Thesis (MScEng)--Stellenbosch University, 2012. / ENGLISH ABSTRACT: The optimal design of a heat exchanger for a specified heat transfer, pressure drop and set of ambient conditions entails minimising space, weight, material usage and overall cost. However, the variables which influence the performance as well as the overall cost of a heat exchanger are not related in a simple way and it is not obvious which variables play the most important roles (Perry & Green, 1997:11-44). Air cooled heat exchangers (ACHEs) are normally designed in three stages, by different experts in the field, and with the aid of specially designed software. This project combines these thermal, mechanical and cost estimation processes into a MS Excel model which makes it easier to see the influence that design parameters have on the overall cost of the heat exchanger. A thermal model was created to design an API 661 (2006) ACHE. The results from this model compared well with those of HTRI Xchanger Suite 6.0 software, with HTRI being more conservative in its design mode. A mechanical design model was then developed, which uses as inputs the outputs of the thermal design. The output from this mechanical design model is the minimum material thicknesses based on the stress criteria of Appendix 13 of ASME VIII div 1 (2007) Boiler and Pressure Vessel Code. An experiment on a finned tube bundle was performed in a wind tunnel facility to determine performance characteristics and compare these to existing correlations in literature. The results showed that both the heat transfer coefficient (h) and loss coefficient (Eu) correlations proposed by Ganguli et al. (1985) closely predict the measured data, and were consequently used in the thermal design model. During this experiment it was also shown that the tube bundle reached 8 % - 9 % of its allowable internal fouling factor, due to rust build up inside the tubes, and in a testing period of only nine days. The thermal and mechanical models were then combined with a cost estimation process to perform both a thermal and mechanical parametric study. The thermal study showed that to obtain an optimal solution, the design must attempt to maximise the length, increase the width rather than the number of bays, make use of two bundles per bay with fewer but larger fans and employ a large number of tube rows with the least number of tube passes. These guidelines were used to create an initial design; Excel Solver was then applied to locate the optimum combination of bundle length and width that result in the minimum heat exchanger cost. Two mechanical considerations were investigated, both requiring additional welding and thus increased welding cost. Firstly the use of stay plates result in reduced required plate thicknesses according to the stress criteria since it provides additional stiffness in the header box design. Secondly the use of more (but smaller) nozzles as opposed to less (but larger) nozzles was also considered. The mechanical parametric study showed no specific trends, but both considerations should still be checked as it can be cost beneficial in a specific design. / AFRIKAANSE OPSOMMING: Optimale ontwerp van ’n warmteoordraer vir ’n gespesifiseerde warmteoordrag, drukval en stel van omgewingstoestande behels die