116964 research outputs found
Sort by
Energy, gender, and development : a study of the linkages between access to reliable electricity, primary healthcare, and rural livelihoods, through a gender lens
Get PDFAccess to modern energy services has often been framed as a golden thread that enables development, but how these energy transitions affect the quality of public services, productive applications and gender equity has received little attention. In the context of unavailable, unreliable, or unaffordable sources of grid-based electricity for millions of vulnerable households, efforts to leverage electricity as a pathway for development outcomes have focused on innovation of decentralized renewable energy applications for livelihoods, healthcare, and education. Practitioners and researchers have highlighted the need for gender disaggregated research that can aid the design and implementation of gender inclusive programs in the energy sector. In this dissertation, I explore three questions at the nexus of energy and development using a gender lens. Specifically, I focus on how the delivery of primary healthcare is related to reliable electricity access, and how sustainable energy solutions can provide pathways for equitable, gender inclusive rural livelihoods. In the first paper, I develop a series of econometric models using India’s primary health center facility data and find that the lack of electricity access was associated with a significant and large decrease in the number of deliveries, number of in-patients, and number of out-patients treated at the facilities. Lower levels of electricity access were disproportionately associated with adverse effects on women’s access to safe and quality healthcare. In the second paper, through a series of interviews with rural entrepreneurs in India who recently bought Decentralized Renewable Energy Assisted Machines (DREAMs) to improve their livelihoods, I explore the ways in which the entrepreneurs’ gender influenced their access to the enabling environment for DREAMs, and their agency in the process of acquiring and using the DREAMs. Building on this fieldwork, in the third paper, I present a gender-disaggregated analysis of multi-level impacts that entrepreneurs experience post their adoption of DREAMs, and the mechanisms through which these impacts translate to building different forms of livelihood resilience for their households. Together, these papers provide insights for policy, program design, and program evaluation, and contribute to the emerging literature on the complex relationships between energy, gender, and development.Public Affair
Design of self-centering rocking steel braced frames : an engineering demand trade-off approach considering enhanced performance
Get PDFSelf-centering rocking braced frames have gained attention as an innovative seismic-resistant structural system due to its promising performance resulting in limited structural damage and recentering capabilities. While most rocking braced frame systems rely on rocking about their base only, recent research has shown that installing an additional rocking joint above the base can help to reduce unintended damage due to higher mode contributions. However, there is a growing research need to develop a better understanding of the impact of rocking design parameters on the behavior of these structures. Additionally, it is critical to investigate not only structural performance but also how design parameters affect the potential for nonstructural damage that can negatively impact post-event functionality and repair costs. This study aims to further inspect the effects of rocking design parameters on structural response and engineering demands, such as inter-story drift ratio and floor acceleration, together with the goal of producing designs that reduce the total damage in the building system, including structural and nonstructural damage. This dissertation presents the modeling, design, and analysis of three- and nine-story rocking frames for a wide range of design parameters. The study discusses the nonlinear seismic behavior of rocking systems and compares the influence of rocking parameters related to recentering, energy dissipation, and higher mode effect mitigation. In the study, a metric is proposed to evaluate the effectiveness of designs in reducing engineering demands such as inter-story drift ratio and floor acceleration. The proposed metric is used as the primary indicator to compare design alternatives with the intent of mitigating damage to nonstructural components considering a relative prioritization between acceleration-sensitive and drift-sensitive components. Based on this work, recommendations are made related to design of the post-tensioning, energy dissipation elements, number of rocking joints, and the location of upper rocking joints that are most effective in reducing the engineering demands pertinent to nonstructural components, while also considering cost-effectiveness in terms of total frame weight. From the results of this study, it is observed that self-centering systems with higher energy dissipation coefficients and an additional rocking joint work best to reduce engineering demands and reduced frame weight. Depending on the importance of reducing floor acceleration vs. inter-story drift, the benefit of providing more or fewer post-tensioning elements is discussed. The study also includes suggestions for future research to further improve the numerical models and design approaches for rocking braced frames beyond the current research scope.Civil, Architectural, and Environmental Engineerin
Methods for mitigating non-idealities in analog-domain compute accelerators
Get PDFArtificial Intelligence (AI) has given rise to a new era of computing paradigms that demand extraordinary amounts of parallel processing. Researchers have responded by giving increasing attention to Compute in Memory (CiM) and reduced precision computing methods to meet the new levels of parallelism required and overcome the challenges presented by the Memory Wall Bottleneck problem. Of these, methods based on promising emerging materials and non-volatile memories (NVM) are often plagued by non-ideal characteristics and stochastic behavior. These include device-to-device and cycle-to-cycle variation, resistance drift, reliability concerns, parasitic effects, and non-linear current-voltage (I-V) relationships. Hence improving design solutions to tolerate stochastic behavior is paramount to the success of the emerging computing paradigms described in this work.
