When used in a wireless mesh network, TCP has shortcomings in the areas of throughput and fairness among traffic flows. Several methods have been proposed to deal with TCP's weakness in a wireless mesh, but most have been evaluated with simulations rather than experimentally. We evaluate several major enhancements to TCP – pacing, conservative windows, and delayed ACKs – to determine whether they improve performance or fairness in a mesh network operating in the BYU Computer Science building. We also draw conclusions about the effectiveness of wireless network simulators based on the accuracy of reported simulation results.

Um sistema de aprendizado supervisionado é um programa capaz de realizar decisões baseado na experiência contida em casos resolvidos com sucesso. As regras de classificação induzidas por um sistema de aprendizado podem ser analisadas segundo dois critérios: a complexidade dessas regras e o erro de classificação sobre um conjunto independente de exemplos. Sistemas de aprendizado têm sido desenvolvidos na prática utilizando diferentes paradigmas incluindo estatística, redes neurais, bem como sistemas de aprendizado simbólico proposicionais e relacionais. Diversos métodos de aprendizado podem ser aplicados à mesma amostra de dados e alguns deles podem desempenhar melhor que outros. Para uma dada aplicação, não existem garantias que qualquer um desses métodos é necessariamente o melhor. Em outras palavras, não existe uma análise matemática que possa determinar se um algoritmo de aprendizado irá desempenhar melhor que outro. Desta forma, estudos experimentais são necessários. Neste trabalho nos concentramos em uma tarefa de aprendizado conhecida como classificação ou predição, na qual o problema consiste na construção de um procedimento de classificação a partir de um conjunto de casos no qual as classes verdadeiras são conhecidas, chamado de aprendizado supervisionado. O maior objetivo de um classificador é ser capaz de predizer com sucesso a respeito de novos casos. A performance de um classificador é medida em termos da taxa de erro. Técnicas experimentais para estimar a taxa de erro verdadeira não somente provêem uma base para comparar objetivamente as performances de diversos algoritmos de aprendizado no mesmo conjunto de exemplos, mas também podem ser uma ferramenta poderosa para projetar um classificador. As técnicas para estimar a taxa de erro são baseadas na teoria estatística de resampling. Um ambiente chamado AMPSAM foi implementado para ajudar na aplicação dos métodos de resampling em conjuntos de exemplos do mundo real. AMPSAM foi projetado como uma coleção de programas independentes, os quais podem interagir entre si através de scripts pré-definidos ou de novos scripts criados pelo usuário. O ambiente utiliza um formato padrão para arquivos de exemplos o qual é independente da sintaxe de qualquer algoritmo. AMPSAM também inclui ferramentas para particionar conjuntos de exemplos em conjuntos de treinamento e teste utilizando diferentes métodos de resampling. Além do método holdout, que é o estimador de taxa de erro mais comum, AMPSAM suporta os métodos n-fold cross-validation --- incluindo o leaning-one-out --- e o método bootstrap. As matrizes de confusão produzidas em cada iteração utilizando conjuntos de treinamento e teste podem ser fornecidas a um outro sistema implementado chamado SMEC. Este sistema calcula e mostra graficamente algumas das medidas descritivas mais importantes relacionadas com tendência central e dispersão dos dados. Este trabalho também relata os resultados experimentais a respeito de medidas do erro de classificação de três classificadores proposicionais e relacionais bem conhecidos, utilizando ambos os sistemas implementados, em diversos conjuntos de exemplos freqüentemente utilizados em pesquisas de Aprendizado de Máquina. / A learning system is a computer program that makes decisions based on the accumulative experience contained in successfully solved cases. The classification rules induced by a learning system are judged by two criteria: their classification error on an independent test set and their complexity. Practical learning systems have been developed using different paradigms including statistics, neural nets, as well as propositional and relational symbolic machine learning. Several learning methods can be applied to the same sample data and some of them may do better than others. Still, for a given application, there is no guarantee that any of these methods will work or that any single method is necessarily the best one. In other words, there is not a mathematical analysis method that can determine whether a learning system algorithm will work well. Hence, experimental studies are required. In this work we confine our attention to the learning task known as classification or prediction, where the problem concerns the construction of a classification procedure from a set of data for which the true classes are known, and is termed supervised learning. The overall objective of a classifier is to be able to predict successfully on new data. Performance is measured in terms of the error rate. Error rate estimation techniques not only provide a basis for objectively comparing the error rate of several classifiers on the same data and then estimating their future performance on new data, but they can also be a powerful tool for designing a classifier. The techniques of error rate estimation are based on statistical resampling theory. In this work, rules induced complexity of propositional and relational learning systems as well as several resampling methods to estimate the true error rate are discussed. An environment called AMPSAM has been implemented to aid in the application of resampling methods to real world data sets. AMPSAM consists of a collection of interdependent programs that can be bound together either by already defined or by new user defined scripts. The environment uses a common file format for data sets which is independent of any specific classifier scheme. It also includes facilities for splitting data sets up into test and training sets using different methods. Besides holdout, which is the most common accuracy estimation method, AMPSAM supports n-fold cross-validation --- including leaving-one-out --- and bootstrap. The confusion matrices produced in each run using those test and training sets can be input to another system called SMEC. This system calculates and graphically displays some of the most important descriptive measures related to central tendency and dispersion of those data. This work also reports the results of experiments measuring the classification error of three well known propositional and relational classifiers, using the implemented systems, on several data sets commonly used in Machine Learning research.

