137091 research outputs found
Sort by
Ferroelectric and Monolithic 3D (M3D) Memory Technologies for High-Performance, Energy-Efficient Computing
No full textPresent-day hardware with limited on-chip memory capacity falls short of meeting the speed and energy-efficiency requirement for data-intensive applications such as Artificial Intelligence due to the need for expensive data transfer between processor and off-chip memory. The objective of this research is to mitigate the energy-inefficiency and latency by developing ultra-high density cache memory and computational memory for enabling near-memory computing and in-memory computing. First, a back-end-of-line (BEOL)-compatible dual-gate (DG) W-doped In2O3 (IWO) field-effect transistor (FET) with high-performance, low-leakage and improved reliability is demonstrated, which is promising as access/read transistors for realizing M3D embedded-DRAM. The proposed DG IWO FET synergistically improves the device performance and VT stability, thereby breaking the performance-stability trade-off typically observed in BEOL FETs. Furthermore, a comprehensive reliability model is developed to elucidate the VT shift mechanisms. Second, we study and develop ferroelectric (FE) FET based computational memory for accelerating in-situ training and edge inference in deep neural networks (DNNs). On that front, we develop a scaled Si-channel FE-FinFET process and explores its potential application as a 2bit/cell weight cell for feature extraction and DNN inference. Lastly, this thesis proposes a novel idea of engineering ferroelectric gate stack using superlattice (SL) structure which enhances multi-state programmability of FEFET and thus can enable denser memory and novel functionality such as analog weight cell to accelerate in-situ training and inference of DNN. We experimentally demonstrate a BEOL SL FEFET analog weight cell with a record-high 1000 analog conductance states exhibiting a high degree of linearity and symmetry. We also demonstrate 8 non-overlapping programmed states for the SL FEFET which enables enhanced multi-level-cell operation (3bit/cell) over conventional FEFETs.Ph.D.Electrical and Computer Engineerin
Enhancing Oral Tissue Regeneration using FTY720-nanofibers as a Biomaterial-Based Immunotherapy
No full textOrofacial clefts represent the most prevalent congenital craniofacial abnormality. Surgical complications in cleft palate repair can lead to postoperative oronasal fistula (ONF), a persistent communication between the oral and nasal cavities. ONF negatively impacts a child's eating and speaking abilities, consequently affecting their overall life quality. The current gold standard methods for ONF repair use human allograft tissues; however, these procedures have risks of graft infection and/or rejection. A promising regenerative therapy involves local delivery of FTY720, an FDA-approved immunomodulatory drug, to harness the body’s immune response to create a more favorable healing milieu. FTY720 reduces lymphocyte migration to inflammatory sites and promotes immune cell transition towards a pro-regenerative state. We hypothesize that local bilayer FTY720-NF delivery can accelerate oral cavity wound healing and modulate key lipids, genes, and proteins to reduce inflammation and promote tissue remodeling. This thesis aims to repurpose FTY720 to harness the reparative wound healing system as a biomaterial-based, pro-regenerative immunotherapy following cleft palate and ONF repair. This will be achieved through two specific aims: 1) engineering a bilayer, biomaterial scaffold to deliver FTY720-NF, modulate immune cell recruitment, and promote a regenerative oral wound healing environment, 2) investigating FTY720-NF immunomodulation of key lipid, gene, and protein mediators to enhance ONF tissue regeneration. This research will establish critical insight in the role of immunomodulation for oral wound healing and develop an efficacious treatment alternative for pediatric patients following cleft palate surgery.Ph.D.Biomedical Engineerin
An Uncertainty Quantification-based Methodology for Resource Allocation toward Technology Maturation
No full textTo address the aviation industry's need to decarbonize amid rising travel demand, this dissertation proposes a comprehensive, data-driven methodology to optimize testing strategies for novel technologies, such as hybrid-electric propulsion (HEP). The research is structured in three parts: first, it quantifies how component-level uncertainties impact system performance, identifying battery cell-specific energy as the primary driver of variability through sensitivity analysis. Second, it develops a multi-attribute decision-making framework that holistically prioritizes component testing based on risk and uncertainty, consistently ranking the battery as the highest priority. Third, it translates these priorities into a practical, adaptive test plan that dynamically reallocates resources in response to new information. By integrating these areas, this work provides a cohesive, end-to-end framework for technology maturation that overcomes the limitations of traditional, static methods. The resulting systematic and technology-agnostic approach offers a versatile tool for managing complex engineering development in aerospace and other industries
Group Robustness under the Microscope: How Class Imbalance Shapes the Effectiveness of Last-Layer Retraining
No full textMachine learning models trained with empirical risk minimization (ERM) often rely on spurious correlations, leading to severe performance degradation on underrepresented subpopulations. This thesis provides a comprehensive examination of group robustness in fine-tuned neural networks, focusing on how class imbalance and data composition shape the effectiveness of last-layer retraining (LLR) and related balancing techniques. Through extensive experiments on four benchmark datasets spanning vision and language domains—Waterbirds, CelebA, CivilComments, and MultiNLI—we uncover several surprising and previously underexplored behaviors in worst-group accuracy (WGA).
