93887 research outputs found
Sort by
Deep Learning Approaches for Automatic Acoustic Detection of the Bachman\u27s Sparrow and its Application to Assessing its Response to Prescribed Burns in Subtropical Habitats of Central Florida
No full textAutomatic birdsong detection is a powerful tool that can enable cost-efficient monitoring of birds at large time and spatial scales. For the development of birdsong classifiers, the use of deep learning algorithms such as convolutional neural networks (CNNs) have shown considerable success. However, to successfully train a CNN-based birdsong classifier, a large dataset with labeled recordings of the target species is needed, a condition that represents a major limitation for researchers and conservationists. In this study, the influence of some machine learning techniques used to overcome data scarcity such as transfer learning, data augmentation, and Siamese Neural Networks were evaluated on the performance of Bachman\u27s Sparrow (Peucaea aestivalis)detection. For this purpose, CNN-based models were trained under five different treatments in nine different training set sizes. In addition, to estimate the generalization ability of each model, the birdsong detection performance was evaluated with different signal-to-noise ratio for each of the treatments. My results reveal that the use of pre-trained networks allows obtaining good results with false positive rates close to zero even with low training set sizes. In addition, the indiscriminate use of data augmentation can decrease the detection capability of the model, particularly for recordings with more background noise. My results indicated that increasing the training set does not necessarily improve model performance, so it is recommended to incorporate techniques that measure the potential of a dataset beyond its size
Classroom Translanguaging Practices and Secondary Multilingual Learners in Indiana
No full textMany multilingual learners who use a language other than English at home face academic challenges from English monolingualism prevalent in the U.S. school system. English as a New Language (ENL) programs teach English to these learners while playing a role in reinforcing English monolingualism. For educational inclusivity and equity for multilingual learners, it is imperative to center their holistic language repertoires in ENL classrooms; however, this can be challenging due to individual and contextual factors. Using translanguaging as a conceptual framework, this qualitative case study explores how high school multilingual learners’ languages are flexibly used in ENL classes and how the students think about such classroom translanguaging practices. I used ethnographic methods to observe ENL classroom activities and instructional practices, interview the participants, and collect photos and documents in a high school in Indiana for a semester. The participants were an English-Spanish proficient ENL teacher and four students from Puerto Rico, Mexico, Honduras, and the Democratic Republic of the Congo, whose language repertoires included Spanish, Lingala, French, Arabic, and English. The findings describe the difficulties and possibilities of incorporating all students’ multilingual-multisemiotic repertoires in ENL classes. The classroom language practices primarily constituted of Spanish and drawing; some instructional activities and practices, such as the multigenre identity project and the teacher’s use of Google Translate, well integrated the students’ multilingual-multisemiotic repertoires. When the students engaged in English writing, they frequently used machine translation, such as Google Translate, through dynamic processes involving evaluation. While the students perceived such classroom translanguaging practices generally positively, they considered using machine translation as a problem, a resource, or an opportunity. With these findings, I argue that multilingual learners’ competence to use their own languages and machine translation technology freely and flexibly is a valuable resource for learning and should be encouraged and developed in ENL classrooms. To do so, ENL teachers should use instructional activities and practices considering students’ dynamic multilingualism. TESOL teacher education should develop such competence in teachers, and more multilingual resources should be provided to teachers. In the case of a multilingual classroom with singleton students, building mutual understanding, empathy, and equity-mindedness among class members should be prioritized. Finally, I recommend that the evolving multilingual technologies, such as machine translation, be actively used as teaching and learning resources for multilingual learners
Forest Response Following the Loss of Ash