Οι μνήμες τεχνολογίας NAND Flash χρησιμοποιούνται ευρέως για αποθήκευση δεδομένων λόγω της χαρακτηριστικής πυκνότητας, της χαμηλής απαιτούμενης ισχύος, του χαμηλού κόστους, της υψηλής διεκπεραιωτικής ικανότητας και της αξιοπιστίας τους. Η ανάπτυξη της πολυεπίπεδης τεχνολογίας (MLC) έχει καταστήσει δυνατή την αντικατάσταση των σκληρών δίσκων οδήγησης (HDDs) στις φορητές συσκευές και ορισμένους υπολογιστές με NAND μνήμες. Βεβαίως, οι NAND μνήμες δεν διακρίνονται για την απουσία λαθών κατά την αποθήκευση, αλλά στηρίζονται σε τεχνικές διορθώσεις λαθών (ECC) για να επιτύχουν την κατάλληλη αξιοπιστία. Διάφορα φαινόμενα οδηγούν σε λάθη αποθήκευσης στις Flash μνήμες. Σκοπός της παρούσας διπλωματικής εργασίας είναι η ανάλυση αυτών των μηχανισμών εισαγωγής λαθών και η μελέτη από φυσικής πλευράς της τεχνολογίας των MLC NAND Flash μνημών. καθώς και η πειραματική αξιολόγηση τους και η εξαγωγή των αναγκαίων συμπερασμάτων. / --

Um sistema de aprendizado supervisionado é um programa capaz de realizar decisões baseado na experiência contida em casos resolvidos com sucesso. As regras de classificação induzidas por um sistema de aprendizado podem ser analisadas segundo dois critérios: a complexidade dessas regras e o erro de classificação sobre um conjunto independente de exemplos. Sistemas de aprendizado têm sido desenvolvidos na prática utilizando diferentes paradigmas incluindo estatística, redes neurais, bem como sistemas de aprendizado simbólico proposicionais e relacionais. Diversos métodos de aprendizado podem ser aplicados à mesma amostra de dados e alguns deles podem desempenhar melhor que outros. Para uma dada aplicação, não existem garantias que qualquer um desses métodos é necessariamente o melhor. Em outras palavras, não existe uma análise matemática que possa determinar se um algoritmo de aprendizado irá desempenhar melhor que outro. Desta forma, estudos experimentais são necessários. Neste trabalho nos concentramos em uma tarefa de aprendizado conhecida como classificação ou predição, na qual o problema consiste na construção de um procedimento de classificação a partir de um conjunto de casos no qual as classes verdadeiras são conhecidas, chamado de aprendizado supervisionado. O maior objetivo de um classificador é ser capaz de predizer com sucesso a respeito de novos casos. A performance de um classificador é medida em termos da taxa de erro. Técnicas experimentais para estimar a taxa de erro verdadeira não somente provêem uma base para comparar objetivamente as performances de diversos algoritmos de aprendizado no mesmo conjunto de exemplos, mas também podem ser uma ferramenta poderosa para projetar um classificador. As técnicas para estimar a taxa de erro são baseadas na teoria estatística de resampling. Um ambiente chamado AMPSAM foi implementado para ajudar na aplicação dos métodos de resampling em conjuntos de exemplos do mundo real. AMPSAM foi projetado como uma coleção de programas independentes, os quais podem interagir entre si através de scripts pré-definidos ou de novos scripts criados pelo usuário. O ambiente utiliza um formato padrão para arquivos de exemplos o qual é independente da sintaxe de qualquer algoritmo. AMPSAM também inclui ferramentas para particionar conjuntos de exemplos em conjuntos de treinamento e teste utilizando diferentes métodos de resampling. Além do método holdout, que é o estimador de taxa de erro mais comum, AMPSAM suporta os métodos n-fold cross-validation --- incluindo o leaning-one-out --- e o método bootstrap. As matrizes de confusão produzidas em cada iteração utilizando conjuntos de treinamento e teste podem ser fornecidas a um outro sistema implementado chamado SMEC. Este sistema calcula e mostra graficamente algumas das medidas descritivas mais importantes relacionadas com tendência central e dispersão dos dados. Este trabalho também relata os resultados experimentais a respeito de medidas do erro de classificação de três classificadores proposicionais e relacionais bem conhecidos, utilizando ambos os sistemas implementados, em diversos conjuntos de exemplos freqüentemente utilizados em pesquisas de Aprendizado de Máquina. / A learning system is a computer program that makes decisions based on the accumulative experience contained in successfully solved cases. The classification rules induced by a learning system are judged by two criteria: their classification error on an independent test set and their complexity. Practical learning systems have been developed using different