Real-time monitoring of active distribution systems can be done by using micro-phasor measurement units (µPMUs). In the first part of this report, an innovative μPMU placement algorithm is presented to completely observe the system and facilitate state estimation for unbalanced distribution systems. The proposed algorithm considers practical constraints such as single-phase laterals, distributed loads and variable tap-ratios and ensures complete phase observability while minimizing the number of μPMUs needed. However, complete observability of distribution systems may not be economically and practically viable. Hence, in the second part of this report the challenge of limited availability of μPMUs in distribution systems is addressed using deep learning (DL). The proposed DL-based method offers superior accuracy with fewer μPMUs compared to conventional least squares method, even during topology changes. The robustness of the deep neural network (DNN) used in DL is further evaluated by considering realistic measurement errors in the μPMUs, ensuring the practicality of the approach.In the third part of this report, the research delves into the verification of the DNN for distribution system state estimation (DSSE) by analyzing its performance under input perturbations. A mixed-integer linear programming (MILP) approach is proposed to analytically verify the DNN's robustness. This ensures that the DNN output remains bounded given a bounded perturbation in the input. This essentially builds trust in using DNN-based solutions for critical tasks such as power grid monitoring. Finally, due to the scalability challenges of MILP problems, a linear bound propagation method is proposed for DNN verification. The proposed method significantly increases the speed (of verification) and makes the verification problem scalable for wider and deeper DNNs. In conclusion, this research provides a comprehensive monitoring framework for distribution system operators, empowering them to effectively manage the addition of distributed energy resources. It combines optimal μPMU placement, a robust DNN-based DSSE, and a rigorous DNN verification analyses. By ensuring reliable and trustworthy system monitoring, the approaches developed in this report can contribute to a more stable, resilient, and renewable-rich power distribution grid.

Over the past decade, the success of deep learning models has largely depended on the availability of extensive training data and the assumption that training and target data distributions are independent and identically distributed (i.i.d.). However, deviations from these conditions often reveal the models' brittleness, as they struggle with distribution shifts—discrepancies between training and testing datasets arising from various factors. To address this challenge, this dissertation leverages Generative Adversarial Networks (GANs) to parameterize distribution shifts, using a novel single-shot target aware (SiSTA) adaptation technique. This approach updates a GAN with a target domain example to generate synthetic samples that facilitate the effective adaptation of predictive models to target conditions. Beyond synthetic data generation, GANs are integral to digital imaging restoration tasks such as image denoising and super-resolution. However, they often perform poorly when the input data deviates from the training distribution. To address this, a technique called SPHInX (Style Projection Heads for Inverting X) is developed to enhance GAN inversion capabilities, thereby improving the model's ability to handle out-of-distribution images. However, GANs struggle with semantic shifts caused by label shifts and class imbalances. Vision-Language Models (VLMs) are more effective in these scenarios. This dissertation introduces CREPE (CLIP Representation Enhanced Predicate Estimation), a framework that leverages VLMs to improve nuanced visual relationship prediction by better contextualizing the visual representations. A key part of mitigating model failures involves understanding when and why these failures occur. This dissertation proposes a novel strategy for estimating failures by parameterizing the decision rules learned by predictive models through VLMs. This approach refines the mechanism for failure estimation, allowing for more precise identification and correction of failures across various scenarios. When there is a scarcity of data, the challenge becomes even more pronounced. In such cases, the problem is commonly modeled as a distribution of small tasks. This dissertation addresses this issue by exploring the use of knowledge graphs to dynamically modulate the weights of predictive models. This approach enables the models to adapt their decision rules effectively, enhancing flexibility and effectiveness in real-world applications. Overall, this dissertation presents robust methodologies for understanding and mitigating adverse effects of distribution shifts {on the performance of deep learning models, significantly advancing the adaptability and reliability of these models in dynamic environments. These contributions lay a foundation for future research into developing artificial intelligence systems that are capable of sustaining reliable performance across varying conditions.