No full textThis study is part of an ongoing project established in 2007 as an emerald ash borer (Agrilus planipennis) trapping study. The primary objective of this thesis was to assess how forests have changed following the loss of ash over a 15-year period, from 2007 to 2022. The study aimed to answer three key questions: 1) How has the loss of ash affected the regeneration and recruitment patterns of ash and other tree species? 2) Which specific species are driving forest regeneration and recruitment in the absence of ash? 3) Can the data be used to predict future changes in forest composition following the loss of ash? To achieve these objectives, I conducted analyses of 44 forests representing 19 different forest types in Indiana, Michigan, and Ohio. Forest composition was examined at the overstory, midstory and understory levels to capture the overall impact of ash loss on forest regeneration and succession. Despite the decline or total loss in overstory ash, ash regeneration continued at most sites from 2017 to 2022. However, ash regeneration was not observed at 15 sites, and competition from established shade-tolerant species, such as red maple and American elm, was identified as a primary factor hindering ash recruitment. The loss of ash did not lead to significant changes in forest types. Forest types at all sites remained stable, with only minor shifts in dominant species. Tree species that were already present in the overstory filled the space left by the lost ash, thereby maintaining the existing forest types. In terms of driving species, my findings demonstrated a rapid increase in red maple dominance across all sites, particularly in the absence of ash. Red maple exhibited vigorous growth, suggesting it plays a crucial role in shaping future forest composition following ash loss. Other species, including black cherry, American elm, and American basswood, also contributed to forest regeneration, with some sites showing a shift toward a higher abundance of these species. Although the loss of ash has led to changes in forest composition, existing species have shown resilience and adaptability in filling the niche left by the lost genus. The long-term effects of emerald ash borer infestation on forest ecosystems are complex and multifaceted. This study provides insights into the regeneration and recruitment patterns of ash and other tree species following ash loss. The data suggest that while forests have experienced loss of ash, the remaining understory and midstory individuals of this species may contribute to future overstory recruitment. Furthermore, the dominance of red maple highlights its potential as a key driver of forest composition in the wake of ash loss
Advancement of Additive Manufacturing for Monopropellant Catalyst Beds
No full textMonolithic catalyst beds have been used extensively in other industries and are gaining interest for space propulsion applications. Additive manufacturing of monolithic supports allows for catalyst beds with a wider range of geometries than could be produced using conventional methods, potentially allowing for higher performance monoliths that can compete with conventional packed beds in performance. Achieving these gains requires a consistent, even, and well-adhering washcoating procedure for the additively manufactured supports, one which works well on varied geometry and on support materials that can be readily printed. I conducted an extensive development process on improving methods of surface preparation and coating for high temperature ceramic monoliths that resulted in improvements in the state of the art. The materials and methods used are appropriate for rocket grade hydrogen peroxide, hydrazine, or other monopropellants with similar operating temperatures. Using existing published coating methods resulted in uneven coating distribution and poor adhesion. I demonstrate that this was due to the substrate surface morphology producing a hydrophobic effect. Surface morphology plays a significant role in coating coverage and adhesion and differences in initial support surfaces likely account for much of the variation in results seen across the literature. I present a method of controlled thermochemical surface etching using pure sodium hydroxide at 420°C that can reliably produce a roughened hydrophilic surface from a variety of starting morphologies. I also present several modifications to the primer formulation that improve evenness of coverage, the most significant of which is the inclusion of a surfactant at a concentration of 1 g per 36 g water. The surface treatment and coating formulation improvements combine well and produce an even coating with strong adhesion to the substrate. I also conducted preliminary work on the investigation of novel geometric designs for monolithic catalyst beds, and on the reactivity of different transition metal oxide catalysts for rocket grade hydrogen peroxide decomposition
Exploiting Magnetic Correlations in Low-dimensional Hybrid Quantum Systems: Towards Next-Generation Spintronic Devices
