62588 research outputs found
Sort by
Confluences of Peace and Power: A Microhistory of Peacemaking in Trans-Appalachia, 1765
No full text{"value":" In 1765, representatives from the British Empire, led by Deputy Superintendent of Indian Affairs George Croghan, and diplomats representing diverse Indigenous nations that had spent the last two years resisting the expansion of that empire during the so-called Pontiac’s War met to discuss the terms of a lasting and mutually beneficial peace. This dissertation is a microhistory of those negotiations. As such, it is a study of peace, both as a choice and a practice, during a time much more commonly seen as exceedingly violent. The broader historiography of colonialism in North America squarely focuses on the violence of dispossession of Indigenous peoples during this time, which is understandable considering how widespread and deleterious this process was for Indigenous communities. However, such a focus simultaneously creates the sense that violence and dispossession were teleological processes within the structure of colonialism. This dissertation instead uses a focused diplomatic and ethnohistorical lens to argue that peace was possible and supported by Indigenous and British imperial agents alike, and therefore a later return to dispossessive violence by the nascent United States was a choice and not a structural inevitability. Such an approach uncovers diverse voices, including voices erased by the archive while centering the agency and impact of often historically marginalized peoples. In the process, this dissertation intervenes within this historiography of Pontiac’s War itself broadly Indigenous-Anglo American relations, as well as a still developing historiography of eighteenth-century peace studies.","attr0":"abstract"
Compression Aided Differentially Private Learning for Large Machine Learning models
No full textFederated learning (FL) has revolutionized collaborative model training by enabling decentralized systems to learn from distributed data while preserving local privacy at edge devices. However, the real-world deployment of FL faces two major challenges: ensuring communication efficiency and maintaining robust privacy guarantees, particularly in non-IID settings where client data distributions vary significantly. This research addresses these challenges by integrating Gradient Sparsification with Quantization (GSQ) and Differential Privacy (DP) into the Federated Averaging (FedAvg) algorithm. Through systematic experimentation, this study evaluates the impact of these enhancements on model performance, scalability, and privacy-utility trade-offs.The experiments, conducted on datasets of varying complexity, culminate with CIFAR-100, where different privacy settings—high, medium, and low—reveal the intricate balance between privacy and utility. The results demonstrate that GSQ effectively reduces communication overhead, allowing resource-constrained devices such as smartphones and IoT sensors to actively participate in federated learning without overwhelming their bandwidth or computational resources. At the same time, DP safeguards individual data by adding controlled noise to gradient updates, ensuring robust privacy protections. While stronger privacy guarantees slightly reduce model accuracy due to added noise, careful calibration of compression and privacy parameters enables a balance where both data security and model performance are maintained
Acoustic and flow measurements of porous plate designs for aerodynamic noise mitigation
No full textAn experimental study investigates parametrically the effects of porosity on the acoustic and aerodynamic fields about lifting- and non-lifting surfaces at two separate aeroacoustic facilities using microphone arrays and hot-wire anemometry. A single dimensionless porosity parameter characterizes the flow noise generated by a turbulent boundary layer and informs the design of the porous edge test specimens, including perforated flat plates and flat-plate extensions with a blunt or sharp trailing edge.The strong tonal peak due to vortex shedding from blunt trailing-edges diminishes in magnitude as the porosity parameter increases, and high-porosity plates eliminate this tone from the acoustic spectra. Single-microphone measurements indicate further that the porous plates examined can reduce low-frequency noise and increase high-frequency excess noise levels by up to 10 dB. DAMAS beamforming of the porous plates with sharpened edges reveal similar results on the acoustic spectra and identify that the principal effect of edge porosity on the acoustic source regions is a reduction in low-frequency noise and an increase in high-frequency noise across the entire plate. Noise generated by porous edges in the low-frequency range by the trailing- and leading-edge regions can be reduced by up to 20 dB, and porous edges increase high-frequency noise by up to 20 dB. Plates with the same dimensionless porosity perform similarly, where plates with circular holes perform slightly better (2 dB) than their counterparts with square holes at reducing low-frequency noise the most and increasing high-frequency noise the least in wind tunnel testingHot-wire anemometry of the flow field about blunt porous trailing edges reveals a down- ward shift of the bluntness-induced vortex-shedding peak in the spectra of turbulent velocity fluctuations, which are not seen in the acoustic spectra. In addition, flow field measurements for both the blunt-edged and sharp-edge plates indicate significant increases in turbulence intensity at the plate surface which are believed to be caused by the presence of holes and related to the increase in noise seen at high frequencies.The wing of a remote-controlled glider is modified with porous plates near the trailing edge to demonstrate reductions in surface pressure level fluctuations on a flying vehicle at the owl scale. Measurements of these fluctuations on the wing and fuselage indicate the capacity of porous plates to modestly reduce surface pressure levels in select frequency ranges and settings of aerial vehicles.</p