Achieving a zero net emissions economy in line with science-based climate goals requires coordinated technical efforts at an unprecedented scale and pace. This work identifies and addresses high impact gaps in systems and tools needed to facilitate decision-making analyses, with a focus on optimization in the context of California’s building energy transition programs. To utilize waste heat from a concentrating solar power tower to deliver process or comfort cooling, technology options are compared across a range of ambient temperatures with each technology being optimized at each design point. Tools for modeling sorption cooling cycles are developed that lower the barrier to entry and enable near real-time design iteration and optimization. An analysis of dual fuel space heating technology potential demonstrates the impact of adopting a connected and optimized control strategy on the operating cost and emissions reduction tradeoff. Based on experience with these analyses, additional discussion points to gaps required to support further optimization analyses and faster iteration of calculation tools used in the regulated energy efficiency industry.

Layer-wise extrusion of cement pastes, mortars, or concrete is the most commonly used technique in three-dimensional (3D) concrete printing. Understanding the behavior of the printed binder after placement is crucial for optimizing the properties of 3D-printed elements. This research, conducted in two stages, fresh and hardened state responses, elucidates the post-extrusion mechanics of cementitious binders to enhance print quality. In Stage-1, a novel technique for characterizing 3D printable mortar binders using the green compression test (GCT) is introduced. Equations based on GCT parameters were established to predict buildability, the maximum height that can be sustained without significant deformation or failure. These equations were able to predict buildability and failure mechanisms over time accurately. Stage-2 investigates the mechanical response of hardened 3D printed binders, focusing on inter-layer and inter-filament interfaces, mixture types, and fiber content. Variation in interface positioning and the addition of fibers (0.28% by volume) improved flexural response while maintaining comparable compression strength. However, it did not eliminate anisotropy in compression, and mechanical properties remained inferior to cast counterparts. Next, a numerical model was developed, using cohesive zone finite elements to represent joints and an orthotropic visco-elastic-visco-plastic material model for the bulk filament. This model effectively predicted the mechanical response of 3D printed elements, accurately capturing anisotropy under uniaxial compression. This highlighted the importance of properly characterizing joints and selecting material models. Finally, ultra-high performance 3D printable mixtures were developed using a low water-to-binder ratio mixture with a higher content of water-reducing agent, achieving compressive strengths exceeding 100 MPa at 28 days. This mixture resulted in reduced anisotropy while providing strengths comparable to mold-cast specimens. Incorporating high fiber volume (1.5% by volume) into this mixture significantly enhanced compression and flexural responses. Composite material sections, created by printing different mixtures in various layers, showed comparable mechanical responses while improving cost and environmental efficiency. The findings of this research contribute to precise failure prediction during printing, propose methods for better mechanical responses in printed products, and offer insights into cost-effective and environmentally efficient section design through composite material printing using 3D concrete printing.

Glioblastoma multiforme (GBM) is an aggressive, high-grade glioma for which available treatments produce limited success. The lack of effective GBM therapies stems from insufficient knowledge of its heterogeneous biology. Omics analytical techniques allow opportunities to improve understanding of GBM and include radiomics, which is a quantitative approach to imaging, and transcriptomics, which describes gene expression data. A specialized high-grade glioma cohort of spatially matched radiomics and transcriptomics revealed associations between magnetic resonance image (MRI) signal and underlying tumor biology. T1-weighted contrast-enhanced (T1+C) MRI measurements identify regions of blood-brain barrier (BBB) breakdown by detecting contrast agent leakage into the central nervous system. Anti-inflammatory immune cells were present in areas of high T1+C signal, suggesting contrast leakage and immune cell transmigration occur at the same compromised BBB sites. The T2-weighted (T2W) and mean diffusivity (MD) metrics register water presence and diffusion. Angiogenic genes such as VEGFR1, VEGFR2, and ROBO4 were expressed in areas of high T2W and MD signal and areas with histologically-measured angiogenic vasculature, indicating T2W and MD indirectly register sprouting angiogenesis. These findings illustrate the biological processes visualized by MRI. Radiation resistance is another reason for limited therapeutic success in GBM and has been linked to glioma stem cells (GSCs), which are a highly radiation-resistant subpopulation of GBM tumors. Single-cell transcriptomics of irradiated GSCs explored radiation-induced gene expression changes. A novel subpopulation expanded post-irradiation, termed the radiation-induced cluster (RIC). The RIC was observed in multiple patient-derived GSC lines, demonstrating its presence across unique patient phenotypes. High expression of RIC marker genes was also associated with worse GBM patient survival. Further, RIC marker genes were related to external stress responses but were not previously identified as part of the GSC radiation response, suggesting the RIC confers radiation resistance through a previously unknown mechanism. Additional study of this novel subpopulation will provide insight into the biology of GSC and GBM radiation resistance to guide future therapeutic development.