We first diagnose failure modes in widely used class-balancing strategies. While subsetting, upsampling, and upweighting are intended to improve robustness, we show that upsampling and upweighting frequently cause catastrophic WGA collapse in imbalanced datasets, whereas subsetting may remove structurally important minority-subgroup examples. To address these limitations, we introduce mixture balancing, a hybrid method that combines subsetting with controlled resampling. Mixture balancing consistently improves WGA and mitigates the instability exhibited by prior methods.
We further analyze why LLR—despite modifying only the final classification layer—achieves disproportionate gains in robustness. Contrary to hypotheses based on neural collapse or maximum-margin convergence, our results demonstrate that LLR’s effectiveness is primarily driven by improved group balance in the held-out retraining set, even when group labels are unavailable. Finally, we identify a structural phenomenon, spectral imbalance, in which minority-group feature covariances systematically exhibit larger top eigenvalues, revealing deeper geometric causes of group disparity.
Together, these findings offer a unified perspective on data balance, model scaling, and representation geometry, and provide practical guidance for designing robust models under spurious correlations
Serix: A Reflective Game-Based Learning Experience to Support Young Adults in Navigating Life Transitions
No full textNavigating significant life transitions such as graduation, career changes, relocation, and relationship shifts—presents unique challenges for young adults, often requiring them to adapt to new circumstances and develop effective coping strategies. This study investigates the interplay of personal experiences, social dynamics, and external stressors during these transitions and explores how Game-Based Learning (GBL) can support young adults in managing these challenges. Through user-centered research, including semi-structured interviews and iterative design testing, this study develops Serix, an innovative card-based board game. Serix combines engaging mechanics, reflective prompts, and narrative-driven scenarios to encourage self-reflection, foster adaptability, and support decision-making. The findings demonstrate the potential of GBL to empower young adults with tools to navigate complex transitions while maintaining an enjoyable and meaningful experience.M.S.Industrial Desig
Modeling Urban Flood Risks for Resilient Infrastructure, Emergency Services, and Public Safety
No full textUrban flooding poses severe threats to Emergency Medical Services (EMS), with response times increasing by an average of 25% during flood events and up to tenfold in extreme cases. This dissertation develops a comprehensive framework for estimating and mitigating EMS delay risks during urban floods through advanced flood situation awareness and traffic disruption modeling. The research addresses critical gaps in current methodologies by integrating real-time sensing, data-driven modeling, and risk assessment to support dynamic EMS resource management. The first study establishes a framework for EMS delay risk estimation during urban floods, revealing significant inaccuracies and inequities in state-of-the-art methods when evaluated against real-world EMS data. The analysis demonstrates that lower-income communities not only experience disproportionately greater impacts from urban flooding, but also facing more flood risk underestimation, highlighting the need for more equitable risk assessment approaches. The second study addresses the data quality challenges in flood modeling by developing an error correction method for AI-based river streamflow forecasting. This method reduces the impact of measurement inaccuracies during flood events, improving the reliability of hydrological predictions that serve as critical inputs for urban flood modeling and subsequent EMS risk assessments.The third study introduces a novel application of roadside flood sensors for real-time traffic state estimation during floods. Using a double-layered Bayesian Network framework, this research develops an optimization approach for sensor positioning that maximizes the value of information for traffic disruption assessment, accounting for probabilistic dependencies and uncertainties across connected road networks. Collectively, these studies contribute to building a dynamic, real-time EMS delay risk estimation and mitigation system. The integrated framework supports proactive decision-making for emergency response operations, including ambulance deployment optimization and critical infrastructure recovery prioritization, ultimately enhancing community resilience to urban flooding disasters
Innovations in Characterization and Design of Durable Sustainable Cementitious Systems Utilizing Pozzolanic Materials
No full textThis dissertation investigates how pozzolanic materials influence both the
microstructural and macrostructural properties of cementitious systems, with the overarching goal of advancing the design and implementation of sustainable, durable concretes. Through a combination of experimental, analytical, and microstructural characterization approaches, this work provides new insights into the design of these low-clinker systems, the performance of reclaimed ashes as a supplementary cementitious material (SCM), and the long-term mechanisms of durability observed in ancient Roman concrete.