A Study of Process-Structure-Property Relationships for Novel Micro-Metal Additive Manufacturing Technologies
No full textOver the course of the last four decades, additive manufacturing has fundamentally altered the perception of how parts can be designed and assembled. More specifically, metal additive manufacturing (MAM) provides the potential to fabricate parts with limitless geometric complexity and novel material properties. As the technology matures, MAM is becoming increasingly utilized in industries such as aerospace, automotive, health care, and energy. One of the main factors halting its progression is the size limitation of parts, at both extremes. Most parts can only be built as large as a printing unit can encompass. Alternatively, most parts can only be built as small as current technology allows. Micro-metal additive manufacturing (MMAM) provides an alternative method of fabricating small parts with unmatched design freedom compared to traditional manufacturing processes. In this study, three MMAM processes were used to print components for microstructural characterization and mechanical testing, in comparison to their MAM counterpart. The initial research focused on a comparison between a powder bed fusion and sinter-based technology. Selective laser melting (SLM), metal binder jetting (MBJ), and digital light processing (DLP) were chosen to determine the respective capabilities of each technology. Scanning electron microscopy (SEM), namely energy dispersive x-ray spectrometry (XEDS) and electron backscatter diffraction (EBSD), was used to characterize their resultant microstructures. No qualitative difference in microstructure or chemistry was found between the MMAM and MAM for either powder bed fusion or sinter-based additive manufacturing components. Rather, a difference in their mechanical performance was directly related to their processing conditions and feedstock material. Components produced by powder bed fusion exhibited higher densities (≥ ~98%), tensile strength (~ 560-750 MPa), and surface roughness (~ 3-5 µm Ra). Sinter-based components exhibited much higher toughness (elongation ~ 50-75%) and lower surface roughness (~1-3 µm Ra).
The latter research focused on a characterization of a novel metal material jetting (MMJ) as compared to metal binder jetting (MBJ), considering both their as-sintered condition (no post-processing) and after post-processing with hot isostatic pressing (HIP). HIP was found to greatly improve densification for MBJ but had little to no effect for MMJ. The difference in microstructures was entirely process related. A difference in sintering profiles and levels of contamination resulted in two distinct secondary phases. δ-ferrite was formed during high-temperature sintering of MBJ, while (Mn, Cr)Cr2O4 and SiO2 oxides were formed due to oxygen contamination in the MMJ printing process. MBJ components exhibited a medium-strength (500-560 MPa), high-elongation (~ 75-98%) mechanical response, predominantly influenced by density and Hall-Petch strengthening. Alternatively, MMJ samples exhibited a high-strength (~ 630-670 MPa), low-elongation (~ 20%) mechanical response, more closely related to powder bed fusion. The mechanical strength was influenced by density, Hall-Petch strengthening, and dispersion strengthening. The findings of this research provided significant insight into MMAM process related microstructures and mechanical properties not yet reported in literature.
</p
Biomechanical and Biophysical Models and Coarse-grained Simulations for Virus-Cell Adhesion
No full textA critical event during the process of cell infection by a viral particle is attachment, which is driven by adhesive interactions and resisted by bending and tension. Understanding the contact mechanics of virus-cell system is the key to predicting the adhesive interaction and finding a way to blocking the adhesion. The biophysics of this process has been studied extensively, but the additional role of externally applied force or displacement, the role of glycocalyx and spike-like protrusion are generally lacking. To get a better understanding of the adhesive force-displacement response of viral particles against the cell membrane, two continuum models were built: one in which the viral particle is cylindrical, representative of a filamentous virus such as Ebola, and another in which it is spherical, representative of viruses such SARS-Cov-2 and Zika. The models are built based on the case where the deformations are small. We consider the two main features: bending stiffness and tension on the membrane when it\u27s deformed so that the total energy can be written as the sum of three contributions: bending energy, tension energy and adhesion energy. We are able to describe the force-deflection and contact-deflection relationships and seek the moment when there is no adhesion or pull-off force between the virus particle and cell membrane for bending dominated and tension dominated cases.
