22689 research outputs found
Sort by
Quantifying Heat Loss in an Industrial Condensate Network and Proposing Methods of Condensate Heat Reclamation for Cost and Energy Savings
Get PDFWith the advent of carbon footprint regulations and a stronger focus on watchful energy consumption, industrial plants that constitute a proportionally large share of pollution are paramount in the fight against climate change. A promising, widely applicable method is condensate recovery, which recycles residual heat energy and liquid condensate back into the process where it can be sustainably repurposed. Within a food-grade Polyols sweetener plant, individual unit operations must be analyzed to identify the most consequential points of condensate reuse. This study critically scrutinizes individual sections of the plant’s process flow to isolate areas that would most benefit from process improvement projects. Multiple factors including projected energy savings, resulting cost savings, and impact are examined to narrow down the most ideal locations for investment. Further studies and deeper, continual observation of operating conditions and parameters would be invaluable in gaining a complete understanding of the process with concrete findings to justify future funding and inspire innovation to tackle the energy crisis
The Wagner Group in Africa: An Effective Kremlin Tool?
No full textHow effective is the Wagner Group and its affiliated network in achieving Russia’s strategic objectives in Africa? I explore this question by conducting an analysis of Soviet and Russian activities since 1945 in select African countries through the lens of the diplomatic, informational, military, and economic (DIME) instruments of national power. Based on this analysis, I conclude that Russia’s objectives in Africa are enduring, although its means and methods for achieving these objectives have evolved. By using the Wagner Group as a multi-purpose military tool, the Kremlin has been able to expand its geopolitical influence across Africa and exploit Africa’s natural resources while minimizing the risks and costs associated with direct Russian intervention.
During the Cold War, the Soviet Union cultivated political alliances, developed economic ties, secured access to strategic resources, and established a military presence in Africa based on geopolitical ambitions, economic motives, and its need for external legitimacy and recognition as a superpower. Two decades after the fall of the Soviet Union, the Kremlin under the leadership of President Putin aggressively re-engaged with African countries by leveraging historic Soviet ties to re-establish and expand its influence using all four DIME instruments of national power. The new Russian approach to its engagements in Africa is more pragmatic and less ideological, leveraging the employment of the Wagner Group and its affiliated network to secure its geopolitical and economic interests in unstable African regions that are rich in natural resources.
The Wagner Group has achieved tactical success in some African countries like the Central African Republic, while experiencing failure in others like Mozambique. However, the value of the Wagner Group to the Kremlin goes beyond these tactical outcomes because of the minimum risks and costs associated with their employment, which has allowed Russia to position itself as an unconditional security partner to African countries. This narrative is very effective in those African countries where authoritarian regimes seek coup proofing or in other fragile states experiencing growing insurgencies. Thus, the Wagner Group and offshoots like the Africa Corps are effective multi-purpose tools in achieving Russia’s strategic objectives in Africa
On the Role of Explanation-Seeking on Inferences and Learning in Early Childhood
Get PDFExplanation plays an important role in guiding causal inference and learning in both adults and children. Past research on the development of explanatory reasoning has largely focused on asking children to generate explanations or to evaluate candidate explanations. These explicit tests of explanatory thinking leave open the question of whether and how young children spontaneously use explanations to guide their thinking in the absence of prompting. The work in this dissertation sought to characterize spontaneous explanatory thinking in early childhood. In Chapter 2, I asked whether children indeed spontaneously seek explanations. Through an analysis of children’s representation of coincidence, I found that children as young as 4 years old are sensitive to explanatory information, and readily use that information to reason about coincidental events. Notably, older children (and adults) spontaneously generated their own explanations for unusual observations when none were available. Next, in Chapter 3, I asked whether children’s spontaneous explanatory thinking empowers their further inferences. Using thinking about other people’s competence and performance as a case study, I found that preschool-aged children not only recognize explanatory information (both in terms of the explanations offered by the long-term characteristics of individual social agents, and by the short-term situational constraints individuals may face when performing a task), they also successfully used these explanations to support novel inferences about others’ competence and performance. In Chapter 4, I examined the very early foundations of explanation-seeking by studying infants. I found evidence that infants who see surprising events (here, in the social domain) appear to spontaneously seek explanations for those events. Furthermore, infants appear to consider multiple candidate explanations for a surprising observation. Collectively, the studies reported in this dissertation help to characterize the developmental origins of a foundational human capacity—the capacity to seek and represent explanations for our observations. This process, which is active from early in life, operates automatically, does not rely on linguistic competence, and supports inference and learning throughout development
Large-Scale Multi-Physics Topology Optimization with Consideration of Additive Manufacturing Constraints
No full textThis work focuses on creating topology optimization (TO) methods and algorithms for designing components that are fabricated by additive manufacturing and whose behavior is governed by multiple physics. The ability to design self-supporting parts without needing additional supports is beneficial and often necessary in additive manufacturing. One common method for achieving this is to enforce overhang constraints, but current techniques face challenges when scaling to three-dimensional, large-scale design problems. To overcome this limitation, we propose a novel approach that employs the Augmented Lagrangian method and proposes a novel disjunctive aggregation function. This new method introduces a rigorous solution to an existing issue in overhang constraint modeling, enhancing the design of self-supporting structures for additive manufacturing.
