137091 research outputs found
Sort by
Physics-Aware Statistical Learning for Scientific Computing and Decision Making in Engineering Applications
No full textMotivated by the convergence of physical systems and data-driven technologies under the industry 4.0 paradigm, the research addresses two central challenges: how to emulate nonlinear dynamics governed by physical laws, and how to reduce the cost of sensing and inference in inverse problems. This dissertation investigates physics-aware statistical learning as a principled and computationally efficient framework for modeling complex engineering systems and supporting high-stakes decision-making. By integrating physical knowledge with statistical and machine learning techniques, the proposed approaches enable interpretable, scalable, and uncertainty-aware modeling for digital twin development, advanced manufacturing, and environmental monitoring.
Chapter 2 presents a physics-aware surrogate modeling framework for nonlinear dynamical systems, with a focus on aircraft-UAV collisions. The full-order model derived from finite element analysis (FEA) is computationally expensive, so the proposed method constructs a reduced-order model using proper orthogonal decomposition (POD) and then builds a two-stage statistical model: a multivariate Gaussian process (GP) captures the mapping from physical parameters to external forces, and a function-to-function regression (based on functional principal component analysis) learns the dynamic system response. The resulting surrogate achieves high prediction accuracy while offering significant speed-up compared to direct FEA, making it a practical tool for assessing collision severity under varying impact conditions. We show that the proposed physics-aware statistical model can achieve a 12\% out-of-sample mean relative error, and is more than times faster than Finite Element Analysis (FEA).
Chapter 3 addresses the inverse problem of ultrasonic weak bond detection in advanced manufacturing. We propose a physics‑aware statistical ML framework for ultrasound testing in advanced manufacturing: combining function‑to‑function Gaussian Process regression with Amortized Variational Inference to estimate latent physical variables reflecting adhesive bond integrity and strength. This approach integrates rich functional data with physical modeling for robust integrity assessment. The method enables estimation of interfacial stiffness from high-dimensional ultrasound data while mitigating the effects of ultrasonic couplants and measurement artifacts. By incorporating frequency-domain information and enforcing smoothness in the inference network, the model achieves physically interpretable and real-time estimations of bond quality. Validation on both simulated and experimental data demonstrates the method’s ability to perform scalable and accurate inspections, providing a valuable tool for nondestructive evaluation and quality assurance in adhesive-bonded structures.
Chapter 4 focuses on sparse sensor allocation for detecting leaking emission sources, such as methane, over large spatial domains. The problem is formulated as a bilevel optimization where the upper-level objective minimizes the integrated mean squared error (IMSE) of estimated emission rates, while the lower level solves a constrained inverse problem accounting for non-negativity, sparsity, and physical uncertainties such as wind conditions. Two solution algorithms are proposed: repeated sample average approximation (rSAA) and stochastic bilevel approximation (SBA), both implemented with GPU acceleration. The framework significantly improves inference accuracy and adapts sensor deployment based on underlying physics and uncertainty, demonstrating its utility in environmental monitoring and smart sensing systems. Convergence analysis is performed to obtain the performance guarantee, and numerical investigations show that the proposed approach can allocate sensors according to the parameters and output of the forward model. Computationally efficient code with GPU acceleration is developed.
Finally, Chapter 5 summarizes the original contributions and outlines future research directions. This dissertation introduces a unified framework for physics-aware statistical learning that enables interpretable and efficient modeling of complex engineering phenomena. Both the methodologies and applications advance the state of the art in integrating physical knowledge with statistical learning, laying a foundation for next-generation decision-making systems in engineering applications.Ph.D.Industrial Engineerin
Coverage Control for Constrained Heterogeneous Multi-Robot Teams
No full textCoverage control is a framework for effectively distributing a team of mobile robots over a domain of interest through coordinated control. Deploying real robots typically requires the consideration of their physical limitations and external constraints, such as connectivity maintenance or collision avoidance. This thesis investigates the incorporation of heterogeneous robot characteristics and various constraints into the coverage control framework. The most fundamental characteristic of a mobile robot is its mobility, which can also be seen as a constraint. This thesis begins with a coverage problem where a team of robots with different maximum speeds need to optimally cover a domain in terms of travel time. We then discuss how robots with different mobility types, e.g., ground and water, can perform coverage on a domain comprised of multiple terrain types while remaining in their compatible regions. In addition to mobility constraints, connectivity maintenance is crucial for coverage control, as it allows robots to collectively cover a domain through data sharing. Therefore, we examine how robots can establish and maintain a sparse communication network during coverage. Finally, we introduce a coverage control framework for aerial robots that effectively covers a given domain by balancing sensor resolution with the area of sensing and leveraging coverage redundancy.Ph.D.Electrical and Computer Engineerin