To extend the application of our continuum models by accounting for an additional layer of complexity, we extended our models by introducing the Winkler foundation as representative of a deformable membrane-virus interface. The previous models fall short in representing the deformable nature of adhesive receptors, such as the Spike protein and transmembrane immunoglobulin and mucin domain (TIM) family that mediate adhesion of SARS-CoV-2 and Ebola viruses, respectively. To address these limitations, we borrow the idea of representing adhesion by a cohesive zone, which is characterized by a cohesive stress in addition to the work of adhesion. By employing the new model, we are able to predict the pull-off force needed to break the adhesion between the virus and the cell in two cases. By comparing the force-separation curves simulated by the model and experimental data, we found that the cohesive zone model can effectively explain the AFM pull-off force trace when incorporating additional physical aspects of viral-cell adhesion systems, including receptor density and receptor-ligand elasticity.
A common feature of virus-cell adhesion mechanisms is a spike-like protrusion. In the third part of the dissertation, we explore the influence of spike biophysical properties on adhesion. The virus internalization process initiates when the spike-shaped protrusions bind to their receptor. This process is a crucial focus of the therapies such as vaccine and drug development. Understanding their structure and function is vital for developing effective treatments and vaccines against viral infections. We developed a coarse-grained model to study how the physical properties such as flexibility could affect the binding.</p
English Language Arts Pre-Service Teachers\u27 Perceptions of Curricula and Beliefs about Teaching Autonomy and Self-Efficacy
No full textSecondary (grades 7-12) English Language Arts (ELA) teacher preparation programs require students to spend time in a clinical setting, but the types of curricula that student teachers experience in these settings can differ. Sparse research exists on secondary ELA student teachers\u27 experiences with different types of curricula. This qualitative case study aimed to examine how student teachers viewed the feasibility and acceptability of the curricula they experienced during student teaching, and the effect, if any, the curricula had on their sense of autonomy and self-efficacy. The participants in this study included seven university students pursuing certification in secondary ELA education. Participants participated in two interviews, one before student teaching or toward the beginning of student teaching and one toward the end or after student teaching, and completed two reflections during their student teaching experience. Within-case findings are presented, as are seven between-case themes including participants\u27 overall positive beliefs of curricula, participants\u27 beliefs about the feasibility and acceptability of standardized curricula, the impact of curricula on autonomy, the prevalence of text-centered curricula and its impact on increased autonomy, the impact of curricula on self-efficacy, and the impact of implementing the mentor teacher\u27s materials on self-efficacy, and the impact of the student teacher\u27s relationship with the mentor teacher on self-efficacy. The findings of this study are useful to teacher education program leaders who wish to support student teachers in implementing various curricula types during student teaching.</p
Optimizing Fairness in Graph Neural Networks: Multi-Objective Trade-offs and Subgroup Discovery
No full textGraph Neural Networks (GNNs) have demonstrated exceptional effectiveness in spam detection by modeling the relational structures among reviewers, reviews, and products. However, structural biases, particularly those arising from skewed node degree distributions, lead to persistent fairness concerns across user groups. This thesis systematically addresses these challenges through three primary contributions. First, we propose VarReduction-PC, a stable and scalable multi-objective optimization (MOO) algorithm designed to generate high-quality Pareto fronts, thereby establishing a reliable foundation for fairness optimization. VarReduction-PC integrates variance reduction techniques within the predictor-corrector framework to ensure robustness under mini-batch training conditions commonly encountered in large-scale graph datasets. Second, we introduce a fairness certification framework for graph-based spam detection by formulating multiple fairness criteria as linear constraint systems and characterizing their trade-offs. The proposed stochastic search method enables efficient exploration of Pareto-optimal solutions without requiring explicit preference specifications. Third, we investigate hidden behavioral subgroup structures among users, propose a new subgroup-sensitive attribute, and design a joint training framework that simultaneously infers subgroup membership and enhances fairness calibration. Extensive experiments conducted on real-world review graph datasets validate the effectiveness of the proposed methods. Our approaches consistently improve fairness across multiple criteria while preserving detection performance and exhibit superior Pareto front coverage compared to baseline methods based on fairness regularization or adversarial debiasing. Overall, this thesis enhances the stability, scalability, and fairness of GNN-based spam detectors, contributing to the development of equitable graph learning.</p
