No description available.

No description available.

Indiana University-Purdue University Indianapolis (IUPUI) / Convolutional Neural Networks have come a long way since AlexNet. Each year the limits of the state of the art are being pushed to new levels. EfficientNet pushed the performance metrics to a new high and EfficientNetV2 even more so. Even so, architectures for mobile applications can benefit from improved accuracy and reduced model footprint. The classic Inverted Residual block has been the foundation upon which most mobile networks seek to improve. EfficientNet architecture is built using the same Inverted Residual block. In this thesis we experiment with Harmonious Bottlenecks in place of the Inverted Residuals to observe a reduction in the number of parameters and improvement in accuracy. The designed network is then deployed on the NXP i.MX 8M Mini board for Image classification. / 2023-10-11

Urban environments have been identified as a non-point source contributor of nutrient loadings into watersheds. Interannual surges of nutrient loadings into local water systems are more damaging than mean interannual nutrient loadings. Virginia has outlined the need to reduce urban nutrient loadings. Mean interannual nutrient loadings and interannual nutrient loadings variability are expected to increase under climate change (CC). However, there are few studies that provide a predictive framework for abating nutrient loadings under CC. Thus, there is a lack of information regarding how effective water quality policy will be in the future. Using the Difficult Run watershed in Fairfax County, VA, as a site of study, we used mathematical programming to compare how the costs of abating nutrient loads differed under differing climates in the Mid-Atlantic. We first compared the costs of abating mean interannual nutrient loadings in the watershed based on historical climate conditions to those predicted for CC. We then evaluated how changes in the interannual variability of nutrient loadings for CC affect the costs of meeting watershed goals. We found that abating mean interannual nutrient loadings was substantially costlier for CC relative to meeting the same goals under historical climate conditions. Further, we found that the costs of abating interannual nutrient loadings variability increased under CC relative to meeting the same goals under historical climate. One implication of this study suggests that policy makers seeking to meet water quality goals over time must front-load supplemental BMPs today in order to offset the changes predicted for CC. / Master of Science

Over the past ten years, there has been increased use of electronic power processing in alternative, sustainable, and distributed energy sources, as well as energy storage systems, transportation systems, and the power grid. Three-phase voltage source converters (VSCs) have become the converter of choice in many ac medium- and high-power applications due to their many advantages, such as high efficiency and fast response. For transportation applications, high power density is the key design target, since increasing power density can reduce fuel consumption and increase the total system efficiency. While power electronics devices have greatly improved the efficiency, overall performance and power density of power converters, using power electronic devices also introduces EMI issues to the system, which means filters are inevitable in those systems, and they make up a significant portion of the total system size and cost. Thus, designing for high power density for both power converters and passive components, especially filters, becomes the key issue for three-phase converters. This dissertation explores two different approaches to reducing the EMI filter size. One approach focuses on the EMI filters itself, including using advanced EMI filter structures to improve filter performance and modifying the EMI filter design method to avoid overdesign. The second approach focuses on reducing the EMI noise generated from the converter using a three-level and/or interleaving topology and changing the modulation and control methods to reduce the noise source and reduce the weight and size of the filters. This dissertation is divided into five chapters. Chapter 1 describes the motivations and objectives of this research. After an examination of the surveyed results from the literature, the challenges in this research area are addressed. Chapter 2 studies system-level EMI modeling and EMI filter design methods for voltage source converters. Filter-design-oriented EMI modeling methods are proposed to predict the EMI noise analytically. Based on these models, filter design procedures are improved to avoid overdesign using in-circuit attenuation (ICA) of the filters. The noise propagation path impedance is taken into consideration as part of a detailed discussion of the interaction between EMI filters, and the key design constraints of inductor implementation are presented. Based on the modeling, design and implementation methods, the impact of the switching frequency on EMI filter weight design is also examined. A two-level dc-fed motor drive system is used as an example, but the modeling and design methods can also be applied to other power converter systems. Chapter 3 presents the impact of the interleaving technique on reducing the system passive weight. Taking into consideration the system propagation path impedance, small-angle interleaving is studied, and an analytical calculation method is proposed to minimize the inductor value for interleaved systems. The design and integration of interphase inductors are also analyzed, and the analysis and design methods are verified on a 2 kW interleaved two-level (2L) motor drive system. Chapter 4 studies noise reduction techniques in multi-level converters. Nearest three space vector (NTSV) modulation, common-mode reduction (CMR) modulation, and common-mode elimination (CME) modulation are studied and compared in terms of EMI performance, neutral point voltage balancing, and semiconductor losses. In order to reduce the impact of dead time on CME modulation, the two solutions of improving CME modulation and compensating dead time are proposed. To verify the validity of the proposed methods for high-power applications, a 100 kW dc-fed motor drive system with EMI filters for both the AC and DC sides is designed, implemented and tested. This topology gains benefits from both interleaving and multilevel topologies, which can reduce the noise and filter size significantly. The trade-offs of system passive component design are discussed, and a detailed implementation method and real system full-power test results are presented to verify the validity of this study in higher-power converter systems. Finally, Chapter 5 summarizes the contributions of this dissertation and discusses some potential improvements for future work. / Ph. D.