No full textIn recent years, correlated magnetic phenomena have emerged as a unique resource for enabling alternative computing, memory, and sensing applications. This has led to the exploration of novel magnetic hybrid platforms with the promise of improved figures of merit over the state-of-the-art. In this dissertation, we delve into several example platforms where magnets interact with various other degrees of freedom, resulting in enhanced figures of merit and/or the emergence of novel functionalities.First, we investigate the possibility of utilizing the collective resonant mode of nanomagnets to enhance the electric field sensitivity of quantum spin defects. While quantum systems have garnered significant attention in recent years for their extraordinary potential in information processing, their potential in the field of quantum sensing remains yet to be fully explored. Quantum systems, with their inherent fragility to external signals, can be harnessed as powerful tools to develop highly efficient sensors. In this dissertation, we explore the potential of a specific type of quantum sensor, namely the quantum spin defects as an electric field sensor, when integrated with a nanomagnet/piezoelectric composite multiferroic. This integration yields at least an order of magnitude enhancement in sensitivity, presenting a promising avenue for quantum sensing applications.Next, we shift our focus towards harnessing magnetic correlation in the emerging class of atomically thin magnets known as van der Waals magnets. These magnets provide distinctive opportunities for controlling and exploiting magnetic correlations. Specifically, these platforms allow for tunable magnetic interactions by twisting two vertically adjacent layers of the magnet, features that are unique to van der Waals materials. By capitalizing on such twist degrees of freedom, we demonstrate the creation of twist-tunable nanoscale magnetic ground states. This capability opens up avenues for applications such as high-density memories and magnon crystals.Interestingly, the same material platform also allows for exploiting magnetic correlation by controlling the local electrical environment. We uncover the symmetry-allowed spincharge coupling mechanisms in the heterostructures of such magnets, a prediction that has received experimental support. Utilizing such understanding, we propose a setup for the electrical generation of magnons. Magnons—the elementary excitation of spin waves—have garnered a lot of attention these days due to their potential to couple various diverse physical systems and in the field of low dissipation computing. Our findings offer a potential pathway towards the realization of magnon-based spintronic devices
Utilizing Embedded Sensing for the Development of Piezoresistive Elastodynamics
No full textObtaining full-field dynamicmaterial state awareness would have profound and wideranging implications across many fields and disciplines. For example, achieving dynamic state awareness in soft tissues could lead to the early detection of pathophysiological conditions. Applications in geology and seismology could enhance the accuracy of locating mineral and hydrocarbon resources for extraction or unstable subsurface formations. Ensuring safe interaction at the human-machine interfaces in soft robotic applications is another example. And as a final representative example, knowing real-time material dynamics in safety-critical structures and infrastructure can mitigate catastrophic failures. Because many materials (e.g., carbon fiber-reinforced polymers composites, ceramic matrix composites, biological tissues, cementitious and geological materials, and nanocomposites) exhibit coupling between their mechanical state and electrical transport characteristics, self-sensing via the piezoresistive effect is a potential gateway to these capabilities. While piezoresistivity has been mostly explored in static and quasi-static conditions, using piezoresistivity to achieve dynamic material state awareness is comparatively unstudied. Herein lies the significant gap in the state of the art: the piezoresistive effect has yet to be studied for in-situ dynamic sensing.In this thesis, the gap in the state of the art is addressed by studying the piezoresistive effect of carbon nanocomposites subject to high-rate and transient elastic loading. Nanocomposites were chosen merely as a representative self-sensing material in this study because of their ease of manufacturability and our good understanding of their electro-mechanical coupling. Slender rods were manufactured using epoxy, modified with a small weight fraction of nanofillers such as carbon black (CB), carbon nanofibers (CNFs), and multi-walled carbon nanotubes (MWCNTs), and subject to loading states such as steady-state vibration at structural frequencies (102 − 104 Hz), controlled wave packet excitation, and high-strain rate impact loading in a split-Hopkinson pressure bar. This work discovers foundational principles for dynamic material state awareness through piezoresistivity.Three major scholarly contributions are made in this dissertation. First, an investigation was pursued to establish dynamic, high-strain rate sensing. This investigation clearly demon strated the ability of piezoresistivity to accurately track rapid and spatially-varying deformation for strain rates up to 102 s −1. Second, piezoresistivity was used to detect steady-state vibrations common at structural frequencies. Utilizing simple signal processing techniques, it was possible to extract the excitation frequency embedded into the collected electrical measurements. The third contribution examined the dynamic piezoresistive effect through an array of surface-mounted electrodes on CNF/epoxy rods subject to highly-controlled wave packet excitation. Electrode-spacing adjustments were found to induce