paradigms including statistics, neural nets, as well as propositional and relational symbolic machine learning. Several learning methods can be applied to the same sample data and some of them may do better than others. Still, for a given application, there is no guarantee that any of these methods will work or that any single method is necessarily the best one. In other words, there is not a mathematical analysis method that can determine whether a learning system algorithm will work well. Hence, experimental studies are required. In this work we confine our attention to the learning task known as classification or prediction, where the problem concerns the construction of a classification procedure from a set of data for which the true classes are known, and is termed supervised learning. The overall objective of a classifier is to be able to predict successfully on new data. Performance is measured in terms of the error rate. Error rate estimation techniques not only provide a basis for objectively comparing the error rate of several classifiers on the same data and then estimating their future performance on new data, but they can also be a powerful tool for designing a classifier. The techniques of error rate estimation are based on statistical resampling theory. In this work, rules induced complexity of propositional and relational learning systems as well as several resampling methods to estimate the true error rate are discussed. An environment called AMPSAM has been implemented to aid in the application of resampling methods to real world data sets. AMPSAM consists of a collection of interdependent programs that can be bound together either by already defined or by new user defined scripts. The environment uses a common file format for data sets which is independent of any specific classifier scheme. It also includes facilities for splitting data sets up into test and training sets using different methods. Besides holdout, which is the most common accuracy estimation method, AMPSAM supports n-fold cross-validation --- including leaving-one-out --- and bootstrap. The confusion matrices produced in each run using those test and training sets can be input to another system called SMEC. This system calculates and graphically displays some of the most important descriptive measures related to central tendency and dispersion of those data. This work also reports the results of experiments measuring the classification error of three well known propositional and relational classifiers, using the implemented systems, on several data sets commonly used in Machine Learning research.

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第22704号 / 工博第4751号 / 新制||工||1743(附属図書館) / 京都大学大学院工学研究科建築学専攻 / (主査)教授 池田 芳樹, 教授 西山 峰広, 准教授 聲高 裕治 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

Thesis (MScEng (Mechanical and Mechatronic Engineering))--University of Stellenbosch, 2010. / ENGLISH ABSTRACT: The Pebble Bed Modular Reactor (PBMR) is a Generation IV graphite-moderated helium cooled nuclear reactor which is being developed in South Africa. The PBMR design is based on the German Arbeitsgemeinschaft Versuchreaktor (AVR). The AVR was decommissioned in December 1988 due to operational and safety problems. The PBMR project has put a lot of emphasis on safety and therefore all safety issues relating to the AVR have to be addressed before this technology can be implemented. After the decommissioning of the AVR plant, technicians found radioactive isotopes of cesium 55Cs137, 55Cs134, silver 44Ag110 and strontium 38Sr90 as well as graphite dust in the primary coolant loop of the reactor. These isotopes as well as the graphite dust have to be removed from the helium coolant stream because it can be potentially harmful to equipment, personnel and the general public. The main objective of this thesis is therefore to investigate a separation method for removing the graphite dust (and with it the radioactive isotopes) from the helium coolant stream and also test this method under different operating conditions and geometrical configurations to determine its dust separation efficacy. The device chosen to investigate is the Ranque-Hilsch vortex tube. The Ranque-Hilsch vortex tube (RHVT) is a simple device having no moving parts that produces a hot and cold air stream simultaneously at its two ends from a compressed air source. The vortex generated by the vortex generator located at the inlet of the RHVT causes strongly rotating flows similar in