First, this dissertation addresses the memory duplication incurred in CiM designs that accelerate 2D convolution by investigating data-movement techniques for improved energy efficiency. The proposed Systolic RAM demonstrates how the best practices of accelerator design and CMOS-compatible analog computing can be united to accelerate 2D direct convolution and minimize the cache-bandwidth requirement exacerbated by IM2COL. Systrolic RAM introduces two unique contributions: (1) an improved data movement pattern optimized specifically for 2D CNNs, and (2) signed-8-bit charge-domain computation demonstrated within the otherwise-unused BEOL SRAM metal layers. The non-ideal effects of analog computation on CNN accuracy are mitigated through a retraining approach.
In the next chapter, a neural-network framework for modeling non-ideal behavior of ReRAM is introduced and applied to devices fabricated using 2D hexagonal boron nitride (hBN). Device-to-device and cycle-to-cycle resistive switching variations are well captured by the model. This introduced ML approach is intended to expedite the creation of accurate device models and is relevant to research domains where physical understanding might otherwise take years to develop. Despite this, conventional ML approaches on their own cannot circumvent the need for physics-based understanding.
The final chapter introduces Differentiable Interpolation (DI) for representing physical circuit behavior by interpolating directly between measured samples. Unlike conventional interpolation, DI support auto-differentiation for compatibility with a leading deep learning library. DI is compared to the regression methods explored in Chapter 3 and demonstrates low loss with reduced compute overhead. This method is applied to model the undesired non-linearity of 1T1R bitcells found in crossbar arrays for vector-matrix multiplication (VMM). Then, an analog neural network is trained to mitigate the effect of non-linearity on the classification objective. The realized non-linear VMM provides a tractable method for simulating accurate circuit behavior, is computed without the need to run SPICE simulations at runtime, and improves the accuracy of a simulated analog neural network.Electrical and Computer Engineerin
Of ants and entomologists
Get PDFHere, I present a work concerning the behavior of ants and recruitment of new researchers into the field of entomology. This thesis presents work that intends to further understanding of insects both through novel research efforts, teaching, and recruitment. First, we test whether queen acceptance by workers of the fungus-farming ant Mycocepurus smithii can be predicted from the fungal genotypes of fungus gardens and from the ant genotypes of interacting workers and queens. We found that aggression toward queens is correlated with ant genotype, and that odor cues that ants may have acquired from their native fungal cultivar do not override the cues correlated with ant genotype during queen acceptance by workers. Second, we present novel predatory interactions between Neivamyrmex texanus and the fungus-farming ant Trachymyrmex septentrionalis (McCook, 1881) in central Texas, and between Nomamyrmex esenbeckii (Westwood, 1842) and numerous prey species, including the leafcutter ant Atta texana (Buckley, 1860), in southern Texas. Comparing predatory by Neivamyrmex on T. septentrionalis in central Texas to predation by Nomamyrmex on A. texana colonies in south Texas, we note a marked difference in response to army ant predation. Third, to increase knowledge of local insect diversity and to help recruitment into the field of entomology, the UT Austin Ecology, Evolution, and Behavior Undergraduate Student Organization conducted an insect survey on the urban campus of UT Austin. This insect survey was designed to understand the factors that impact insect habitat quality of urban green spaces. Following the insect survey, we used an end-semester questionnaire to understand the learning experiences by undergraduate participants. Answers to our questionnaire revealed important experiential outcomes from our diverse group of participating early-career researchers. Our project demonstrates the multifaceted usefulness of college campuses for conducting accessible, reproducible biodiversity surveys and promoting recruitment to the field of entomology. By successfully making new discoveries concerning the behavior of ants and involving early career scientists in entomological research, this thesis completes both its aims in terms of ants and entomologists.Ecology, Evolution and Behavio
Hexagonal boron nitride atomristor devices and applications
Get PDFRecently, nonvolatile resistive switching effects have been reported in atomic-scale metal-insulator-metal structures based on two-dimensional (2D) transitional metal dichalcogenide (TMD) and hexagonal boron nitride (h-BN) monolayers. These atomristor structures using 2D monolayers and gold (Au) electrodes showed outstanding nonvolatile resistive switching and retention characteristics even at the ultrathin monolayer thicknesses. However, numerous combinations of 2D monolayers and metal electrodes necessitate further exploration of atomristor research, so we must take a more strategic approach to address the serious readout margin and sneak path problems to advance toward practical neuromorphic computing applications. To mitigate these technical challenges, we made use of a polycrystalline h-BN monolayer and electrochemically active