Low water-to-solid ratio limestone–calcined clay cement (LC3) systems were designed using a particle packing approach to link packing density with hydration, strength development, and environmental efficiency. Strong early-age correlations between the particle packing index (PPI), compressive strength, and an Environmental Performance Indicator (EPi) confirm particle packing as a predictive framework for designing high-performance, low-impact mixtures.
Reclaimed coal ash was evaluated as a sustainable SCM alternative to conventional fly ash, focusing on mitigation of alkali–silica reaction (ASR) and sulfate attack. For ASR mitigation performance, reclaimed ashes outperformed inert fillers, but they did not match Class F fly ash, indicating that higher replacement rates may be required. MicroXRF analysis revealed key differences in alkali transport among standard ASR test methods, supporting the 56-day AASHTO T380 as a representative evaluation protocol. To investigate sulfate attack, LC3 based engineered cementitious composite (ECC) incorporating reclaimed ashes was assessed using both mechanical expansion tests and microXRF imaging. These low-clinker, fiber-reinforced systems demonstrated superior resistance to external sulfate attack compared to portland cement controls, owing to the combined effects of LC3’s dense matrix and fiber-induced crack-width control. MicroXRF provided spatially resolved quantitative data on sulfate penetration and diffusivity, revealing very low diffusion coefficients and confirming the excellent durability of these systems.
The final component of this work applied a novel, non-destructive analytical approach of combining microXRF and solid-sample XRD to a 2000-year-old Roman concrete sample. This technique preserved spatial resolution while enabling phase identification and compositional mapping across the heterogeneous microstructure. The findings point towards the use of seawater and “hot-mixing” of the lime and support the interpretation that Roman concrete durability arises from progressive phase development over time, creating an impermeable microstructure, while also enabling potential post-pozzolanic reactions and self healing to occur.
Collectively, this thesis provides both design methodologies and analytical frameworks for developing durable, low clinker cementitious systems. The particle packing approach offers a quantitative guide for sustainable mixture design; reclaimed coal ashes expand the resource base for SCMs; and spatially resolved characterization techniques reveal fundamental relationships between composition, microstructure, and durability. From the ancient Romans, use of locally available materials, allowance for chemical evolution over time, and
harnessing material-environmental interactions to the benefit offer a strategy for
enhancing the durability and sustainability of modern concretes
Pneumatic Gibbot: Design and Simulation of a Pneumatic Brachiating Passive Dynamic System
No full textAmong locomotion types in robotics, climbing is an area where there is still much to be explored and improved. The locomotion of brachiation (arm swinging) is an energy-efficient mode of climbing most utilized by primates, like the gibbon. It allows for movement that utilizes energy conservation in a similar way to walking and can be simply modeled as a double pendulum. With this inspiration, we created a passive dynamic system that mimics the brachiation movement of a gibbon, inspired by the Gibbon Robot, “Gibbot”, prototype which introduced a unique design that allowed attachment anywhere on a metal wall. This thesis explored the process and challenges of updating the Gibbot to a continuous contact passive pneumatic design that could successfully mimic the graceful brachiation movement. The addition of pneumatics allowed a broad range of surface applications while introducing interesting design and simulation challenges. Once built, the prototype’s functionality was evaluated through lab testing. A simulation was then created in MATLAB, modeled after this design. Prototype data was then used to align the simulation more accurately with real-world behavior. This effort was evaluated and then the simulation was used to explore beyond the constraints of the prototype in the lab and provide information for further design improvement.M.S.Mechanical Engineerin
FMCW Lidar Architectures: Classification, Simulation, and Performance Assessment
No full textLidar, first developed in the 1960s, uses laser light to measure distance. Initially used for rangefinders and altimeters, lidar's role expanded with advancements in hardware and real-time data processing, making it a vital tool in fields like autonomous vehicles, archaeology, and forestry. Unlike radar, which generates a sparser dataset, lidar produces dense point clouds for detailed 3D mapping. While lidar offers higher spatial resolution, it struggles in adverse weather conditions.