speed to that of a gas centrifuge. The gas centrifuge is used for isotope separation. The RHVT, in theory, can therefore be implemented to separate the graphite/silver isotopes from the helium coolant with the added benefit of either cooling or heating the coolant and was thus selected as the separation technique to be tested experimentally. The dust separation efficiency of the RHVT was tested experimentally using different grades of graphite dust, different fluids, various inlet volumetric flow rates and volume fractions and different RHVT geometries. The experimental results showed that the RHVT has a dust separation efficiency of more than 85 %. A regression analysis was also done with the experimental data to obtain a correlation between the different operating conditions (such as volumetric flow rate) and the dust separation efficiency that can be used to predict the dust efficiency under different operating and geometric conditions (such as the PBMR environment). An analytical model is also presented to describe the ‘temperature separation’ phenomenon in the RHVT, using basic thermo-physical principals to gain a better understanding of how the RHVT works. A CFD analysis was also attempted to supplement the analytical analysis but the solution did not converge and therefore only the preliminary results of the analysis are discussed. / AFRIKAANSE OPSOMMING: Die “Pebble Bed Modular Reactor” (PBMR) is `n vierde generasie grafiet gemodereede en helium verkoelde reaktor wat in Suid-Afrika ontwikkel word. Die PBMR ontwerp is gebaseer op the Duitse Arbeitsgemeinschaft Versuchreaktor (AVR) wat buite werking gestel is in Desember 1988 as gevolg van operasionele en veiligheidsprobleme. Die PBMR projek lê baie klem op veiligheid en daarom moet alle veiligheidskwessies van die AVR eers aangespreek word voor die tegnologie geimplementeer kan word. Nadat die AVR buite werking gestel is, het AVR tegnisie radioaktiewe isotope van cesium 55Cs137, 55Cs134, silwer 44Ag110 en strontium 38Sr90 asook grafiet stof in die primêre stroomkring van die reaktor gevind. Hierdie isotope sowel as die grafiet stof moet uit die helium verkoelingsmiddel in die primere stroomkring van die reaktor verwyder word aangesien dit dalk skadelik kan wees vir toerusting, personeel en die publiek. Die hoofdoelwit van hierdie tesis is dus om `n skeidingstekniek te ondersoek wat die stof (en dus ook die radioaktiewe isotope) uit die helium verkoelingsmiddel kan verwyder. Hierdie tegniek moet dan getoets word onder verskillende operasionele en geometriese toestande om die skeidingsbenuttingsgraad te bepaal. Die toestel wat gekies is om ondersoek te word is die “Ranque-Hilsch Vortex Tube”. Die “Ranque-Hisch Vortex Tube” (RHVT) is a eenvoudige uitvindsel wat geen bewegende parte bevat nie en wat warm en koue lug gelyktydig produseer vanaf `n saamgepersde lugbron. ‘n Baie sterk roteerende vloei word gegenereer in die RHVT wat dieselfde snelhede bereik as die lug in `n gas-sentrifugeerder. Die gas- sentrifugeerder word gebruik as `n isotoopskeidingsapparaat. In teorie kan die RHVT dus ook gebruik word om partikels te skei as gevolg van die sterk roteerende vloei, met die voordeel dat dit ook die lug kan verhit en verkoel. As gevolg van hierde redes is die RHVT gekies as die skeidingstegniek om te ondersoek en dus experimenteel te toets. Die benuttingsgraad van die RHVT se vermoë om die grafiet stof van die lug te skei was gevolglik eksperimenteel getoets deur gebruik te maak van verskillende gehaltes grafiet stof, verskillende vloeistowwe (lug of helium), verskillende inlaat volumevloeitempos en volume fraksies en RHVT geometrieë. Die experimentele resultate het getoon dat die RHVT `n benuttingsgraad van meer as 85 % het. `n Regressie analise was ook gedoen met die eksperimentele data om `n korrelasie tussen die verskillende opersionele toestande (soos volumevloeitempo) en die stof skeiding benuttingsgraad te kry. Hierdie korrelasie kan dan gebruik word om die stofskeidingsbenuttingsgraad onder ander operasionele en geometriese omstandighede, soos die PBMR omgewing, te voorspel. `n Analitiese model word ook voorgestel om die “temperatuur-skeidings” meganisme in die RHVT te verduidelik, met die hulp van basiese termo-fisiese beginsels, om beter te verstaan hoe dit werk. Daar was ook gepoog om `n CFD analise te doen wat die analitiese model kon aanvul, maar die numeriese oplossing het nie gekonvergeer nie en dus word net die voorlopige resultate van dié analise bespreek.

