62588 research outputs found
Sort by
Assessing the Efficacy of Process-Specific Topology Optimization for Direct-Ink Write 3D-Printed Hierarchical Composites and Structures
No full textPolymers are ubiquitous in modern society. From food packaging to structural components inaerospace applications, plastics can be found. For applications like the latter, there is a requirement for exceptional performance characteristics - whether that be with respect to mechanical stiffness or strength, thermal or electrical conductivity, or other desirable engineering outcomes. As a result, much effort has been exerted in industry and academia toward optimizing the performance of structural components made out of polymers. Performance optimization of structural components, generally speaking, is a multi-layered problem. One layer of this problem is the material design problem – tailoring the properties of the material (by changing/adding to the manufacturing process, leveraging composites, etc.) in the desired component to have certain characteristics. Another layer is the structural design problem – the geometric features of the component (such as shape and topology) that govern its performance in service, such as under mechanical or thermal loading. For polymers, additive manufacturing (AM) is a tool which is very amenable to tailoring, both in terms of material behavior and structural geometry. Its use, in conjunction with structural optimization techniques (e.g. topology optimization, or TO) is the subject of this thesis. First, numerical and experimental benchmarking is presented on minimum mechanical compliance optimization and material extrusion AM techniques (fused filament fabrication, or FFF, and direct-ink writing, or DIW). A major focus of the benchmarking work is the analysis of the impact of mechanical anisotropy and material extrusion orientation on the outcomes of minimum compliance TO designs. Finite element calculations are performed using Matlab scripts and commercial software (ABAQUS) to complement flexural testing of AM specimens made of ABS polymer or epoxy-based polymer matrix composite inks. Next, analysis of multi-material topology optimization (MMTO) and multi-material additive manufacturing (MMAM) is presented. An analysis of choices regarding extrusion of material at the interfaces between materials is presented with corresponding experimental testing (e.g., printing and flexural testing) of specimens with disparate interface designs. Finally, numerical work towards the multi-material thermomechanical optimization of structures with orthotropic material behavior is presented. The main contributions of the thesis work lie in: (i) establishing baseline numerical and experimental evaluations of TO/AM structures using standard TO and AM methods (e.g. SIMP, material-extrusion AM with common infill patterns) (ii) quantifying the impact of process-specific design methodologies for the orientation design of orthotropic composites (iii) presenting a new application of optimization in multiphysics and multi-material TO for isotropic/orthotropic materials that considers elastic compliance and thermal conductance and (iv) initial experimental assessments of printing methodologies for their integrity at material interfaces in multi-material TO structures
On the Inelastic Response of Materials, Structures, and Composite Systems: Numerical & Experimental Studies in Cyclic Plasticity and Hypoplasticity
No full textThis dissertation advances inelastic design across several case studies involving elastoplastic shakedown in metallic components and hypoplasticity in composite systems with soil-structure interfaces. For metallic components that are subject to repeated and/or extreme thermomechanical loading conditions, conventional design, in applications that are not limited by high-cycle fatigue, employs first-yield criteria in order to avoid failures due to cyclic plasticity. However, yield-limited designs, which limit the design space to purely elastic behaviors, often fail to produce acceptable solutions for materials and structures in extreme environments. In contrast, inelastic design offers potential enhanced lifing, allowable loading levels, and light-weighting opportunities compared to traditional first-yield approaches. Across many industries, the deployment of these types of inelastic analyses (e.g., shakedown) are underutilized due, in part, to a lack of awareness and experimentally validated modeling and design guides.In this dissertation, several metals-based case studies that are relevant to aerospace and civil engineering applications are presented: thin-walled tubular structures and thin auxetic sheets. Experimentally, new demonstrations of shakedown behavior and shakedown design (avoiding alternating plasticity and/or ratchetting) at elevated and ambient temperatures are made for a next-generation single-crystal superalloy that is