minimalisering van ruimte, gewig, materiaalverbruik en algehele koste. Die veranderlikes wat egter die werkverrigting, sowel as die algehele koste, van ’n warmteoordraer beïnvloed, hou nie in ’n eenvoudige sin met mekaar verband nie, en dit is nie vanselfsprekend watter veranderlikes die belangrikste rolle speel nie (Perry & Green, 1997:11-44). Lugverkoelde warmteoordraers (air-cooled heat exchangers of ACHEs) word normaalweg in drie fases ontwerp deur verskillende kundiges in die veld en met behulp van spesiaal ontwerpte programmatuur. Hierdie studie kombineer dié termiese, meganiese en kosteberamingsprosesse in ’n MS Excel-model, wat dit makliker maak om van te stel wat die invloed wat ontwerpparameters op die algehele koste van die hitteruiler is. ’n Termiese model is geskep om ’n “API 661 (2006) ACHE” te ontwerp. Die resultate van hierdie model het goed vergelyk met dié van die HTRI Xchanger Suite 6.0-program, met HTRI meer konserwatiew in die ontwerp af. Na die termiese model geskep is, is ’n meganieseontwerp-model ontwikkel, wat as insette die uitsette van die termiese ontwerp gebruik het. Die uitset van hierdie meganieseontwerp-model is die minimum materiaaldikte gebaseer op die spanningskriteria van Bylae 13 van “ASME VIII div 1 (2007) Boiler and Pressure Vessel Code.” Daar is ’n eksperiment op ’n vinbuisbundel in ’n windtonnelfasiliteit uitgevoer om werkverrigtingskarakteristieke te bepaal en dit met bestaande korrelasies in die literatuur te vergelyk. Die resultate het getoon dat sowel die warmteoordragskoëffisiënt (h) en die verlieskoeffisient (Eu) korrelasies, voorgestel deur Ganguli et al. (1985), die data wat gemeet is akkuraat voorspel, en gevolglik is die korrelasies in die termieseontwerp-model gebruik. Gedurende die eksperiment is ook getoon dat die buisbundel 8 % - 9 % van sy toelaatbare interne-aanpakkingfaktor bereik het vanweë roesopbou binne-in die buise, en dit in ’n toetsingtydperk van slegs nege dae. Die termiese en meganiese modelle was toe gekombineer met ’n kosteberamingsproses om ’n termiese sowel as ’n meganiese parametriese studie uit te voer. Die termiese studie het getoon dat, om ’n optimale oplossing te verkry, die ontwerp moet poog om die lengte te maksimeer; die wydte eerder as die aantal strate (bays) te vermeerder; van twee bundels per straat gebruik te maak met minder, maar groter waaiers; en ’n groot aantal buisrye met die kleinste hoeveelheid buisdeurvloeiweë in te span. Hierdie riglyne is gebruik in ’n aanvanklike ontwerp, waarna die Excel Solver gebruik is om die optimale kombinasie van bundellengte en –wydte vas te stel met die oog op die laagste moontlike warmteoordraerkoste. Twee meganiese oorwegings is ondersoek wat albei addisionele sweiswerk sou vereis en dus tot verhoogde sweiskoste sou lei. Eerstens lei die gebruik van ankerplate (stay plates) tot ’n vermindering in die vereiste plaatdiktes volgens die spanningskriteria, aangesien dit addisionele stewigheid in die spruitstukhouerontwerp bied. Tweedens is die gebruik van meer (maar kleiner) spuitstukke teenoor minder (maar groter) spuitstukke ook oorweeg. Die meganiese parametriese studie het geen spesifieke voorkeurneigings getoon nie, maar altwee oorgewings moet nog getoets word want dit kan koste voordelig word in 'n spesifieke ontwerp.