<p dir="ltr">Learning, planning, and representing knowledge at multiple levels of temporal abstractions provide an agent with the ability to predict consequences of different courses of actions, which is essential for improving the performance of sequential decision making. However, discovering effective temporal abstractions, which the agent can use as skills, and adopting the constructed temporal abstractions for efficient policy learning can be challenging. Despite significant advancements in single-agent settings, temporal abstractions in multi-agent systems remains underexplored. This thesis addresses this research gap by introducing novel algorithms for discovering and employing temporal abstractions in both cooperative and competitive multi-agent environments. We first develop an unsupervised spectral-analysis-based discovery algorithm, aiming at finding temporal abstractions that can enhance the joint exploration of agents in complex, unknown environments for goal-achieving tasks. Subsequently, we propose a variational method that is applicable for a broader range of collaborative multi-agent tasks. This method unifies dynamic grouping and automatic multi-agent temporal abstraction discovery, and can be seamlessly integrated into the commonly-used multi-agent reinforcement learning algorithms. Further, for competitive multi-agent zero-sum games, we develop an algorithm based on Counterfactual Regret Minimization, which enables agents to form and utilize strategic abstractions akin to routine moves in chess during strategy learning, supported by solid theoretical and empirical analyses. Collectively, these contributions not only advance the understanding of multi-agent temporal abstractions but also present practical algorithms for intricate multi-agent challenges, including control, planning, and decision-making in complex scenarios.</p>

<p dir="ltr">Managing transient heat loads has become more challenging with the increasing electrification of ground, air, and marine vehicles. Doing so requires novel designs of thermal management systems, or in some cases, novel retrofits of legacy TMSs to accommodate the addition of more electrified subsystems. However, design tools that are well suited for examining and optimizing the dynamic response of TMS over candidate operation or mission profiles are limited. In this thesis, a principled methodology and associated tools for the enumeration and dynamic optimization of all feasible architectures of an air cycle machine are presented. Graph-based modeling is pivotal for exploring and optimizing ACM architectures, providing a structured representation of system components and interactions. By modeling the ACM as a graph, with vertices and edges representing components and interactions, respectively, various component configurations and performance metrics can be systematically analyzed. This approach enables efficient exploration of design alternatives and consideration of dynamic boundary conditions (representing, for example, a complex mission profile) during optimization. Another unique contribution of this thesis is a novel application of a multi-state graph-based modeling approach for developing dynamic models of turbomachinery components. By representing multiple states within each control volume or component and connecting them through power flows, this approach accurately captures both first and second law dynamics, enabling the computation of dynamic entropy generation rates. A detailed case study demonstrates the optimization of ACM architectures based on entropy generation minimization and dynamic bleed air flow rate minimization. This study highlights the trade-offs between different optimization criteria and the potential for generalizing the tool to more complex thermofluid systems in thermal management applications. The results underscore the importance of entropy-based analysis in comparing the thermodynamic losses across various system architectures.</p>