Additionally, we tackle the challenge of streamlining large-scale, multi-physics topology optimization in the open-source software OpenFOAM. The disconnect between academic work and industry applications can be attributed to the proof-of-concept nature of many academic studies and the complexity of real-world applications, which often involve multiple performance criteria from different physical simulations. This work integrates multi-physics simulations into a unified framework and provides seamless parallel algorithms to handle large-scale problems. This capability is essential for real-world applications that may require fine mesh resolutions, such as optimizing a heat sink that requires both fluid flow and heat transfer analysis. We leverage OpenFOAM’s parallelization capabilities and its wide range of tools to address these complex, large-scale challenges.
By proposing new methodologies to solve both the overhang constraint scalability and multi-physics TO challenges, this dissertation advances the applicability of topology optimization to large-scale problems, including to scales relevant for industrial use
Cache and Memory Optimized Data Structures for High Performance Applications
Get PDFAs modern data and compute infrastructure expands, designing data structures that maximize performance becomes increasingly crucial for enabling complex analyses on ever-larger datasets. There has been an increasing divergence between compute and memory performance on processors, also known as the processor-memory gap. This gap creates a conflict, especially with the highly dynamic nature of modern datasets. Maximizing memory performance often requires packing data tightly together, which, by its very nature, makes updating that data more difficult.
This thesis demonstrates that focusing on data layout can lead to the design of new data structures that overcome the traditional tradeoff between dynamism and memory locality, resulting in improved performance in both analytics and updates. The research encompasses several intersecting lines of work, presenting improvements to various data structures. These improvements either enhance the locality of existing data structures to increase their analytic capabilities or increase the updateability of tightly packed data structures.
The focus is on foundational data structures such as sets, graphs, and key-value stores, which are fundamental building blocks in other systems. Through various artifacts, this thesis illustrates how prioritizing locality can improve all aspects of data structures, including point queries, range queries, point updates, and batch updates. Additionally, the research highlights related improvements, such as size optimizations and methods to increase parallelism on modern large multicore machines. One of the primary data structures examined is the Packed Memory Array, which had mainly been used in theoretical contexts in prior work. Throughout this thesis, the Packed Memory Array, with the theoretical and practical improvements described, is shown to outperform best practices in many practical situations
Estimating Heterogeneity of Harm-Benefit Balance of Immune Checkpoint Inhibitors in Non-Small Cell Lung Cancer
Get PDFPrognosticators of who will benefit and who will be harmed by immune checkpoint inhibitors (ICIs) to treat non-small cell lung cancer (NSCLC) are poorly understood and lead to sub-optimal outcomes. Evidence gaps can be filled using real world data (RWD) from the broad population of ICI users, including high-risk patients; however, measurement error in adverse event detection must be overcome. The three research aims of this dissertation leveraged Johns Hopkins electronic healthcare records (EHR) and SEER-Medicare data to provide a foundation for improved measurement of ICI-related harms, as well as elucidation of important heterogeneity of treatment-related harms and benefits across NSCLC patient subgroups. In the first research aim, we validated five RWD-operable case definitions for immune-related adverse events (irAE) against clinician chart review in a registry of Johns Hopkins NSCLC patients treated with ICIs. We found several that were sufficiently valid for measurement of severe irAEs, which are of especially high clinical concern. In the second research aim, we assessed the severe irAE risk as well as the survival benefits of different ICI-containing regimens in a large sample of SEER-Medicare NSCLC patients to quantify the harm-benefit balance of adding chemotherapy to ICI therapy (ICI + chemo). We found evidence of co-occurring benefits and harms of added chemotherapy, including among high-risk subgroups such as patients with pre-existing autoimmune diseases. In the third research aim, we implemented a novel predictive framework for individual treatment effects to place patients into categories of harm-benefit tradeoffs. We found the harm-benefit balance of adding chemotherapy to ICI appears to favor younger patients with adenocarcinoma and fewer comorbidities who are diagnosed at a later stage, while patients with co-occurring harms and benefits are more likely to have baseline autoimmune diseases. Taken together, the aims address prevailing research gaps and demonstrate the utility of RWD to inform optimal use of ICIs. Our findings support the judicious use of ICI + chemo in high-risk patients, given the survival benefits. Predictive harm-benefit models can ensure risk-informed treatment choices that overlay with patient and provider priorities regarding quality of life versus length of life
Understanding mammalian cell metabolism and developing cell culture strategies for enhanced manufacturing of biotherapeutics
No full textMammalian cells, like Chinese hamster ovary (CHO) and human embryonic kidney (HEK293), are used as workhorses for biotherapeutics production owing to their ability to grow in large scale suspension cultures and produce high product titers with desirable product quality attributes (PQAs). However, bioprocessing faces challenges due to inefficient cell metabolism resulting in accumulation of toxic by-products, hindering cell proliferation, and causing cell death. Sub-optimal formulations of cell culture media (CCM), an expensive raw material, further reduces process yields. In this thesis, we studied the synergistic relationship between cell metabolism and CCM by employing bioanalytical tools to develop cell culture strategies for enhanced biotherapeutic production.