Can momentum trading strategies beat Dutch or German stock market indices? If so, dothose strategies show significant positive net returns? For the period from March 2009 to March 2016this appears to be the case for only one out of the nine momentum trading strategies investigated withrespect to the Dutch stock market and for none of those same momentum trading strategiesinvestigated with respect to the German stock market. Furthermore, this research finds that the netmomentum returns seem to be winner- instead of loser-portfolio driven and that the longer the holdingperiod, the higher the net momentum returns realized.

No description available.

<p>Deep Neural Networks (DNNs) are constantly evolving, enabling the power of deep learning to be applied to an ever-growing range of applications, such as Natural Language Processing (NLP), recommendation systems, graph processing, etc. However, these emerging neural workloads present large computational demands for both training and inference. In this dissertation, we propose optimizations that take advantage of the unique characteristics of different emerging workloads to simultaneously improve accuracy and computational efficiency.</p> <p><br></p> <p>First, we consider Language Models (LMs) used in NLP. We observe that the design process of LMs (pre-train a foundation model, and subsequently fine-tune it for different downstream tasks) leads to models that are highly over-parameterized for the downstream tasks. We propose AxFormer, a systematic framework that applies accuracy-driven approximations to create accurate and efficient LMs for a given downstream task. AxFormer eliminates task-irrelevant knowledge, and helps the model focus only on the relevant parts of the input.</p> <p><br></p> <p>Second, we find that during fine-tuning of LMs, the presence of variable-length input sequences necessitates the use of padding tokens when batching sequences, leading to ineffectual computations. It is also well known that LMs over-fit to the small task-specific training datasets used during fine-tuning, despite the use of known regularization techniques. Based on these insights, we present TokenDrop + BucketSampler, a framework that synergistically combines a new regularizer that drops a random subset of insignificant words in each sequence in every epoch, and a length-aware batching method to simultaneously reduce padding and address the overfitting issue.</p> <p><br></p> <p>Next, we address the computational challenges of Transformers used for processing inputs of several important modalities, such as text, images, audio and videos. We present Input Compression with Positional Consistency (ICPC), a new data augmentation method that applies varying levels of compression to each training sample in every epoch, thereby simultaneously reducing over-fitting and improving training efficiency. ICPC also enables efficient variable-effort inference, where easy samples can be inferred at high compression levels, and vice-versa.</p> <p><br></p> <p>Finally, we focus on optimizing Graph Neural Networks (GNNs), which are commonly used for learning on non-Euclidean data. Few-shot learning with GNNs is an important challenge, since real-world graphical data is often sparsely labeled. Self-training, wherein the GNN is trained in stages by augmenting the training data with a subset of the unlabeled data and their pseudolabels, has emerged as a promising approach. However, self-training significantly increases the computational demands of training. We propose FASTRAIN-GNN, a framework for efficient and accurate self-training of GNNs with few labeled nodes. FASTRAIN-GNN optimizes the GNN architecture, training data, training parameters, and the graph topology during self-training.</p> <p><br></p> <p>At inference time, we find that ensemble GNNs are significantly more accurate and robust than single-model GNNs, but suffer from high latency and storage requirements. To address this challenge, we propose GNN Ensembles through Error Node Isolation (GEENI). The key concept in GEENI is to identify nodes that are likely to be incorrectly classified (error nodes) and suppress their outgoing messages, leading to simultaneous accuracy and efficiency improvements. </p> <p><br></p>

No description available.

[pt] Nesta dissertação é proposto um novo método para a avaliação da eficiência de uso do espectro a ser aplicado para os sistemas de telecomunicações pontoárea, incluindo radiodifusão, e ponto-a-ponto. A definição proposta para o índice de eficiência é baseada nos aspectos técnicos, mas outros aspectos que podem influenciar na determinação da eficiência dos sistemas de telecomunicações, como os econômicos, sociais, políticos e estratégicos também são discutidos. O método é testado com dados reais de sistemas disponibilizados pela Anatel e no sítio dos prestadores de serviços de telecomunicações no Brasil. É muito importante avaliar a eficiência dos sistemas de telecomunicações, pois sendo o espectro de radiofreqüências um recurso limitado, pela tecnologia e capacidade de gerenciamento, a introdução de novos serviços de telecomunicações somente será possível com a substituição, pela realocação ou extinção, de determinadas aplicações menos eficientes. Apesar da regulamentação nacional e internacional constantemente referir-se à importância do uso eficiente das faixas de freqüências, não existe um método harmonizado de avaliação, o que impede a verificação deste aspecto pelos órgãos reguladores. As novas tecnologias e as oportunidades de desenvolvimento social que elas proporcionam demonstram a necessidade da boa gerência do espectro de radiofreqüências e de que cada uma de suas porções seja utilizada de forma eficiente e adequada a cada tipo de aplicação. / [en] In this work a new method for the assessment of the spectrum use efficiency is proposed to be applied to point-area, including broadcasting, and point-to-point telecommunications systems. The proposed definition of the efficiency index is based in technical aspects, but other subjects that affect the determination of a telecommunications system efficiency, as economics, social, political and strategic issues are also discussed. The method is tested with data of real systems available from the Brazilian Regulatory Agency (Anatel) and at the telecommunications service providers web sites. It is very important to assess the telecommunications systems efficiency, because being the radio spectrum a scarce resource, limited by technology and management capacity, the introduction of new telecommunications services will only be possible by replacement, refarming or suppression of some less efficient applications. Despite national and international regulations usually refers to the importance of efficient use of the frequency bands, there is no harmonized assessment method to be applied and verified by the regulators. The new technologies and the opportunities for social development provided by them, bring to the light the need for a good radio spectrum management and how each part of it must be used efficiently and assigned properly for each type of application or telecommunications service.