Studien syftar till att undersöka eventuell optimering av reservdelslagret som Stort AB hanterar åt sin kund. Det har genomförts genom att klassificera artiklarna i reservdelslagret enligt en ABC-klassificering baserat på antal lagerrörelser per artikel. Vidare skapades en optimeringsmodell som prioriterar artikelplacering utifrån ABC-klassificeringen. Optimeringsmodellen löstes med hjälp av linjärprogrammering i programvaran LINGO 20.0 med den objektiva funktionen att minimera avståndet artiklarna rör sig i reservdelslagret. Resultatet av optimeringen presenteras i en zonindelad skiss av reservdelslagret baserat på ABC-klassificeringen. För att bekräfta att resultatet från optimeringen faktiskt optimerade reservdelslagret så jämfördes det optimerade avståndet med nuvarande avstånd som artiklarna rör sig. Modellen är indelad i två höjdnivåer där nivå ett går att plocka för hand och nivå två kräver truckkörning. Det försvårar beräkningen för hur mycket modellen optimerade reservdelslagret eftersom det var svårt att jämföra tiden (avståndet) det tar att plocka för hand med tiden (avståndet) det tar att plocka med truck. Problemet löstes genom att göra ett intervall som varierar tiden (avståndet) det tar att plocka en artikel med truck. Studien resulterade i att reservdelslagret optimerades genom att avståndet minimerades i jämförelse med den befintliga artikelplaceringen. / The purpose of this study is to evaluate eventual optimization of the spare part warehouse that Stort AB manages for its customer. This has been carried out by classifying the products in the spare parts warehouse according to an ABC-classification based on the warehouse movement frequency of each product. Thereafter an optimization model was created that prioritizes the product placement according to the classification. The optimization model was solved by using linear programming in the software LINGO 20.0 with the objective function to minimize distance that the products move in the spare part warehouse. The result of the optimization is presented in a zoned sketch of the spare parts warehouse based on the ABC-classification. In order to verify that the optimization actually optimized the spare parts warehouse, the total optimized travel distance of the products was compared to the current total travel distance of the products. The optimization model is divided into two different height levels where a forklift is required for the highest level. This poses trouble for calculating the degree of optimization that the model provides since it is hard to compare the time (distance) it takes to pick a product by hand and the time (distance) it takes to pick a product with a forklift. A workaround for that problem is to calculate the degree of optimization on an interval that varies the time (distance) it takes to pick a product with a forklift. This study resulted in an optimization by reducing the distance traveled in the warehouse compared to the current product placement.