A popularização dos dispositivos móveis em todas as camadas sociais tem motivado o desenvolvimento de aplicações educacionais móveis, denominadas aplicações de m-learning. Neste cenário, as aplicações existentes, mesmo possuindo diversos benefícios e facilidades no que diz respeito ao ensino e aprendizagem, apresentam problemas e desafios relacionados, sobretudo no que se refere ao desenvolvimento, reuso e padronização arquitetural. Por outro lado, na vertente do reúso de software, percebe-se uma crescente adoção do conceito de Linha de Produtos de Software (LPS). Esse paradigma possibilita às organizações explorar as similaridades e variabilidades de seus produtos, aumentando a reutilização de artefatos e, como consequência, diminuindo custos e tempo de desenvolvimento. Neste trabalho é apresentada uma LPS voltada ao domínio das aplicações de m-learning, denominada M-SPLearning. A proposição da M-SPLearning envolveu desde o estudo inicial para a obtenção de uma análise de domínio efetiva, até a implementação dos componentes previamente analisados. A LPS concebida teve seus respectivos produtos avaliados experimentalmente no cenário industrial, fornecendo evidências de que sua utilização pode acelerar o time-to-market de produtos de m-learning, com um número reduzido de defeitos. / The popularity of mobile devices in all social classes has motivated the development of mobile educational applications, called m-learning applications. The existing applications, even having many benefits and facilities in relation to teaching and learning, also have problems and challenges, especially regarding the development, reuse and architectural standardization. Particularly, there is an increasing adoption of the concept of Software Product Line (SPL) in researches related to reuse. This paradigm enables organizations to explore the similarities and variabilities of their products, increasing the reuse of artifacts and, consequently, reducing costs and development time. This work presents an SPL focused on the domain of m-learning applications, named M-SPLearning. The development of M-SPLearning has comprised since the initial study for an effective domain analysis until the implementation of the components previously analyzed. Such SPL had its products experimentally evaluated in the industrial scenario, providing evidences that its use can accelerate the time-to-market of m-learning applications, with a reduced number of faults.

A popularização dos dispositivos móveis em todas as camadas sociais tem motivado o desenvolvimento de aplicações educacionais móveis, denominadas aplicações de m-learning. Neste cenário, as aplicações existentes, mesmo possuindo diversos benefícios e facilidades no que diz respeito ao ensino e aprendizagem, apresentam problemas e desafios relacionados, sobretudo no que se refere ao desenvolvimento, reuso e padronização arquitetural. Por outro lado, na vertente do reúso de software, percebe-se uma crescente adoção do conceito de Linha de Produtos de Software (LPS). Esse paradigma possibilita às organizações explorar as similaridades e variabilidades de seus produtos, aumentando a reutilização de artefatos e, como consequência, diminuindo custos e tempo de desenvolvimento. Neste trabalho é apresentada uma LPS voltada ao domínio das aplicações de m-learning, denominada M-SPLearning. A proposição da M-SPLearning envolveu desde o estudo inicial para a obtenção de uma análise de domínio efetiva, até a implementação dos componentes previamente analisados. A LPS concebida teve seus respectivos produtos avaliados experimentalmente no cenário industrial, fornecendo evidências de que sua utilização pode acelerar o time-to-market de produtos de m-learning, com um número reduzido de defeitos. / The popularity of mobile devices in all social classes has motivated the development of mobile educational applications, called m-learning applications. The existing applications, even having many benefits and facilities in relation to teaching and learning, also have problems and challenges, especially regarding the development, reuse and architectural standardization. Particularly, there is an increasing adoption of the concept of Software Product Line (SPL) in researches related to reuse. This paradigm enables organizations to explore the similarities and variabilities of their products, increasing the reuse of artifacts and, consequently, reducing costs and development time. This work presents an SPL focused on the domain of m-learning applications, named M-SPLearning. The development of M-SPLearning has comprised since the initial study for an effective domain analysis until the implementation of the components previously analyzed. Such SPL had its products experimentally evaluated in the industrial scenario, providing evidences that its use can accelerate the time-to-market of m-learning applications, with a reduced number of faults.