artificial signal filtering by containing larger portions of the injected wave packets. The strain state in the rod was found after employing an inverse conductivity-to-mechanics model, thereby demonstrating the possibility of deducing actual in-situ strains via this technique. A digital twin in ABAQUS was constructed, and an elastodynamic simulation was conducted using identical dynamic loading, the results of which showed very good agreement with the piezo-inverted strains.This work creates the first intellectual pathway to full-field dynamic embedded sensing. This work has far-reaching potential applications in many fields, as numerous materials exhibit self-sensing characteristics through deformation-dependent changes to electrical properties. Therefore, piezoresistive elastodynamicshas the incredible potential to be applied not just in structural applications but in other potentially innovated applications where measuring dynamic behavior through self-sensing materials is possible
Intelligent Solid Waste Classification System Using Deep Learning
No full textThe proper classification and disposal of waste are crucial in reducing environmental impacts and promoting sustainability. Several solid waste classification systems have been developed over the years, ranging from manual sorting to mechanical and automated sorting. Manual sorting is the oldest and most commonly used method, but it is time-consuming and laborintensive. Mechanical sorting is a more efficient and cost-effective method, but it is not always accurate, and it requires constant maintenance. Automated sorting systems use different types of sensors and algorithms to classify waste, making them more accurate and efficient than manual and mechanical sorting systems. In this thesis, we propose the development of an intelligent solid waste detection, classification and tracking system using artificial deep learning techniques. To address the limited samples in the TrashNetV2 dataset and enhance model performance, a data augmentation process was implemented. This process aimed to prevent overfitting and mitigate data scarcity issues while improving the model\u27s robustness. Various augmentation techniques were employed, including random rotation within a range of -20° to 20° to account for different orientations of the recycled materials. A random blur effect of up to 1.5 pixels was used to simulate slight variations in image quality that can arise during image acquisition. Horizontal and vertical flipping of images were applied randomly to accommodate potential variations in the appearance of recycled materials based on their orientation within the image. Additionally, the images were randomly scaled to 416 by 416 pixels, maintaining a consistent image size while increasing the dataset\u27s overall size. Further variability was introduced through random cropping, with a minimum zoom level of 0% and a maximum zoom level of 25%. Lastly, hue variations within a range of -20° to 20° were randomly introduced to replicate lighting condition variations that may occur during image acquisition. These augmentation techniques collectively aimed to improve the dataset\u27s diversity and the model\u27s performance. In this study, YOLOv8, EfficientNet-B0 and VGG16 architectures were evaluated, and stochastic gradient descent (SGD) and Adam were used as the optimizer. Although, SGD provided better test accuracies compared to Adam.Among the three models, YOLOv8 showed the best performance, with the highest average precision mAP of 96.5%. YOLOv8 emerges as the top performer, with ROC values varying from 92.70% (Metal) to 98.40% (Cardboard). Therefore, the YOLOv8 model outperforms both VGG16 and EfficientNet in terms of ROC values and mAP. The findings demonstrate that our novel classifier tracker system made of YOLOv8, and supervision algorithms surpass conventional deep learning methods in terms of precision, resilience, and generalization ability. Our contribution to waste management is in the development and implementation of an intelligent solid waste detection, classification, and tracking system using computer vision and deep learning techniques. By utilizing computer vision and deep learning algorithms, our system can accurately detect, classify, and localize various types of solid waste on a moving conveyor, including cardboard, glass, metal, paper, and plastic. This can significantly improve the efficiency and accuracy of waste sorting processes
Spectral Rigidity and Flexibility of Hyperbolic Manifolds
No full textIn the first part of this thesis we show that, for a given non-arithmetic closed hyperbolic n manifold M, there exist for each positive integer j, a set M1, ..., Mj of pairwise nonisometric, strongly isospectral, finite covers of M, and such that for each i, i0 one has isomorphisms of cohomology groups H∗ (Mi , Z) = H∗ (Mi ,Z) which are compatible with respect to the natural maps induced by the cover. In the second part, we prove that hyperbolic 2- and 3-manifolds which arise from principal congruence subgroups of a maixmal order in a quaternion algebra having type number 1 are absolutely spectrally rigid. One consequence of this is a partial answer to an outstanding question of Alan Reid, concerning the spectral rigidity of Hurwitz surfaces