International Telemetering Conference Proceedings / October 23-26, 2000 / Town & Country Hotel and Conference Center, San Diego, California / Advances in spectrally efficient Feher’s Quadrature Phase Shift Keying (FQPSK) and Feher‘s Quadrature Amplitude Modulation (FQAM) patented technologies, commercial and government “dual-use” FQPSK products and Test & Evaluation (T&E) results are highlighted in this overview paper. US and international customer requirements/systems, programs and recent deployments and standardization programs are also described. FQPSK doubles the spectral efficiency of PCM/FM Telemetry and of Feher patented Gaussian Minimum Shift Keying (GMSK or FGMSK) and of alternatives, while FQAM quadruples the spectral efficiency of these systems. The predominant focus of this presentation is on an overview and advances of IRIG 106-00 standardized FQPSK technologies. FQAM systems will also be described. Use of FQPSK for applications such as telemetry, data links, clear mode, TDMA ,CSMA and CDMA, OCDMA, WCDMA as well as OFDM – COFDM will be included in the presentation.

The goal of this work is to present a conjugate duality treatment of composed programming as well as to give an overview of some recent developments in both scalar and multiobjective optimization. In order to do this, first we study a single-objective optimization problem, in which the objective function as well as the constraints are given by composed functions. By means of the conjugacy approach based on the perturbation theory, we provide different kinds of dual problems to it and examine the relations between the optimal objective values of the duals. Given some additional assumptions, we verify the equality between the optimal objective values of the duals and strong duality between the primal and the dual problems, respectively. Having proved the strong duality, we derive the optimality conditions for each of these duals. As special cases of the original problem, we study the duality for the classical optimization problem with inequality constraints and the optimization problem without constraints. The second part of this work is devoted to location analysis. Considering first the location model with monotonic gauges, it turns out that the same conjugate duality principle can be used also for solving this kind of problems. Taking in the objective function instead of the monotonic gauges several norms, investigations concerning duality for different location problems are made. We finish our investigations with the study of composed multiobjective optimization problems. In doing like this, first we scalarize this problem and study the scalarized one by using the conjugacy approach developed before. The optimality conditions which we obtain in this case allow us to construct a multiobjective dual problem to the primal one. Additionally the weak and strong duality are proved. In conclusion, some special cases of the composed multiobjective optimization problem are considered. Once the general problem has been treated, particularizing the results, we construct a multiobjective dual for each of them and verify the weak and strong dualities. / In dieser Arbeit wird, anhand der sogenannten konjugierten Dualitätstheorie, ein allgemeines Dualitätsverfahren für die Untersuchung verschiedener Optimierungsaufgaben dargestellt. Um dieses Ziel zu erreichen wird zuerst eine allgemeine Optimierungsaufgabe betrachtet, wobei sowohl die Zielfunktion als auch die Nebenbedingungen zusammengesetzte Funktionen sind. Mit Hilfe der konjugierten Dualitätstheorie, die auf der sogenannten Störungstheorie basiert, werden für die primale Aufgabe drei verschiedene duale Aufgaben konstruiert und weiterhin die Beziehungen zwischen deren optimalen Zielfunktionswerten untersucht. Unter geeigneten Konvexitäts- und Monotonievoraussetzungen wird die Gleichheit dieser optimalen Zielfunktionswerte und zusätzlich die Existenz der starken Dualität zwischen der primalen und den entsprechenden dualen Aufgaben bewiesen. In Zusammenhang mit der starken Dualität werden Optimalitätsbedingungen hergeleitet. Die Ergebnisse werden abgerundet durch die Betrachtung zweier Spezialfälle, nämlich die klassische restringierte bzw. unrestringierte Optimierungsaufgabe, für welche sich die aus der Literatur bekannten Dualitätsergebnisse ergeben. Der zweite Teil der Arbeit ist der Dualität bei Standortproblemen gewidmet. Dazu wird ein sehr allgemeines Standortproblem mit konvexer zusammengesetzter Zielfunktion in Form eines Gauges formuliert, für das die entsprechenden Dualitätsaussagen abgeleitet werden. Als Spezialfälle werden Optimierungsaufgaben mit monotonen Normen betrachtet. Insbesondere lassen sich Dualitätsaussagen und Optimalitätsbedingungen für das klassische Weber und Minmax Standortproblem mit Gauges als Zielfunktion herleiten. Das letzte Kapitel verallgemeinert die Dualitätsaussagen, die im zweiten Kapitel erhalten wurden, auf multikriterielle Optimierungsprobleme. Mit Hilfe geeigneter Skalarisierungen betrachten wir zuerst ein zu der multikriteriellen Optimierungsaufgabe zugeordnetes skalares Problem. Anhand der in diesem Fall erhaltenen Optimalitätsbedingungen formulieren wir das multikriterielle Dualproblem. Weiterhin beweisen wir die schwache und, unter bestimmten Annahmen, die starke Dualität. Durch Spezialisierung der Zielfunktionen bzw. Nebenbedingungen resultieren die klassischen konvexen Mehrzielprobleme mit Ungleichungs- und Mengenrestriktionen. Als weitere Anwendungen werden vektorielle Standortprobleme betrachtet, zu denen wir entsprechende duale Aufgaben formulieren.