Synthesis and Modifications of Palladium Icosahedral Nanocrystals for Electrocatalytic Reactions
No full textEffective electrocatalysts are critical in fuel cells (e.g., direct formic acid fuel cells, DFAFCs) for accelarating the electrode reactions and thus the conversion of chemical energy into electricity. However, most of current catalysts are made of platinum group metals (PGMs), which are notoriously expensive and have extremely low abundances in the Earth’s crust. To achieve cost-effective use of PGMs (i.e., high mass activity), it is imperative to enlarge the atom utilization efficiency and the electrochemically surface area (ECSA) without sacrificing the specific activity (SA). For oxygen reduction reaction (ORR, the cathode reaction of DFAFCs), current electrocatalysts, including 3-nm Pt nanocrystals and hollow Pt nanocrystals (i.e., nanoframes), have large ECSAs but low SAs due to the poorly-defined surface structure. In this work, I used Pd icosahedra as the template to synthesize Pt nanocrystals with both large ECSAs and high SAs. I first synthesized Pd icosahedra with uniform sizes tunable in the range of 7–20 nm by combining Ostwald ripening and seeded growth. Using the uniform Pd icosahedra as a template, I developed synthetic strategies to obtain small Pt cones and Pt icosahedral nanoframes enclosed by compressively-strained {111} facets. The twin defect of vertices and edges of a Pd icosahedron makes them preferential sites for the nucleation and deposition of newly-formed Pt atoms. The small Pt cones and Pt icosahedral nanoframes were synthesized by confining the Pt adatoms to the vertices and edges, respectively. The conformally-deposited Pt replicated the surface structure of the vertices or edges, generating Pt nanocrystals covered by compressively-strained {111} facets. After removing the Pd template, both the small cones and nanoframes were found to exhibit greatly-enhanced SA and MA toward ORR, compared to the commercial Pt/C catalyst. In addition, I also demonstrated a synthetic method to obtain PdHx icosahedra (x = 0–0.7) that showed superior catalytic performance toward formic acid oxidation (FAO, the anode reaction of DFAFCs). Firstly, the phase transformation from Pd to PdHx was significantly facilitated by achieving a “single-phase pathway”. Then, PdHx icosahedra (x = 0–0.7) were easily obtained by leveraging the “single-phase pathway” and showed enhanced SA due to the insertion of hydrogen atoms. The tensile strain in Pd icosahedra greatly decreased the chemical potential for hydrogen atoms, making it hard for hydrogen to escape from PdHx (x=0–0.7) icosahedra at elevated temperatures or during FAO. Taken together, this dissertation focuses on the rational synthesis of Pd icosahedra-based nanocrystals with well-controlled shapes, strain distributions, and compositions, which were found to exhibit remarkable catalytic performances toward ORR or FAO, the essential electrode reactions involved in DFAFCs.Ph.D.Chemical and Biomolecular Engineerin
QUASI-PERIODIC BRAIN PATTERNS AND SLEEP DURING REST AND N-BACK TASKS
No full textQuasi-periodic brain patterns (QPPs) provide a representation of functional architecture between the Default Mode Network (DMN) and Task Positive Network (TPN) during functional magnetic resonance imaging (fMRI). Previous research has identified fluctuations in QPPs during both rest and task performance that are affected by attentional focus and arousal (Abbas et al., 2019). Individual sleep duration with daytime sleepiness has been shown to affect arousal levels and reduce DMN connectivity in young adults (Ward et al., 2013), which alters connectivity between the DMN and the TPN. This study examines how fluctuations in arousal affect brain network patterns during both rest and working memory tasks in healthy college-age students. We monitored sleep quality using wrist actigraphy watches for three days before the scan. Participants were separated into good and poor sleepers based on sleep efficiency scores. Each participant underwent an fMRI and performed 0-back and 2-back tasks, as well as a rest scan. Overall we found that good sleepers were faster and more accurate in 0-back than 2-back compared to poor sleepers. The differences between poor and good sleepers in the 2-back task were less drastic, and this could be due to a more cognitively demanding task requiring more resources in both sleep groups. During rest and 0-back TPN subnetwork frontoparietal control network (FPCN) is positively correlated with DMN but decouples during the 2-back task. The data suggests that the network driving the differences between DMN and TPN is FPCN. Additionally, poor sleepers have lower FPCN amplitude and lower positive correlation with DMN than good sleepers during rest. Overall, our results suggest that the relationship between functional connectivity and brain networks changes with arousal and relates to task performance.UndergraduateNeuroscienc
Thulium Atoms Embedded in a Solid-State System of Neon for Quantum and Sensing Applications