Med fokus på att öka lönsamheten för sina medlemmar strävar Norra Skog ständigt efter att förbättra och effektivisera sin logistik för att hantera det ständigt föränderliga flödet av massaved från skogsavlägg till industri. För att uppnå detta har företaget implementerat Woodflow, en avancerad programvara för logistikoptimering. I linje med det ökande intresset för att minska transportkostnader och främja hållbarhet inom transportnätverket, har det också börjat väcka intresse för att undersöka möjligheterna till att utöka användningen av tågtransporter för massaved. Detta examensarbete syftar till att analysera och optimera Norra Skogs transportflöde av massaved med målet att minimera transportkostnader och undersöka möjligheterna till att minska koldioxidutsläpp inom transportflödet. Arbetet inleddes med en grundlig litteraturstudie för att skapa en förståelse för leveranskedjan för massaved inom skogsindustrin samt ge en bakgrund till den optimeringsteori som ligger till grund för programvaran Woodflow.  Genom användning av Woodflow utfördes olika scenarioanalyser för att utvärdera olika strategier och metoder för att optimera transportflödet. Relevanta data för projektet samlades in och bearbetades för att förbereda indata till programvaran och möjliggöra analys av informationen om transportflöden för massaved. De undersökta scenarierna inkluderade ett scenario som speglade det faktiska transportflödet år 2023 och ett scenario med optimerade lastbilstransporter som användes som referenspunkt. Vidare genomfördes ett scenario med fria tåg- och lastbilstransporter med syftet att undersöka det optimala transportflödet som minimerar transportkostnader. Införande av nya terminaler och uteslutande av en befintlig terminal undersöktes i ett scenario och dessutom utfördes en analys av minimering av koldioxidutsläpp och transportkostnader genom att tillämpa en premiekostnad som representerade lastbilstransporternas högre koldioxidutsläpp. Till sist genomfördes ett framtidsscenario som syftar till att undersöka hur Norra Skog bör planera för sina transporter i framtiden utifrån prognoser av ökade tillgångsvolymer. Genom att noggrant utvärdera dessa olika scenarier söker arbetet efter de mest effektiva och hållbara lösningarna för Norra Skogs transportflöde för massaved. Resultatet visar på att det finns potential till stora kostnadsbesparingar och betydande minskningar av koldioxidutsläpp genom att optimera lastbilstransporterna från det faktiska utfallet 2023. Därefter finns ytterligare kostnadsbesparingar att hämta genom att optimera tågflödet och transportera mer volym via detta mer hållbara transportslag. Samtliga optimeringar av tågflödet, förutom framtidsscenariot med ökad tillgång, bidrar till ytterligare minskningar av koldioxidutsläpp och det anses därav vara av intresse för Norra Skog att fortsatt undersöka och utvärdera möjligheten att utöka tågtransporterna. Det är främst terminalen Storuman som visar på stor potential för fler antal tågavgångar, medan Hissmofors och Bastuträsk överlag visar minst potential för utökning av tågtransporter. Vidare visar terminalen Östavall ytterligare potential för ökad transport av massaved via tåg genom att helt ersätta terminalen Ånge. / Focusing on increasing profitability for its members, Norra Skog continuously strives to improve and streamline its logistics to manage the ever-changing flow of pulpwood from forest sites to industry. To achieve this, the company has implemented Woodflow, an advanced logistics optimization software. In line with the growing interest in reducing transportation costs and promoting sustainability within the transport network, there has also been increasing interest in exploring the possibilities of expanding the use of rail transport for pulpwood. This thesis aims to analyze and optimize Norra Skog's transport flow of pulpwood with the goal of minimizing transportation costs and exploring possibilities to reduce carbon dioxide emissions within the transport flow. The work began with a thorough literature review to create an understanding of the supply chain for pulpwood in the forestry industry and then provide a background to the optimization theory underlying the Woodflow software. Using Woodflow, various scenario analyses were conducted to evaluate different strategies and methods for optimizing the transport flow. Relevant data for the project was collected and processed to prepare input for the software and enable the analysis of information on pulpwood transport flows. The scenarios examined included one that reflected the actual transport flow in 2023 and one with optimized truck transports used as a reference point. Furthermore, a scenario with free rail and truck transports was conducted to investigate the optimal transport flow that minimizes transportation costs. The introduction of new terminals and the exclusion of an existing terminal were examined in one scenario, and an analysis of minimizing carbon dioxide emissions and transportation costs was also performed by applying a premium cost representing the higher carbon dioxide emissions of truck transports. Finally, a future scenario was conducted to examine how Norra Skog should plan its transports in the future based on forecasts of increased supply volumes. By carefully evaluating these different scenarios, the work seeks the most efficient and sustainable solutions for Norra Skog's pulpwood transport flow. The results show that there is potential for significant cost savings and substantial reductions in carbon dioxide emissions by optimizing truck transports from the actual outcome in 2023. Further cost savings can be achieved by optimizing the rail flow and transporting more volume with this more sustainable mode of transport. All optimizations of the rail flow, except for the future scenario with increased supply, contribute to further reductions in carbon dioxide emissions and are therefore considered of interest for Norra Skog to continue evaluating the possibility of expanding rail transports. It is mainly the terminal Storuman that shows great potential for an increased number of rail departures, while Hissmofors and Bastuträsk generally show the least potential for the expansion of rail transports. Furthermore, the Östavall terminal shows additional potential for increased transport of pulpwood by rail by completely replacing the Ånge terminal.

