental community. Released from industrial operations, they have contaminated both the atmosphere and groundwater and are considered persistent in both media.1 For over the past 20 years, the practice of synthesizing hydrofluorocarbons as alternatives to chlorofluorocarbons (CFC) has been conducted in an effort to reverse the effects of stratospheric ozone layer depletion. 2,3 However, in doing so these new fluorinated compounds exhibited an unexpected property as a potent global warming greenhouse gas (GHG) with radiative forcing potentials in the range of 100 to 10,000 equivalents greater than carbon dioxide.4 Conversely, HFCs exhibit desirable properties as precision cleaning solvents due to their low surface energy but that use has lead to releases contaminating groundwater resulting in recalcitrant pollution in the form of dense nonaqueous phase liquids (DNAPL).5 The Environmental Protection Agency (EPA) has recently requested studies on the environmental impact of HFCs with respect to a number of petitions received from various environmental action groups imploring the use of the Montreal Protocol as the vehicle by which to achieve elimination of the compounds from industrial operations.6,7 Additionally, results from studies requested by the international community have shown HFCs to exhibit developmental and neurological damage in animal life along with their impact to humans remaining not completely understood.8,9,10 Therefore, the potential hazards of HFCs to human health and the environment necessitates the development of an effective and environmentally responsible technology for their remediation from groundwater. The National Aeronautics and Space Administration (NASA) has employed the use of various halogenated solvents in its spacecraft cleaning operations at its facilities for many years iv and in that time experienced accidental releases which eventually resulted in environmental contamination.11,12,13 Many of the organic solvents employed in these operations consisted of halogenated compounds with most being partially chlorinated and fluorinated hydrocarbons. Through normal use and operation, releases of these materials found their way into the environs of atmosphere, soil and groundwater. Remediation of fluorinated compounds has not followed the successful path laid by clean-up technologies developed for their chlorinated counterparts.14,15,16,17 Fluorinated compounds are resistant however to those methods due to their unreactive nature stemming from the properties of the strong carbon-fluorine bond. 18 This unique bonding property also ensures that their environmental persistence endures. 19 One particular fluorinated groundwater contaminant, the HFC 1,1,1,2,2,3,4,5,5,5-decafluoropentane (DFP), which serves as an excellent cleaning agent and has been used by NASA since the late 1990’s and still remains in use today, was selected as the focus of this study. 20 For this study, various reductive metal systems were evaluated for their capability towards effective degradation of DFP. These included the metals: iron, magnesium, aluminum and zinc and several bimetallic alloys as well as attempts employing some on carbon support. Variations in protic solvent reaction media and acidic metal activation were also explored. The bimetallic reductive catalytic alloy, magnesium with palladium on carbon support Mg(Pd/C) in aqueous media, proved to be the successful candidate with 100% conversion to simple hydrocarbons. Mechanistic evaluation for degradation is proposed via a series of stepwise catalytic reduction reactions. Kinetic studies revealed degradation to obey second order reaction kinetics. Further study should be conducted optimizing an in situ groundwater delivery method for field application. Additionally, the developed technology should be assessed against other v groundwater fluorocarbon pollutants; either as a method for remediating multiple fluorinated polluted sites or as a polishing agent where all other non-fluorinated halogen pollutants have been abated.

Noise can be defined as unwanted sound. There are many cases and applications that reducing noise level is of great importance. Loss of hearing is only one of the effects of continuous exposure to excessive noise levels. Noise can interfere with sleep and speech, and cause discomfort and other non-auditory effects. High level noise and vibration lead to structural failures as well as reduction in life span. The importance of noise issue could be well understood by looking at regulations that have been passed by governments to restrict noise production in society. Industrial machinery, air/surface transportation and construction activities are main contributors in noise production or "noise pollution". Noise Pollution is not only an annoyance; it is an environmental health hazard. Noise can be found anywhere that life exists, in forests, in the workplace, in homes acrossAmericaand even under water. A lot of research has been done about noise pollution in the last 40 years, but yet there is still more to learn about how to control and lessen the affects that noise has on human and animal life. Noise control is a major factor in the planning, design, and construction of transportation corridors. Architects, acoustical engineers and transportation planners are searching for creative ways to eliminate or greatly reduce noise levels. The challenge lies in attaining desired sound levels while simultaneously maintaining or enhancing the visual environment. I will be setting up an experiment to determine what kinds of materials absorb sound waves of varying frequencies most effectively.