Regular Path Queries (RPQs) are at the core of many recent declarative graph pattern matching languages. They leverage the compactness and expressiveness of regular expressions for matching recursive path structures. Unfortunately, most prior works on RPQs only consider breadth-first search as traversal strategy, neglecting other possible graph traversals like depth-first search or a combination of both. Within this paper, we conduct an analysis of graph traversals for RPQs by introducing a generalized graph traversal frame-work subsuming breadth-first search and depth-first search as extreme cases and thus opening up a new design space for graph traversals algorithms. We outline the underlying principles as well as provide comprehensive experimental evaluation using implementations which yield beneficial results regarding evaluation time and peak memory consumption.

Thesis (PhD (Electrical and Electronic Engineering))--University of Stellenbosch, 2011. / ENGLISH ABSTRACT: The focus of this thesis is on the optimal design, control and evaluation of 3-phase permanent magnet radial flux synchronous machines with non-overlapping, concentrated-coil, double layer stator windings for EV hub drive applications. A simple analytical method is developed that can be used as a first design tool. The method uses and predicts the MMF harmonic content for a certain pole-slot combination as well as the harmonic content for the air gap permeance function. These harmonics are then used to calculate the torque and torque ripple of machines with large stator slot openings and surface mounted permanent magnets. A different approach to calculate the iron, stator copper eddy current and magnet losses is presented. This method specifically looks at the machine during field weakening operation when the flux paths are changing in the machine. Flux density information throughout the machine is extracted from a series of static FE solutions, to calculate the losses and to combine this with an empirical formula. Some machine topology choices are compared for use as hub drives in small electric ve- hicles. The parameters that influence the machine design are discussed and evaluated after a multidimensional design optimization is done and an efficient control algorithm is imple- mented. The algorithm works through the entire operating speed range and make use of, automatically generated, 2D look up tables to determine the correct current reference. A stator lamination design is proposed, that combines the use of rectangular preformed coils and semi-closed stator slots. Two prototype machines, one with a good winding factor and the other with a low winding factor, are built and compared. The manufacturing and testing of the two prototype machines are described and shown in detail. / AFRIKAANSE OPSOMMING: Die fokus van hierdie tesis is op die optimale ontwerp, beheer en evaluasie van 3-fase per- manent magneet radiale vloed sinchroon masjiene met nie-oorvleuelende, gekonsentreerde, dubbel laag stator wikkelinge vir EV hub motor toepassings. ’n Eenvoudige analitiese metode is ontwikkel wat as ’n eerste ontwerp gereedskap stuk gebruik kan word. Die metode gebruik en voorspel die MMF se frekwensie inhoud vir ’n sekere pool-gleuf kombinasie sowel as die frekwensie inhoud vir die lug spleet permeansie funksie. Hierdie frekwensie inhoud word dan gebruik om die draaimoment en draaimoment riffel van masjiene met groot stator gleuf openinge en oppervlak magnete te voorspel. ’n Ander benadering om yster, stator koper werwel stroom en magneet verliese te bepaal word voorgestel. Hierdie metode kyk spesifiek na masjiene onder veld verswakking beheer wanneer die vloed paaie verander vanaf die normale. Die vloeddigtheid, regdeur die masjien, word verkry deur om van ’n reeks statiese eindige element oplossings gebruik te maak en dit te kombineer met ’n empiriese verliesberekening. Die parameters wat die masjienontwerp beïnvloed, word bespreek en geëvalueer na ’n mul- tidimensionele ontwerp optimering gedoen is en ’n effektiewe beheer algoritme geïmplimen- teer is. Die algoritme werk vir enige spoed en is gebaseer op die outomaties gegenereerde 2D opsoek tabelle wat die korrekte stroomverwysing gee. ’n Stator laminasie ontwerp word voorgestel wat die gebruik van vooraf vervaardigde spoele en gedeeltelik toe stator gleuwe moontlik maak. Twee prototipe masjiene, een met ’n goeie windingsfaktor en een met ’n swakker windingsfaktor is gebou en vergelyk. Die ver- vaardiging en toetsing van die twee prototipe masjiene word in detail beskryf en gewys.