There are different types of lidar, including pulsed, Amplitude-Modulated Continuous-Wave (AMCW), and Frequency-Modulated Continuous-Wave (FMCW) lidar. FMCW lidar modulates a continuous laser beam’s frequency to calculate distance, offering advantages such as simultaneous velocity and range measurements, better range resolution, and higher immunity to ambient light.
This thesis develops a comprehensive classification of FMCW lidar architectures, analyzing 186 relevant papers to address inconsistencies in naming conventions and definitions within the field. The findings categorize systems by key architectural distinctions, such as homodyne versus heterodyne configurations, modulation schemes (SSB vs. DSB), and demodulation methods (IQ vs. non-IQ). For example, heterodyne systems involve significant differences in frequency and phase between the local oscillator (LO) and received signal, requiring an intermediate frequency, while homodyne systems have nearly identical LO and received signals, allowing direct mixing in the baseband.
The thesis also presents a lidar simulator based on one of the most prominent FMCW lidar architectures identified in the survey, which reflects a DSB Homodyne Non-IQ architecture. This simulator adds realistic features such as band-limited noise, an amplifier, an Analog-to-Digital Converter (ADC), and most notably, speckle noise. It evaluates simulation performance based on theoretical expectations of metrics like probability of detection, probability of false alarm, and signal-to-noise ratio.M.S.Electrical and Computer Engineerin
Scalable Approaches to Manufacturing Nanocellulose-Glass-Fiber Reinforced Composites
No full textGlass fiber reinforced polymer (GFRP) composites are versatile materials with widespread use in the transportation industry, and the increasing demand for more sustainable technologies have motivated the research in the past decade to focus on improving their mechanical properties while reducing their specific weight with the end goal of increasing fuel efficiency. The goal of this study is to further the understanding of the effect of CNC on GRFP performance and to develop a scalable methodology facilitating CNC incorporation into composites for high volume production.
To accomplish this goal, a novel scalable approach of vacuum-assisted slot die coating on a roll-to-roll is adapted to deposit uniform CNC sizing onto GF fabrics. The coating method is studied using computational fluid dynamics (CFD) modeling and validated using experimental trials to predict operating parameters, coating outcomes, and further the understanding of processing conditions of coating CNC dispersions onto GF fabric. The effect of CNC concentration, functionalization, GF type, and the application of vacuum are evaluated using single fiber fragmentation tests (SFFT) and mechanical tests to determine their influence on GF-matrix interfacial adhesion and GFRP laminate properties. The ability of CNC to enhance GFRP properties in high-volume production is explored using pilot-scale sheet molding compound (SMC) manufacturing, using CNC as a lightweight reinforcement and filler to replace heavy CaCO3 filler within resin matrix while enhancing GFRP performance.
Both coating and SMC studies demonstrate the potential of CNC to greatly enhance GFRP interfacial, tensile, flexural, and interlaminar properties, leveraging the mechanical interlocking and chemical bonding with GF and polymer resins whether CNC is deposited at the GF-matrix interface or within the matrix. This work provides complete scalable methodologies and furthers the understanding of using CNC to produce improved GFRPs for high-volume structural applications