Colorectal cancer is the third most common cause of cancer in the United States.In 2024 alone, 150,000 new colorectal cancer cases are expected in the USA. Early detection of these lesions are a key factor in significantly improving the life expectancy of the patient. Most cancers are resected endoscopically, some case warrant a surgery for tumor resection. It is often difficult to recognize a new tumor growth because of the significantly small size of the tumor tissue which makes presurgical tumor localization very crucial. Endoscopic tattooing is the most common way of localizing the tumor mass. In this technique a dye, commonly dark in color to improve identification, is injected in the colon near a suspected tumor site, for easy identification. Current commercially available endoscopic tattoos have a significant problem of diffusing into the surrounding area and in some cases, losing their contrast, which makes is a very unreliable technique. This thesis is focused on developing an endoscopic tattoo, with mucoadhesive properties to solve the problem of diffusion while maintaining a dark contrast while also having dual modal imaging applications such as white light imaging and fluorescence imaging to further aid successful tumor localization.,

The purpose of this study is to understand how external evaluators attempt to make sense of firm performance. Specifically, I investigate the seemingly puzzling phenomenon whereby external stakeholders make negative attributions for “good” organizational performance, i.e., assume that the apparent good performance stems from causes different from those explainable by the firm’s positive motives, qualities, characteristics, or behavior. The study challenges our understanding of how credit and blame are assigned when performance is good or poor, respectively. While foundational work offers general guidance on the sociocognitive aspects of venture evaluation, there is little work explaining how observers identify, collect, and process information to make these judgments within the context firm performance evaluations. The study employs an inductive, grounded approach to explore two main questions: (1) "How do external evaluators make attributions of firm performance?" and (2) "Why does good performance sometimes yield negative attributions?" In doing so, I focus on the factors, processes, and reasoning that lead external evaluators to perceive superior performance as “too good to be true.” The findings offer a more nuanced understanding of firm performance attributions and evaluations, shed light on the roles of different stakeholders and contributing contextual factors, and broaden our understanding of attributional processes in the domain of strategic management.

Methanotrophic bacteria play an important role in reducing methane emissions by consuming methane as their primary energy and carbon source. Thus, it is important to define the identity, diversity, and distribution of methanotrophic bacteria to understand their contributions to methane removal.To investigate methanotrophs from three geochemically distinct tropical peatlands of the Amazon, several treatments were applied to enrich methanotrophs. These enrichments varied in their pH, type of media used, single-carbon substrate(s) added, and the presence or absence of organic nitrogen. Conditions were selected based on the gradient of pH and nutrient concentrations characteristic of the three peatland sites of interest and combined to maximize the variety of culturable methanotrophs. Methanotrophic enrichments were then selected for isolation efforts, establishing various lineages that were purified and then identified by their DNA sequence. This study evaluated whether methanotrophs from geochemically distinct peatlands would prefer distinct culturing conditions under the hypothesis that various methanotrophs will grow best in conditions that mimic the selective pressure(s) of the environment that they arose from. The results show (a) the presence of multiple culturable species of methanotrophic bacteria, including strains of the species Methylosinus sporium, Methylocystis parvus, Methylocapsa acidiphila, Methylocystis hirsute, and Methyloferula stellata, as well as representatives of a potential novel species related to Methylocapsa acidiphila strain B2, (b) observable methanotrophic activity within enrichments from which no responsible organisms were isolated indicates unculturable methanotrophs from Amazon peats exist within the peat communities, and (c) the type of media and carbon sources used for enrichment and isolation of methanotrophs did not have a significant effect on the identity of methanotrophs isolated by traditional methods. Overall, the results show that Amazon peatlands contain a significant presence of culturable and yet-to-be-cultured methanotrophs. In addition to the progress achieved here, further variation in enriching and isolating strategies is still needed.