silver (Ag) metal electrodes with a thermally conductive diamond substrate to improve the memory window and reduce the leakage current in a single atomristor structure. The h-BN atomristor using symmetric Ag electrodes exhibited the largest memory window (~10⁹) and the lowest leakage current (< 0.2 pA) compared to other reported 2D atomristors. The h-BN atomristor using the same electrodes also demonstrated volatile and nonvolatile resistive switching within a single atomristor through tailored electrical and structural designs. For example, the 1.0 × 1.0 μm² h-BN atomristor transitioned from volatile to nonvolatile switching at a current compliance in the tens of nanoampere, whereas the 0.4 × 0.4 μm² h-BN atomristor showed more reliable multistate nonvolatile resistive switching characteristics from minimum measurable to maximum sustained current compliances. Furthermore, the 0.2 × 0.2 μm² h-BN millimeter-wave (mmWave) switch showed outstanding high-frequency and high-bandwidth signal transmission performance of insertion loss (~0.5 dB) and isolation (> 23 dB) across the D-band frequency range from 110 GHz to 170 GHz. Lastly, the h-BN atomristor achieved significant endurance(> 10⁴ cycles) and yield (~77%) through the wrinkle-free and residue-free polypropylene carbonate (PPC) assisted wet transfer method. The findings of this study demonstrated the potential of our atomristor research using a 2D h-BN insulating monolayer and electrochemically active Ag electrodes to achieve a highly-integrated multifunctional neuromorphic crossbar array structure.Electrical and Computer Engineerin
Doctoral thesis recital (double bass)
No full text4 unidentified works.MusicName of supervisor not provided
Using Augmented Reality to Train Undergraduate Engineering Students on Makerspace Tools
Get PDFTraining off erings for undergraduate engineering students vary widely depending on staffi ng, culture, and associated
coursework. Augmented reality (AR) has shown promise in training users for procedural tasks in industry and improving
classroom learning. This paper compares the effi cacy of an AR training tutorial to a traditional Teaching Assistant (TA)-
led demonstration on undergraduate engineering students’ learning of makerspace tools. A group of undergraduate
mechanical engineering students (n = 47) were trained on the power drill/driver with either a TA-led demonstration or an
AR tutorial using a Microsoft HoloLens. Pre- and post-surveys measured students’ knowledge of the tools, self-effi cacy
for tool usage, and perceptions of the tutorial. Both tutorials resulted in a statistically signifi cant increase in student
learning as measured by content knowledge and self-effi cacy. Students did not exhibit a preference for the AR training
compared to the TA training, indicating that either training option is benefi cial and equitable for a range of student needs.
Considering how diff erent tutorial modalities aff ect the learning of diff erent student populations is important when
developing inclusive classroom activities. Using AR training in lieu of or in combination with TA-led training could help
with staffi ng and scaling challenges in academic makerspaces and hands-on design courses while accommodating diff erent
learning preferences and levels of prior experience.Mechanical Engineerin
Multilayer graphene-based capacitive micromachined ultrasonic transducer design, manufacturing, and characterization
Get PDFMicroelectromechanical system (MEMS) sensors have become ubiquitous in consumer electronic and industrial applications. However, there are some fundamental limits to sensitivity and bandwidth of traditional silicon-based MEMS due to relatively high internal stress and thickness. Chemical vapor deposition (CVD)-grown, multi-layer graphene (MLG) is a promising alternative to silicon because it can be made very thin and has low internal stress, which results in higher compliance membranes. Before MLG is used in MEMS, the mechanical properties must be understood. Nanoindentation experiments were performed on 122 separate 434 nm thick MLG membranes between 100-500 µm in diameter to determine internal stress and Young's modulus. There was an increase in internal stress and modulus after multiple loading cycles which was explained by stretching out of stress-relieving wrinkles in the membrane. This behavior also suggests that the observed decrease in internal stress as radius increases is also explained by wrinkles in MLG since they are more prominent in large diameter membranes. Additionally, it was found that the initial roughness state of the metal catalyst used to grow the MLG did not have an effect on the final tension in the film, which suggests that the transfer method and wrinkles in MLG are the dominant factors determining tension. Measured MLG properties were used to quantify the performance differences between MLG and silicon using a capacitive micromachined ultrasonic transducer (CMUT) receiver as an exemplar MEMS sensor. Modeling showed a >4x improved mechanical compliance and bandwidth attributable to the lower mass and lower internal stress of MLG membranes. Finally, a proof-of-concept CMUT was fabricated and its frequency response tested in vacuum and in air using laser doppler vibrometry and a piezo buzzer. Frequency response of the CMUTs match expected modeling results, so the feasibility of MLG-based CMUTs is shown.Mechanical Engineerin
Improving the assessment of attention bias in depression