Crowdsourcing for Indirect Bridge Modal Identification
No full textData-driven methods for bridge vibration monitoring enable frequent, accurate structural assessments; however, the high costs of large-scale deployments of these systems make important condition information a luxury for bridge owners. This work presents an inexpensive smartphone-based monitoring system capable of producing structural information in crowdsensing applications. Here, two methods are developed and tested: (1) absolute mode shape (AMS) identification using asynchronous mobile sensors, and (2) a complete modal identification using pairs of synchronous sensors. The AMS identification is the most extensive real-world study on bridge monitoring with crowdsourced smartphone-vehicle trips and simulated damage detection capabilities. The method analyzes over 500 trips across four bridges with main spans ranging from 30 to 1300 meters in length, representing about one-quarter of US bridges, and extracts absolute value mode shapes. This method is expanded to incorporate two time-synchronous mobile sensors to estimate a spatially dense frequency response matrix. Subsequently, this matrix can be integrated into existing system identification methods and structural health monitoring platforms, including the NExT-ERA and FDD. The methodology was tested numerically and using a lab-scale experiment for long-span bridges. This dissertation demonstrates a bridge health monitoring platform compatible with ride-sourcing data streams that check conditions daily. The result is the potential to commodify data-driven structural assessments on a global scale.</p
Terahertz Spectral Beam Combining and Multispectral Sensing Using Laser Arrays
No full text{"value":"The terahertz (THz) region of the electromagnetic spectrum (0.3-30 THz) falls between that of infrared radiation and microwaves. Different chemical and biomolecular species have unique frequency-dependent absorption characteristics at THz frequencies. To-date, THz sensing is primarily done with THz time-domain-spectroscopy (TDS) systems that are costly and are considerably limited in targeting applications in THz sensing and spectroscopy. Since the advent of THz semiconductor quantum-cascade lasers (QCLs) in 2001, there have been efforts to address such applications with QCLs. However, THz QCLs have unique challenges that need to be addressed before they could be viably utilized for THzspectroscopy. Toward meeting these challenges, here we demonstrate three advancements in the field: spectral beam combining of an array of such QCLs, spectral sensing of high- loss liquids with THz QCLs, and extending the frequency range of high-power THz QCLs to lower frequencies (∼ 2 THz) for broader spectral coverage.
For THz spectroscopy, a high-brightness multispectral source of coherent radiation is required. Toward that goal, frequency tunable THz QCLs and frequency-comb generation in THz QCLs are being explored in literature, albeit the optical power with these tech niques is sill low ( ∼ tens of milliwatts). In this dissertation, planar semiconductor-chip based blazed diffraction-gratings are developed to work in the terahertz frequency range. The grating is manufactured using standard semiconductor processing and optical lithography techniques. Such gratings will potentially find applications in terahertz spectroscopy, interferometry, ultrafast technology, and instrumentation for dispersive optics. The grating is then used to experimentally develop a spectral beam combining (SBC) technique for an array of high-power THz QCLs fabricated on a single semiconductor chip. The mountings and fixtures of SBC system are designed and machined accordingly. Single-lobed beams of four QCLs emitting around 3.2THz and spectrally separated by ∼ 14GHz are combined to achieve collinear propagation with ≲ 0.1◦ of pointing error.
With recent advents in development of high-power THz QCLs, here we propose and develop a unique THz sensing scheme of liquids to extend THz sensing beyond 3 THz by utilizing a custom-designed microfluidic Fabry-Perot Etalon cavity that allows precise control over the channel at micrometer scale, a spectrometer, and single-mode THz QCLs arrays. We demonstrate measuring THz absorption of liquid samples with a broad range of absorption (from 10/cm up to 800/cm), from high-loss and water-based solutions to low-loss oils. This is the first such demonstration of THz transmission measurements through relatively thick channels of liquids for a fast and accurate analysis of their THz optical properties. This technique could pave the way for future THz spectroscopy applications in non-invasive monitoring and analysis of contents and/or concentration of certain chemicals dissolved in solvents, etc.
The aforementioned demonstrations are done with THz QCLs emitting around 3 THz. Nowadays, THz QCLs can be designed to emit hundreds of milliwatts of peak average optical power at 3-4 THz because of relative high gain. In this dissertation, we have demonstrated, for the first-time, single-mode THz QCLs with distributed-feedback (DFB) operating at a much lower frequency of ∼2 THz with peak average power > 100 mW. The developed DFB QCLs have novel four-slit sixth-order Bragg gratings in top metal. Other promising designs based on finite-element simulations are presented as well. These developments expand the spectral range covered by terahertz semiconductor lasers.","attr0":"abstract"