Language has a valuable impact on the world and is a vital aspect of most socialinteractions. Previous research has shown that immigrants who have language barriers struggle with feeling as though they do not belong in their host country. Additionally, having a low sense of identity and life satisfaction can exacerbate these negative feelings. The current study examined how identity and life satisfaction mediate the association of a bilingual’s acculturation attitude with language and sense of belonging to both language groups. The study consisted of 370 participants who were recruited through Prolific and the ASU SONA system and completed a 20 minute online self-reported survey through Qualtrics. I separated the bilinguals into four acculturation groups and then compared assimilation, separation, and marginalization to the integration group and looked at their sense of belonging for both language groups. Overall, I found that those with an assimilation, separation, or marginalized orientation have a lower sense of belonging for both languages through identity with the other language only and not through identity with English. Additionally, life satisfaction only mediates the relationship between separated individuals and their sense of belonging to both languages when compared to integrated individuals.

Clustered regularly interspaced short palindromic repeats (CRISPR)-based DNA Adenine Base Editors (ABEs) represent a groundbreaking advancement in precision genome editing, holding great potential for treating human genetic diseases caused by single nucleotide polymorphisms (SNPs). ABE8e, the most efficient ABE to date, catalyzes the conversion of adenine to guanine, introducing targeted point mutations into genomic DNA. This technology leverages the fusion of CRISPR-Cas9 molecular machinery with a single-stranded DNA (ssDNA)-specific adenosine deaminase evolved from a bacterial tRNA-specific adenosine deaminase (TadA) through directed evolution. This interdisciplinary study, employing computational (molecular dynamics simulations and free energy simulations) and experimental [ensemble Förster Resonance Energy Transfer (FRET), thermal stability assessments, and in vitro activity assays] approaches, reveals the biophysical foundations underlying ABE8e’s 500-fold increase in DNA base editing efficiency compared to other generations. Here the key to this enhancement is shown to be the stable dimerization of the deaminase domain (TadA8e). Its strategic juxtaposition to Streptococcus pyogenes Cas9 (SpCas9) and DNA substrate is achieved via critical interactions involving residues in the TadA8e docking domain (R98 and R129), the residues in the RuvC domain of SpCas9 (E1046) and the phosphates in the backbone of the DNA substrate, uniquely established when TadA8e operates as a stable dimer. It is revealed that T111R and D119N in combination with N122H, mutations introduced to TadA during the evolution of ABE7.10 to ABE8e, drive TadA8e dimerization and enhance DNA editing efficiency of ABE8e. However, ABE8e’s dimerization and aggregation hinder its delivery to the cell using cell-penetrating peptides (CPPs). Thus, a monomeric form of ABE8e is engineered by disrupting the hydrophobic dimerization interface of TadA8e using the QTY code, however, the TadA8e dimers persisted. Ongoing computational simulations aim to identify additional critical residues for efficient disruption of TadA8e dimer.Overall, these findings illuminate the molecular mechanisms driving ABE8e’s improved performance and suggest new engineering strategies aimed at mitigating off-target effects, enhancing editing efficiency, and streamlining cell delivery processes all of which are crucial for the therapeutic application of precision genome editors.