In many fields of science, engineering, and economics large amounts of data are stored and there is a need to analyze these data in order to extract information for various purposes. Data mining is a general concept involving different tools for performing this kind of analysis. The development of mathematical models and efficient algorithms is of key importance. In this thesis we discuss algorithms for the reduced rank regression problem and algorithms for the computation of the best multilinear rank approximation of tensors. The first two papers deal with the reduced rank regression problem, which is encountered in the field of state-space subspace system identification. More specifically the problem is \[ \min_{\rank(X) = k} \det (B - X A)(B - X A)\tp, \] where $A$ and $B$ are given matrices and we want to find $X$ under a certain rank condition that minimizes the determinant. This problem is not properly stated since it involves implicit assumptions on $A$ and $B$ so that $(B - X A)(B - X A)\tp$ is never singular. This deficiency of the determinant criterion is fixed by generalizing the minimization criterion to rank reduction and volume minimization of the objective matrix. The volume of a matrix is defined as the product of its nonzero singular values. We give an algorithm that solves the generalized problem and identify properties of the input and output signals causing a singular objective matrix. Classification problems occur in many applications. The task is to determine the label or class of an unknown object. The third paper concerns with classification of handwritten digits in the context of tensors or multidimensional data arrays. Tensor and multilinear algebra is an area that attracts more and more attention because of the multidimensional structure of the collected data in various applications. Two classification algorithms are given based on the higher order singular value decomposition (HOSVD). The main algorithm makes a data reduction using HOSVD of 98--99 \% prior the construction of the class models. The models are computed as a set of orthonormal bases spanning the dominant subspaces for the different classes. An unknown digit is expressed as a linear combination of the basis vectors. The resulting algorithm achieves 5\% in classification error with fairly low amount of computations. The remaining two papers discuss computational methods for the best multilinear rank approximation problem \[ \min_{\cB} \| \cA - \cB\| \] where $\cA$ is a given tensor and we seek the best low multilinear rank approximation tensor $\cB$. This is a generalization of the best low rank matrix approximation problem. It is well known that for matrices the solution is given by truncating the singular values in the singular value decomposition (SVD) of the matrix. But for tensors in general the truncated HOSVD does not give an optimal approximation. For example, a third order tensor $\cB \in \RR^{I \x J \x K}$ with rank$(\cB) = (r_1,r_2,r_3)$ can be written as the product \[ \cB = \tml{X,Y,Z}{\cC}, \qquad b_{ijk}=\sum_{\lambda,\mu,\nu} x_{i\lambda} y_{j\mu} z_{k\nu} c_{\lambda\mu\nu}, \] where $\cC \in \RR^{r_1 \x r_2 \x r_3}$ and $X \in \RR^{I \times r_1}$, $Y \in \RR^{J \times r_2}$, and $Z \in \RR^{K \times r_3}$ are matrices of full column rank. Since it is no restriction to assume that $X$, $Y$, and $Z$ have orthonormal columns and due to these constraints, the approximation problem can be considered as a nonlinear optimization problem defined on a product of Grassmann manifolds. We introduce novel techniques for multilinear algebraic manipulations enabling means for theoretical analysis and algorithmic implementation. These techniques are used to solve the approximation problem using Newton and Quasi-Newton methods specifically adapted to operate on products of Grassmann manifolds. The presented algorithms are suited for small, large and sparse problems and, when applied on difficult problems, they clearly outperform alternating least squares methods, which are standard in the field.