currently under development (DD6) and a common engineering aluminum alloy (AA5083-O). In concert, material characterization and experiments are used to validate numerical finite element analysis for new shakedown design guides under a range of axial and multiaxial (thermo)mechanical loading conditions. The shakedown case-study results serve to promote more wide-spread adoption of inelastic analysis in realizing new structural concepts and accurately assessing the structural integrity of existing components. Finally, a complementary study of composite systems (at soil-structure interfaces) is presented where potential numerical frameworks are developed based on existing experimental work to support the inelastic design of thick-walled suction elements in off-shore wind applications
Investigation Of The Effect Of Injection Pressure On Cavity Filling Of Hot Runner System
No full textThis research aims to examine the impact of injection pressure on the process of cavity filling in a hot runner injection molding system, under varying temperature conditions. The research utilizes a combination of analytical and experimental approaches to simulate and evaluate the performance of the hot runner system across various injection pressures and temperatures. The aim of this study is to investigate the impact of injection pressure on the process of cavity filling and to determine the most suitable range of injection pressure that can ensure complete cavity filling and optimal quality of the product. The results of the numerical investigation indicate that injection pressure is a critical parameter that affects the cavity filling process in hot runner injection molding systems. The experiments conducted at different temperatures show that injection pressure has a significant effect on cavity filling, with increasing injection pressure leading to better cavity filling. The optimal injection pressure range for complete cavity filling varies depending on the melting temperature. Incomplete filling occurs at lower injection pressures due to the increased material viscosity caused by the lower melting temperature. The findings of this study can be useful to manufacturers seeking to optimize their hot runner injection molding process for enhanced product quality. Future work includes investigating the effectiveness of Rheodrop technology in cavity filling under low injection pressure
Interpretability of Graphical Models
No full textThe success of machine learning has outpaced our ability to understand these models, and explainable machine learning (XML) has emerged as a research field to bridge this gap. Specifically, we focus on explaining inference algorithms and neural networks relevant to graphs, including Belief Propagation (BP) and Graph Neural Networks (GNN).The dissertation has three main contributions. First, we propose two distinct methods to explain BP. We first formulate the problem of explaining BP by finding faithful, concise, and diverse subgraphs as a constrained cross-entropy minimization framework, and solve it by a beam search algorithm. Alternatively, we introduce Shapley values to fairly evaluate the attribution of a BP inference outcome and propose an algorithm that leverages the structure of graphs, the decomposability of Shapley values, and the iterative nature of BP inference to overcome the exponential time complexity of calculating Shapley values. Second, we explore the connection between generated explanations and human perception of GNNs. We investigate the joint impacts of two widely used metrics, namely simulatability and counterfactual relevance, by conducting a survey, and develop a multi-objective optimization algorithm that finds Pareto optimal explanations balancing simulatability and counterfactual relevance. In another work, we introduce a novel concept, "explanation thickness," to evaluate the robustness of salient feature rankings, which better aligns with human cognition. Lastly, we study the robustness of gradient-based explanations against adversarial attacks and propose an effective algorithm based on tractable surrogate bounds. In another work, we aim to find robust and sensitive explanations for Siamese Networks using a constrained optimization framework that utilizes self-learning
John Mitchel: Transnational History and Politics, 1854-1873