Degradation of surface water quality by excess nutrients in stormwater is a substantial environmental and economic problem in the U.S. Phosphorus (P) is often the limiting nutrient for harmful algal blooms and the best target to prevent degradation. Natural treatment strategies such as constructed wetlands (CW) demonstrate effective and economical P management but obstacles exist to implementation. Biological P removal has large land requirements that limit the use of best management practices (BMP) in high land-value areas. Various BMP also utilize sorption processes (SP) for P removal but variations in performance and finite sorption capacity limit SP as a viable long-term removal strategy. However, by understanding variability and making sorption capacity renewable, SP could provide, with shorter retention times, a space-efficient, long-term removal strategy. This multi-study research program developed the urban wetland filter (UWF), a concept intended to overcome the unique limitations of high land-value areas to natural treatment strategies and provide a low-cost, easily implemented BMP to meet P management goals while harvesting sequestered P for use as a fertilizer. Experimental factors included substrate and influent properties pertinent to understanding performance variation and optimizing microbial iron (Fe) reduction for rejuvenation of sorption capacity. Regarding performance, modeling identified major sources of variability including, by order of importance, magnitude of a solution/substrate concentration gradient, length of the "antecedent dry period" between loadings, and pH. Field-scale results confirmed this multifactor dependence of P-removal while also supporting the inclusion of cast-iron filings in substrate to improve P removal. Regarding rejuvenation, results indicated that microbial Fe reduction is capable of releasing previously sequestered P from substrates. A sufficient carbon source was necessary, but microbial inoculation was not necessary to facilitate Fe reduction, which released most of the previously sequestered P, albeit more slowly than P sequestration. Field-scale results indicated that Fe reduction might occur faster under field conditions, possibly due to humic acids, and that inclusion of cast-iron filings enabled additional P removal after rejuvenation by providing a conservative source of Fe for the creation of new sorption sites; however, cast-iron filings may also limit the release of P during rejuvenation. / Ph. D.

Specific conductivity (SC) and total dissolved solids (TDS) are increasingly being used as a parameter used to judge the aquatic health of streams that are impacted by coal mining operations in the Appalachian region of the United States. Due to this, government environmental regulatory bodies have been considering issuing a regulation on SC for all mining operation outfalls. Sulfate typically has the greatest dissolved ion presence in coal mine waters. In literature examining the treatment of mine waters, SC and TDS analysis is typically not reported. The technologies examined in this study were nanofiltration membrane technology and biological sulfate reducing bioreactors. In the nanofiltration study, three different nanofiltration membranes were evaluated for SC reduction: NF270, DK, and NFX. The DK and NFX nanofilters were able to reduce SC levels by an average of 84 percent for both mine waters tested and were able to reach SC levels below the proposed limit of 500 S/cm. The SC levels achieved by the NF270 nanofilters were observed to have much higher variability. The inclusion of microfiltration and simulated-sand filtration were also introduced as a pre-treatment stage in order to determine whether or not nanofiltration performance would improve in terms of SC reduction. In the biological sulfate reducing bioreactor study, multiple bioreactors were established to identify the optimal organic mixture to foster both SC and sulfate reduction. Sulfate reduction began to occur approximately 20 days after the establishment of each bioreactor. SC levels were greater than 13,000 S/cm in each of the bioreactors sampled by the fortieth day of sampling. The probable cause of the increase SC was identified to be the manure/compost used in the study. Future testing should incorporate more sampling in the early phases of experimentation in order to ensure the ability to monitor changes in water quality. / MS