Improving Pig Performance and Efficiency by Attenuating Transport and Immune Stress Responses Through L-Tryptophan Supplementation
No full textIn commercial settings piglet weaning and transportation occur concurrently due to the expansive application of multi-site production systems across the United States and the combination of these events can be defined as an early life stressor. Early life stress is known to reduce pig performance, efficiency, and immune resilience contributing to reduced welfare and increased production losses. To combat the deleterious effects of stress on pigs, the essential amino acid tryptophan (Trp) supplemented above current National Research Council (NRC 2012) recommended levels, improves neuroendocrine responses to stress, lowers plasma cortisol and norepinephrine concentrations, and improves hypothalamic-pituitary-adrenal axis recovery time following stress. However, there are discrepancies concerning the Trp requirement for nursery pigs, suggesting the 2012 recommendations for Trp may be inaccurate for optimizing growth performance and health in modern pigs. Three experiments were conducted to determine the effects of supplemented standard ileal digestible Trp above NRC (2012) recommended levels on performance, feed efficiency, immune vigor, and stress tolerance. The objective of experiment 1 was to eliminate or reduce short- and long-term, transport-induced reductions in piglet feed efficiency and growth by supplementing Trp above NRC (2012) recommendations pre-weaning and/or during the nursery phase. An oral gavage of Trp or a control milk carrier was provided to pre-allotted piglets beginning at day 5 of lactation and continuing to wean. At weaning all pigs were blocked by sex, weaning weight and pre-wean treatment and randomly assigned to transport and post-wean treatments. Pigs were fed in four nursery phases with diets containing 1X or 2X NRC recommended concentrations of SID Trp and a common grower diet fed in 6 phases during the grow-finish period. At market, loin characteristics were measured via ultrasound and carcass data was collected from the packing plant. Pre-weaning Trp supplementation had no effect on preweaning growth performance; however, post-wean Trp increased overall body weight and average daily gain in nursery pigs when Trp was supplemented pre-wean. The objective for experiment 2 was to mitigate the adverse effects of early life transport stress on subsequent immune challenges by providing supplemental Trp during the nursery period. At weaning, pigs were transported for 8 hours and assigned to treatments of vaccine-induced immune challenge and dietary treatment. Pigs were fed standard nursery diets, in four phases, over 35 days with pigs receiving 1X or 2X the NRC (2012) recommended Trp concentration. Half the pigs on each dietary treatment were subjected to a 3-wk vaccine challenge consisting of circovirus, mycoplasma, and influenza vaccines administered in wk 2, 3, and 4 post-wean, respectively. At market, loin characteristics were measured via ultrasound and carcass data was collected. At the culmination of the nursery period, unchallenged pigs supplemented with Trp were heavier compared to control pigs. This resulted from an overall improvement in average daily gain for 2X Trp fed pigs. However, increasing Trp did not affect market weight, loin eye area, or lean percent in market pigs, although 2X Trp increased back fat. The objective of experiment 3 was to determine what ratio of SID tryptophan, relative to lysine, maximizes growth performance and feed efficiency in weaned pigs during the nursery period
Characterizing Synthetic Antigen-Binding Fragments for Isolation of the Toc Complex
No full textProtein translocation across the chloroplast outer membrane is essential for photosynthesis in all plants and certain algae. This is because most chloroplast proteins (over 90%) are encoded in the nucleus, translated in the cytoplasm, and must be imported into the chloroplasts to perform their function. The translocon at the outer chloroplast membrane (TOC) complex orchestrates this vital translocation process and consists of three components in plants: Toc75, Toc33/34 and Toc159. Our overall goal is to elucidate the architecture of the TOC complex to gain mechanistic insights into protein translocation into chloroplasts. However, the major bottleneck preventing structure determination of the TOC complex has been the inability to produce or isolate the complex to sufficient yields and purity for structural studies. We began by using phage display to screen for synthetic antigen-binding fragments (sABs) that bind to the soluble POTRA domains of Toc75 from both Arabidopsis thaliana and Pisum sativum. We then characterized the POTRAsAB interactions using size-exclusion chromatography coupled with small angle X-ray scattering (SEC-SAXS), isothermal titration calorimetry (ITC), and X-ray crystallography. Finally, we show that we can use an affinity tagged sAB to isolate the TOC complex directly from pea biomass. This study has paved the way for high-resolution structural studies of the TOC complex from plants to understand protein translocation mechanisms