No full textJohn Mitchel (1815-1875), was a legendary figure of Irish nationalism. Famous, if not infamous, for his advocacy of political violence and nationalist revolution in Ireland, as well as his support for slavery and the slave-trade in the United States, he made significant contributions to the transnational history of the nineteenth-century Anglosphere. The present work focuses on Mitchel\u27s transatlantic work as a journalist, historian, and ideologue. Before, during, and after the American Civil War, Mitchel was consistently and vociferously engaged with the problems of American culture and politics. He analyzed and interpreted these dynamics according to a strong sense of the relationship of small nations to empires, and with an abiding suspicion of coerced political unions. He edited important and widely circulated journals in the antebellum period in both New York City and Knoxville, Tennessee, respectively. In 1862 he became an editor at the Richmond Daily Enquirer, a paper regarded as the most important and influential organ in the Confederacy. When he fell into opposition to the war policy of Jefferson Davis, he moved to the more hostile, but no less influential, Richmond Daily Examiner. At the end of the American Civil War, Mitchel was arrested in New York City for his seditious writings at the New York Daily News. In 1867 he commenced his final weekly journal. That paper, published out of New York City until 1872, was the Irish Citizen. Like all of Mitchel\u27s journalistic efforts it effortlessly combined reviews and mentions of literature with a strong Irish nationalist viewpoint on current events. It is my contention, in the present work, that Mitchel\u27s journalistic and historical writings constitute a vastly informative, and often overlooked, base of source material for analyzing and clarifying the culture and politics of the mid nineteenth-century Anglosphere, particularly in the United States. Mitchel\u27s views, and their wide and occasionally controversial reception by the public, do much to contribute to our understanding of the American antebellum, Civil War, and Reconstruction periods
Spatiotemporal Information Processing in Living Neuronal Network In Vitro
No full textOur brain can differentiate many objects and time-dependent changes in the object features. This functionality requires information processing in both spatial and temporal domains. Studies showed that cortical microcircuits are intrinsically responsive to spatiotemporal patterns. Confined neural networks in vitro obtained from the brain cortex with an optical interface were used to classify spatiotemporal information in this work. This experimental setup was designed by confining a cortical network in a polydimethylsiloxane (PDMS) well and by co-transfecting the neurons with channel-rhodopsin 2 (ChR2) and jRgeco1a. In line with other studies, our data showed that neurons in dissociated cortical cultures are very susceptible to synchronized bursts, particularly in the confined network due to synaptic scaling. As these bursts represent chaotic activation and do not carry information about the input, we first designed a stimulation protocol to suppress bursts with distributed patterns of optical activation in the network. We found that increases in the Ca2+ baseline level of all neurons were correlated with successful burst suppression and the success rate was higher when the culture medium was replaced by artificial cerebrospinal fluid (ACSF) with high [Mg2+]. In the absence of population bursts, our results showed that the state of the living network could be used to classify input spatial and temporal patterns. Our experimental model shows a paradigm of reservoir computing with a living neural network that could run learning applications with small sets of training data and computational resources. This model can further be used to study toxicity or influences of commercial drugs on spatiotemporal information in neural networks
Juniper Shouldn\u27t Be Here: A YA Horror Novel Excerpt, A Rhododendron, Another Question About Genre in the Academy
No full textThis project takes up the legacies of white supremacist movements in the Inland Northwest alongside the shadow Lovecraft casts over cosmic horror and the anxieties surrounding racial purity that sit at the heart of so many iterations of the monster. I utilize critical horror studies, archival theory, and creative writing methodologies to compose a Young Adult horror novel that addresses intergenerational violence, the process of grieving those who have hurt us, and the paths we face if we attempt to break cyclical patterns within our families, our communities, and our generic traditions
Toward Cultural Bridges and Toppled Hierarchy: Geography and Racial Economy in Gloria Naylor\u27s Sapphira Wade
No full textThis paper considers Gloria Naylor\u27s unpublished, unfinished manuscript Sapphira Wade as a Black feminist reimagining of traditional geographies of domination. I examine in the narrative the ways in which different communities\u27 understandings of the landscape shape their relationships with each other, and then I turn to investigate why some communities construct hierarchical social orders: in order to exploit the land and people around them. In addition to outlining the dynamics of group formation in the narrative, I also trace characters Bascombe and Mari\u27s resistance to domination and posit that they offer a productive alternative to a white identity structured by conquest. I argue that Sapphira Wade constitutes a new "poetics of landscape," by centering Naylor\u27s Black feminist subversion of Norwegian traditional geography, featuring the Sámi Indigenous groups\u27 philosophy, and including ephemeral glimpses of eponymous Sapphira, the largely absent Senegalese woman whom we never directly meet in Naylor\u27s partial draft