Firstly, we identified inhibitory metabolites (IMs) secreted by mammalian cells using LC-MS/MS based metabolomics pipeline. Eight IMs were found to accumulate in cell culture supernatants at levels detrimental to cell growth and protein synthesis. Pathway mapping of IMs revealed amino acids (AAs) as chief contributors toward IMs buildup. A design of experiment (DOE) guided statistical framework was developed to modify AA levels for bioprocess enhancements. Reduced AA levels in CCM lowered accumulation of IMs.
Next, due to low solubility and stability of traditional AAs, dipeptides were tested as alternative nutrients to problematic AAs like tyrosine and cysteine. Supplementation of dipeptides in CHO cultures supported biomass synthesis and protein production. 13C-Labeling experiments and kinetic modeling were performed to elucidate the utilization kinetics of dipeptides. We determined that dipeptides are cleaved both intracellularly and extracellularly and the cleavage rate depends on the structure, composition, and concentration of supplementation. Furthermore, Ala-Cys-Cys-Ala (ACCA) dipeptide dimer boosted growth and improved efficiency of glucose metabolism of CHO cells. High solubility of ACCA in basal medium simplified fed-batch processes by eliminating cysteine requirements from feed medium.
Lastly, induction of cytotoxicity in HEK293 cultures during transient recombinant adeno-associated virus (rAAV) production was characterized. Analysis of rAAV-producing cells revealed caspase-mediated apoptosis as a likely mechanism of cellular death. Inhibition of caspases using small molecule, Z-VAD.fmk, alleviated cell death and increased full to empty capsids ratio, a key PQA for rAAV vectors.
To sum up, cellular metabolism was investigated for CCM development to achieve superior biomanufacturing
Crafting Forests, Claiming Futures: Forest Sciences and the Politics of Anthropogenic Forests in South Korea
No full textThis dissertation examines the complex interplay between scientific expertise, forest ecologies, and environmental rule in postcolonial South Korea. Drawing on 36 months of ethnographic fieldwork in South Korea between 2017 and 2023, as well as extensive archival research, I analyze the transformation of scientific expertise and identities surrounding South Korea’s once notoriously devastated forests. Using the term “anthropogenic forests,” the dissertation points to reforested landscapes that embody both ecological and political histories of Korea. Through this framework, I make two overarching arguments. The first is that South Korea’s widely celebrated reforestation projects, often portrayed as triumphs of national resilience and scientific intervention after Japanese colonialism and the Korean War, are deeply entangled with occluded imperial legacies and their ongoing influences. The second is that Korean forests have become critical sites for experimenting with the ongoing negotiation between past violence/traumas, present challenges, and future horizons.
Each chapter develops several key conceptual contributions. First, I introduce the concept of “scientific emplotment” to elucidate how scientific narratives are woven into the everyday practices of forest scientists and the specific history of Korean forests. Second, I propose the notion of “climate sentinels” to describe the evolving role of forest experts in anticipating, registering, and interpreting radical environmental change. Lastly, I explore how contested “lifelines” in forest sciences reveal underlying tensions in Korea’s ecological horizons. Spanning four chapters, this research complicates the prevailing narrative of forest sciences as a nationalist machine. I demonstrate how the field’s struggles for self-determination and its capacity to address evolving socioecological challenges, such as climate change, are rooted in its historical contexts and everyday practices.