Baigiamajame magistro darbe yra nagrinėjama aktyvaus tarptautiniu mastu diversifikuoto portfelio sudarymo ir valdymo problematika. Techninė pažanga ir investicinių priemonių įvairovė leidžia investuotojams investuoti į įvairias pasaulio rinkas kuriant tarptautinius ar globalius investicijų portfelius. Tai suteikia investuotojams galimybę kovoti su tokiais išsivysčiusių kapitalo rinkų iššūkiais kaip šalies sisteminės rizikos padidėjimas dėl įvairių verslo šakų tarpusavio integracijos, nepakankamais kai kurių rinkų augimo tempais, tačiau iškelia ir naujų iššūkių. Pagrindinis jų yra efektyvaus portfelio sudarymo menko informuotumo, didelio neapibrėžtumo ir rinkų nepastovumo sąlygomis iššūkis. Šiame darbe yra nuosekliai nagrinėjami investicijų portfelių sudarymo metodai ir modeliai įvertinant jų taikymo galimybes darbe analizuojamai problemai spręsti. Remiantis atlikta analize, parenkamas modelis geriausiai tinkantis tolesnei plėtotei, vertinami pagrindiniai mokslininkų siūlomi jo papildymai ir taikymo interpretacijos nuosekliai pereinant prie pirmos tikslo dalies įgyvendinimo – aktyviam tarptautiniu mastu diversifikuoto investicijų portfelio valdymui skirto teorinio – hipotetinio modelio pasiūlymo. Atlikus tarptautinės diversifikacijos privalumų ir trūkumų analizę, pagrindžiamas jos taikymo būtinumas ir nauda. Atlikta pagrindinių tarptautinės diversifikacijos vykdymo būdų apžvalga leidžia parinkti, autoriaus nuomone, geriausią šiuo metu Lietuvos investuotojams prieinamą –... [toliau žr. visą tekstą] / The master thesis examines problems of the active internationally diversified investment portfolio selection and management. Technical progress and variety of investment objects (financial assets) allow investors to invest in different world markets developing international or global investment portfolios. It gives opportunities to deal with such challenges of developed capital markets as high systematic risk and limited growth of some markets, but as a result investor must accept some new challenges. The most important of them is an efficient portfolio selection in terms of low information level, high uncertainty and market volatility level. The models of investment portfolio selection are analyzed evaluating opportunities to use them solving in this thesis examined problem. On the basis of the conducted research a model which is most suitable for further development is chosen, the main supplementary supplements and interpretations of applications are analyzed, leading to the implementation of the first part of the set goal which is the suggestion of theoretical–hypothetical model for active management of an internationally diversified investment portfolio. The careful consideration of the advantages and disadvantages of international diversification proves the necessity and benefits of its application. Main methods’ of international diversification analysis gives opportunity for the thesis author to chose the best investment’s method possible for Lithuanian investors – the... [to full text]

Magister Legum - LLM