No full textInvestigating the magnetic dipole ²F₅⸝₂ - ²F₇⸝₂ transition of thulium and its interactions with a crystal host matrix, we embed thulium atoms in solid-state neon and conduct a spectrum-wide survey to build upon current literature regarding thulium in solid-state noble gases and evaluate Tm:Ne’s potential as a qubit or quantum sensor. Rather than use prior methods of spectrum-hole burning to examine the structure of the system, we employ spectrum-wide burning with a pump-probe experiment. We found ten of twelve predicted features in a thermally stable complex pumped at 1139.65 nm. The energy spacings of this trapping site follow the interval rule, and its allowed transitions are subject to magnetic dipole and electric quadrupole selection rules.UndergraduatePhysic
Lipids, the new player in the game: Identifying Lipid Biomarkers and Potential Efflux Routes after Mild Traumatic Brain Injury
No full textTraumatic brain injury (TBI) is a heterogenous disease that affects millions of people worldwide. The vast majority of TBI incidences are classified as mild and prevalent in contact sports both recreationally and professionally, and military populations. Current methods to diagnose TBI rely heavily on self-reported criterion, which may lead to misdiagnosis in milder injuries. There is a need for quantitative metric to accurately diagnose TBI. Biomarkers have emerged in last couple of decades and shown strong predictive power in patient outcome, especially in the moderate-to-severe injury phenotypes, but mixed result in milder injuries. Lipids may serve as ideal biomarker of TBI due to the brain’s high lipid content and lipophilic nature of the blood brain barrier. The work presented in this thesis sought to investigate lipidome alterations in the brain and the blood and understand the contribution of the glymphatic system to biomarker efflux. A clinically relevant closed head injury model was used to integrate lipid and cytokine alterations after single and repetitive mTBI (Aim 1). I found a reduction of pro-inflammatory cytokines in the brain that correlated with lipid decrease after repetitive mild TBI and lipid increase after single mild TBI. Compare the lipidome alteration in the brain and blood compartments after single and repetitive mTBI (Aim 2). I identified overlapping lipids in the brain and blood compartments and found greater changes in the serum compared to the brain, identified brain-specific lipids that elevated in the serum, and identified lipid biomarker panels that discriminate between injury and sham control with overlapping pathways. Lastly, I examined the contribution of the glymphatic system on surrogate biomarkers and endogenous biomarker release from the brain to peripheral tissues (Aim 3). I found TBI alters fluid flow, evidence of size dependent flux in the brain, and the elevation of endogenous biomarkers in the nasal tissue compartment.Ph.D.Biomedical Engineerin
Scalable Asynchronous Actor-based Approaches for Distributed-Memory Parallel Applications
No full textDistributed-memory parallel applications are the backbone of large-scale data analytics and modern computing, forming foundations that enable the efficient distribution and execution of large-scale computations across heterogeneous and distributed systems. Classical primitives for data analytics, including Bulk Synchronous Parallel (BSP) models, focused on moving data to compute. Although these primitives were good enough for past High Performance Computing (HPC) applications, they are highly inefficient for the fine-grained asynchrony and data distribution required by modern large-scale data analytics applications, motivating the need for new architectural approaches for data analytics. Alleviating these challenges inherent in distributed computing environments, the underlying Fine-Grained Asynchronous Bulk Synchronous Parallel (FABS) model in the Actor-based programming system HClib-Actor proposes moving compute to data via asynchronous active messages. This Actor-based approach presents a lightweight, asynchronous computation model that utilizes fine-grained asynchronous actor messages to express point-to-point remote operations, allowing fine-grained, distributed, asynchronous, and scalable executions across systems. We realize the efficacy of the actor-based approach through exploring multiple perspectives and facets of distributed computing, including algorithm design, runtime systems, and system-level optimizations. The Algorithm perspective ("Crafting Scalable Distributed Applications") elucidates novel algorithms tailored for scalable, distributed applications in a wide variety of application domains. Through theoretical analysis and empirical evaluations, we demonstrate the superior scalability and efficiency in the asynchronous actor-based model compared to traditional parallel computing paradigms. The Runtime perspective ("Empowering Distributed Systems with Actors") delves into enhancements made to HClib-Actor, including extending termination protocols, introducing light-weight global termination schemes, and cloud environment deployment. It details optimizations incorporated into HClib-Actor, showing insights that underscore the critical role of runtime systems in facilitating scalable distributed computing. Finally, the System-Level perspective ("Architecture-Aware Optimization and Tuning") signifies the importance of aligning distributed-memory parallel applications with underlying hardware architecture characteristics for achieving peak performance by delving into hardware-specific optimizations and architectural considerations on diverse hardware platforms. We elucidate the importance of architecture-aware bindings and software-level buffer sizes on distributed computing efficiency and bandwidth. Through the realization of the three perspectives and facets of distributed computing, we offer a comprehensive scalable asynchronous actor-based approach for distributed-memory parallel applications. The synthesis of these perspectives, demonstrated through the systematic transformation methodology as an implementation guideline, provides researchers with a practical framework for adopting actor-based approaches in their own applications that are based on traditional models. By integrating these insights, we provide valuable contributions to the advancement of distributed computing, forming foundations for more efficient and scalable distributed and parallel applications in diverse computing landscapes.Ph.D.Computer Scienc