Get PDFBiased attention for negative information is a malleable target putatively implicated in the maintenance of depression. However, before testing cognitive theory or progressing with intervention research, better assessments of attentional biases are needed. We focus on two issues plaguing this research: suboptimal task design and poor psychometrics of attention bias measures. In Study 1, we highlight the disastrous effects of poor psychometrics when identifying the best predictors of depression. We used cross-validated elastic net models in a sample of N= 217 adults to predict five outcomes using a multifactorial dataset. Although we included 30 measures of attention bias derived from the traditionally-used dot-probe task, none of the metrics arose as important variables in our models. Additional analyses revealed the psychometric properties of these metrics were so poor, they likely precluded them from being useful predictors. In Study 2, we sought to improve measurement by using an eye-tracking-based free-viewing task better suited for capturing attention bias as it manifests in depression in a sample of N= 130 students. Importantly, we examined the psychometrics by applying factor analysis to assess the unidimensionality of each of our measures. While two metrics met our threshold for unidimensionality, others did not, highlighting the importance of testing the psychometrics of each measure, even when derived from the same data. In Study 3, we examined the role of attention bias across three samples: an eye-tracking sample (Sample 1, N= 129), and two online mouse-tracking samples (Sample 2, N= 79 and Sample 3, N= 154). Consistent with our approach in Study 2, we tested the psychometrics of our measures in each sample. Next, we created mixed-effects models examining the interaction of depression severity and stimuli valence on time spent looking at stimuli. Samples 1 and 2 generally supported a depression by valence interaction, where increasing depression severity was associated with greater gaze time for sad versus neutral stimuli. However, these results failed to replicate in Sample 3. Together, these findings contribute methods for enhancing the assessment of attention bias and provide additional evidence for the role of negatively biased attention in depression. Implications and future directions are discussed.Psycholog
Biosynthesis of the unusual spirocyclic ortho-δ-lactone of hygromycin B : mechanistic studies of radical S-adenosyl-L-methionine epimerase HygY and 2-oxoglutarate-dependent nonheme iron oxygenase HygX
Get PDFNatural product antibiotics have played an indispensable role in improving human life since the discovery of penicillin nearly one hundred years ago. Nevertheless, the biological arms race between microbial pathogens and man continues relentlessly; the former threatens to unravel the latter’s most notable medical breakthrough of the 20th century. Fabricating diverse antibiotics with novel mechanisms of action via biochemical pathway engineering is a promising avenue that can help stem the tide of multidrug resistance. However, our dexterity in pathway bioengineering is limited by our fundamental understanding of the late-stage tailoring enzymes that are often crucial for establishing bioactivity. Radical S-adenosyl-L-methionine (rSAM) and non-heme iron (NHI) enzymes are two enzyme superfamilies that are often employed for late-stage functionalization of nascently formed antibiotics. Central to these two enzyme superfamilies is the utilization of radical intermediates that astonishingly react under precise stereocontrol. The aminoglycoside hygromycin B is a prime example of a natural product whose biosynthetic gene cluster (BGC) encodes one of each: rSAM epimerase HygY and 2-oxoglutarate (2OG)-dependent NHI oxygenase HygX. Both catalyze transformations that are unique within their respective enzyme families. An unusual spirocyclic ortho-δ-lactone exists on hygromycin B, its distinguishing feature among aminoglycosides, whose assembly is thought to require these two enzymes and a glycosyltransferase: C2’-OH equatorial-to-axial epimerization by radical SAM (rSAM) enzyme HygY to give talamine, glycosylation of talamine O3’ with destomic aldehyde, presumably by glycosyltransferase HygD, and O2’’-C1’’ spirocyclization by 2OG-dep. NHI enzyme HygX. Spectroscopic evidence of an additional iron-sulfur cluster christened HygY as a new member of the rSAM SPASM/twitch subfamily, while functional characterization of HygY with synthesized substrate and deuterated analogue confirmed its epimerase activity. Fascinatingly, mechanistic studies of a cysteine mutant uncovered the centrality of this residue in dictating the reaction course of HygY. Its mutation abrogated the somewhat rare redox neutral reaction in the wild type to a more typical oxidation seen in many rSAM enzymes. As for HygX, an analogue lacking 2-deoxystreptamine was synthesized to evaluate its competency as a substrate. While the fate of this analogue could not be determined, 2-OG was completely converted to succinate within 30 minutes regardless of the presence or absence of substrate, highlighting an unusual case of complete uncoupling for this class of enzyme. While a horseradish peroxidase coupled assay did not detect the formation of diffusible hydrogen peroxide, intact protein mass spectrometry studies identified protein autooxidation, indicating that the generated iron(IV)-oxo species is capable of hydroxylating itself in the absence of a suitable substrate.Chemistr