This dissertation bridges anthropology, environmental humanities, science and technology studies, and East Asian studies, with a focus on human-made forests in South Korea. It not only provides a nuanced understanding of South Korea’s forest sciences but also emphasizes the broader implications of how postcolonial societies confront their past to generate more-than-human responses to present and future ecological concerns. In doing so, this research illuminates the ways in which scientific expertise, forests, and ecological horizons are co-produced. It also contributes to the growing literature at the intersection of science, environment, state, and societal change in East Asia by offering a nuanced perspective on the evolving role of scientific expertise in mediating between violent legacies and ecological challenges
Deep learning-based 3d optical image reconstruction and surgical guidance
No full textImage-based surgical guidance leverages advanced 3D imaging technologies to assist surgeons during complex surgical procedures to enhance precision and safety. By integrating real-time 3D imaging techniques, such as optical coherence tomography (OCT) and structured light imaging, surgeons can visualize anatomical structures in greater detail, even in minimally invasive surgery such as robot-assisted laparoscopic procedures. Additionally, 3D image-based guidance enables continuous monitoring of the surgical field, reducing the risk of damage to surrounding tissues and improving patient outcomes.
This dissertation consists of developing two different 3D intraoperative imaging modalities. The first is a high-speed volumetric(3D) optical coherence tomography system. OCT allows high-resolution volumetric imaging of biological tissues in vivo. However, 3D-image acquisition and reconstruction can be time-consuming and often suffer from motion artifacts due to involuntary and physiological movements of the tissue, preventing accurate quantitative tissue tracking and assessment. The main purpose of this part of the dissertation is to develop a high-performance OCT system based on novel signal and image processing techniques that achieve real-time and motion-free in vivo OCT volumetric video rate imaging. Two methods using high-order regression and deep learning are developed and demonstrated through ex vivo, in vivo experiments. By designing additional decoders for the CNN, the convergence and accuracy of the reconstruction are significantly improved.
The second project is to develop a real-time single-shot fringe projection profilometry (FPP) system. FPP is being developed as a 3D vision system to plan and guide autonomous robotic intestinal suturing. Conventionally, sinusoidal patterns with multiple frequencies and phase shifts are needed to generate tissue point clouds, resulting in a slow frame rate. In this work, a deep learning-based single-shot FPP system and algorithm, and endoscope-based optical setup, which can reconstruct tissue point clouds with a single pattern using convolutions neural network is proposed and demonstrated. Depth reconstruction was performed using both binary and sinusoidal patterns, showing similar error on both synthesized and experimental data
CHARACTERIZING CELL LINE AND MEDIA PERFORMANCE OVER EXTENDED PERIODS THROUGH COMPUTATIONAL AND BIOINFORMATIC APPROACHES
No full textExtended durations are commonplace in large-scale biomanufacturing. Often cultures must be scaled up to volumes one million times larger than they started in to meet demand, which can take as long as three months. Similarly, concentrated media and feeds must often be made ahead of time for these cultures, and their components must remain in solution until ready for use. However, changes frequently occur over these long periods: cells may behave differently by the time they reach production scales, and concentrated media components may precipitate before being used. At best, these changes will reduce process efficiency; at worst, they could render a batch unusable. In this dissertation, I use a variety of computational and omics approaches to model and predict behavior over biomanufacturing timescales in myriad settings. First, I describe recent advances in engineering microbial consortia for improved bioprocessing and present an approach to improve the computational testing of combinatorial co-culture design. I then analyze ribosome profiling and transcriptomic data for young and old Chinese hamster ovary cells to identify translational changes and potential translational bottlenecks. Here, I quantify the translational efficiency of codons and identify novel RNA and polypeptide motifs that likely serve as translational bottlenecks. Next, I adapt a PHA-producing Methylotuvimicrobium alcaliphilum 20ZR cell line over 120 generations to tolerate the highly saline conditions found in electrocatalytic processes and assess the transcriptomic changes that enabled improved halotolerance. This assessment revealed that sulfate transporters and peptidoglycan biosynthesis are key gene categories involved in the adaptation to high sodium bicarbonate concentrations, and I show that the adapted microbe can efficiently grow on electrocatalytic effluent containing methanol reduced from CO2. Finally, I develop an ionic strength activity coefficient model to facilitate solubility predictions of cell culture media components and show that it accurately models the activity of individual ions in solution. Altogether, these results exemplify the utility of computational and bioinformatic approaches in characterizing cell cultures and media over biomanufacturing timescales