Collision Induced Self Organization in Shape Changing Robots
No full textThis dissertation explores how collective behaviors emerge in ensembles of active, shape changing robots that interact through collisions on a frictional substrate. This work extends the
granular matter paradigm to encompass internally actuated, concave robots, enabling dynamic
self-organization via shape and gait coordination. Initial studies reveal how pinned robot collectives spontaneously settle into low-rattling, repeatable motion patterns, highlighting the role
of environmental coupling in selecting stable configurations. Focusing next on minimal interaction units, the work uncovers a novel binding mechanism in gliding dyads, where timed,
repulsive contacts and shape-induced concavity produce long-lived gliders. Further analysis
shows that breaking time-reversal symmetry via non-reciprocal gaits enables robust, steerable transport through non-commutative dynamics. In both studies, minimal feedback aids
in harnessing and tuning emergent behavior, laying the groundwork for task-oriented control.
Finally, the study extends to many-body systems, revealing how local interactions scale up to
complex structures—such as chains and loops—whose morphologies depend on gait design.
Across all regimes, template-dictated collisional interactions guide the emergence of persistent, programmable behaviors in robotic granular systems.Ph.D.Physic
Brain Networks Underlying Individual Differences in Attention Control
No full textAttention control as a trait variable is a crucial cognitive ability correlated with a range of important life outcomes. However, the underlying individual differences in the brain mechanisms of attention control have not been explored in a large-scale study. I will present the results from my thesis (N=191), in which I investigated where in the brain, trait-level attention control is embedded, and how it interacts with state-level fluctuations during cognitive demand. Using the quasi-periodic pattern analysis, I found that individuals with higher attention control demonstrated more pronounced frontoparietal control network (FPCN) and the default mode network segregation, stronger FPCN and dorsal attention network coupling, and greater FPCN and the locus coeruleus coordination, especially under heightened cognitive demand. Notably, I also observed meaningful differences during resting-state scans, underscoring the trait-like consistency of these neural patterns. These findings highlight that stable individual differences in attention control are reflected in brain networks and neuromodulatory synchronization in response to momentary demands.Ph.D.Psycholog
Testing Darwin's Naturalization Hypothesis and Elton's Bio-resistance Hypothesis in North American Breeding Birds
No full textInvasive species cause environmental and ecological havoc, and it is of great importance to understand why and how they happen. Two influential hypotheses have long been at the center of debate around invasive species: Darwin's naturalization hypothesis, which states that invaders that are less closely related to natives will be more successful, and Elton's bio-resistance hypothesis, which states that communities with higher levels of biodiversity will be more difficult to invade. Many studies have been conducted to test these hypotheses, with mixed results. However, studies which examine these hypotheses through the lens of functional traits and functional diversity are rare. I use data from the Breeding Bird Survey (BBS) to analyze if the native functional diversity has any effect on the non-native species, and if the difference between native and non-native traits is larger, or smaller than expected by chance. I used four different measures of functional diversity: functional richness, functional evenness, functional diversion, and functional dispersion. I found that what affects the non-native species presence, proportion, or longevity is extremely variable depending on which functional diversity metric was used for the
native community, or the metric used to represent the non-natives. Additionally, at the surveyroute scale, Darwin’s hypothesis is neither supported nor refuted. At the scale of bird conservation region however, non-native birds tend to be less similar in mass and more similar in tarsus length to natives. I also found a trend towards bird communities being composed of more large species, even when only Passerines are examined. Overall, this study highlights several issues in the literature that attempts to solve Darwin’s conundrum and the invasion paradox: the metrics used to investigate these hypothesis matter significantly, and more examinations of how these hypotheses relate to functional diversity and functional traits need to be carefully conducted.M.S.Biolog