Shrinking and Expanding Self-similar Solutions to the Ricci Flow and the Mean Curvature Flow
No full textIn studying geometric flows, particularly the Ricci flow and the mean curvature flow, singularities generally occur in finite time. Understanding singularities\u27 structure is crucial in the theory of geometric flows and their applications. Since self-similar solutions to geometric flows often arise as singularity models, they play a significant role in the study of the formation of singularities in geometric flows.In this thesis, we attempt to study the classification and geometry of self-similar solutions to the Ricci flow and the mean curvature flow. Firstly, joint with Huai-Dong Cao, we investigate the geometry of 4-dimensional complete gradient shrinking Ricci solitons with half positive isotropic curvature (half PIC) or half nonnegative isotropic curvature. We prove a certain form of curvature estimates for such Ricci shrinkers, including a quadratic curvature lower bound estimate for noncompact ones with half PIC. As a consequence, we obtain a new and more direct proof of the classification result, first observed by Li-Ni-Wang [92], for gradient shrinking K\"ahler-Ricci solitons of complex dimension two with nonnegative isotropic curvature. Moreover, based on a strong maximum principle argument, we classify 4-dimensional complete gradient shrinking Ricci solitons with half nonnegative isotropic curvature, except the half PIC case. We also treat the half PIC case under an additional assumption that the Ricci tensor has an eigenvalue with multiplicity 3.Secondly, joint with Huai-Dong Cao and Tianbo Liu, for dimension , we study the curvature estimates on complete gradient expanding Ricci solitons with nonnegative Ricci curvature. We show that if the asymptotic scalar curvature ratio is finite, then the Riemann curvature tensor must have at least sub-quadratic decay.Lastly, joint with Jiangtao Yu, we study the convexity of 2-convex expanding solitons to the mean curvature flow in . Under the assumption of being asymptotic to (strictly) mean convex cones, we show that for any -dimensional complete self-expander, 2-convex implies convex.
Inventory Management for Supply Chains with Uncertainty
No full textInventory management is a challenging task due to the presence of uncertainty, such asdemand uncertainty and supply uncertainty. It is necessary for companies to achieve a balance between meeting customer demand and minimizing holding costs. This dissertation explores the impact of uncertainty on supply chain systems, and presents simple heuristics to optimize inventory order up-to-levels, enabling efficient management of the system. In our first topic, we consider a single-stage system with demand uncertainty and a minimum order quantity (MOQ) requirement. Given the challenges associated with implementing complex optimal ordering policies in practice, we propose some easily implementable heuristic ordering policies based on well-known base-stock or (s, S) policies. We propose algorithms to determine optimal ordering parameters in the heuristic ordering policies, and test the performance of our heuristics by numerical experiments. Additionally, a sensitivity analysis is conducted to assess the influence of the MOQ requirement on the system. In our second topic, we study a periodic-review, one-warehouse multiple-retailer (OWMR) inventory system with both demand uncertainty and supply disruptions. We assume that each stage follows a base-stock policy and that the allocation policy is first-come, first-served (FCFS). We provide an explicit cost function using a �top-down� approach and optimize the system by the projection method. We also propose two heuristics for this problem. One heuristic combines a heuristic for serial systems subject to supply disruptions with the decomposition-aggregation (DA) heuristic for distribution systems without disruptions. The other heuristic combines DA and the newsvendor problem with disruptions. In the last topic of this dissertation, we consider a distribution system with demand uncertainty and supply disruptions, which is an extension of the research conducted in the second topic. We derive the cost function of the system by analyzing the inventory dynamics. In addition, we propose three heuristics to achieve near-optimal base-stock levels. Two of these heuristics are modified versions derived from the heuristics discussed in the previous chapter, while the third heuristic is a hybrid that combines the first two heuristics