A Markov Decision Process (MDP) is a model employed to describe problems in which a decision must be made at each one of several stages, while receiving feedback from the environment. This type of model has been extensively studied in the operations research community and fundamental algorithms have been developed to solve associated problems. However, these algorithms are quite inefficient for very large problems, leading to a need for alternatives; since MDP problems are provably hard on compressed representations, one becomes content even with algorithms which may perform well at least on specific classes of problems. The class of problems we deal with in this thesis allows succinct representations for the MDP as a dynamic Bayes network, and for its solution as a weighted combination of basis functions. We develop novel algorithms for producing, improving, and calculating the error of approximate solutions for MDPs using a compressed representation. Specifically, we develop an efficient branch-and-bound algorithm for computing the Bellman error of the compact approximate solution regardless of its provenance. We introduce an efficient direct linear programming algorithm which, using incremental constraints generation, achieves run times significantly smaller than existing approximate algorithms without much loss of accuracy. We also show a novel direct linear programming algorithm which, instead of employing constraints generation, transforms the exponentially many constraints into a compact form more amenable for tractable solutions. In spite of its perceived importance, the efficient optimization of the Bellman error towards an approximate MDP solution has eluded current algorithms; to this end we propose a novel branch-and-bound approximate policy iteration algorithm which makes direct use of our branch-and-bound method for computing the Bellman error. We further investigate another procedure for obtaining an approximate solution based on the dual of the direct, approximate linear programming formulation for solving MDPs. To address both the loss of accuracy resulting from the direct, approximate linear program solution and the question of where basis functions come from we also develop a principled system able not only to produce the initial set of basis functions, but also able to augment it with new basis functions automatically generated such that the approximation error decreases according to the user's requirements and time limitations.

A Markov Decision Process (MDP) is a model employed to describe problems in which a decision must be made at each one of several stages, while receiving feedback from the environment. This type of model has been extensively studied in the operations research community and fundamental algorithms have been developed to solve associated problems. However, these algorithms are quite inefficient for very large problems, leading to a need for alternatives; since MDP problems are provably hard on compressed representations, one becomes content even with algorithms which may perform well at least on specific classes of problems. The class of problems we deal with in this thesis allows succinct representations for the MDP as a dynamic Bayes network, and for its solution as a weighted combination of basis functions. We develop novel algorithms for producing, improving, and calculating the error of approximate solutions for MDPs using a compressed representation. Specifically, we develop an efficient branch-and-bound algorithm for computing the Bellman error of the compact approximate solution regardless of its provenance. We introduce an efficient direct linear programming algorithm which, using incremental constraints generation, achieves run times significantly smaller than existing approximate algorithms without much loss of accuracy. We also show a novel direct linear programming algorithm which, instead of employing constraints generation, transforms the exponentially many constraints into a compact form more amenable for tractable solutions. In spite of its perceived importance, the efficient optimization of the Bellman error towards an approximate MDP solution has eluded current algorithms; to this end we propose a novel branch-and-bound approximate policy iteration algorithm which makes direct use of our branch-and-bound method for computing the Bellman error. We further investigate another procedure for obtaining an approximate solution based on the dual of the direct, approximate linear programming formulation for solving MDPs. To address both the loss of accuracy resulting from the direct, approximate linear program solution and the question of where basis functions come from we also develop a principled system able not only to produce the initial set of basis functions, but also able to augment it with new basis functions automatically generated such that the approximation error decreases according to the user's requirements and time limitations.

The behavior of the electric field together with the electron and ion densities in the vicinity of a nonemitting, plane anode is investigated. The selected approach involves non-linear analysis techniques on the continuum equations for steady-state, isothermal conditions where both ionization and two-body recombination are included. Ions, created through electron bombardment of neutral atoms, are repelled toward two stagnation regions: within or near the sheath boundary and near the plasma interface. These equilibria form as a result of the chemistry present: recombination establishes the latter while ionization stipulates the former. As presented, the sheath is fundamentally unstable - ions are driven toward the negative electrode. Using nitrogen data for a numeric example, the following observations are made: a sufficiently strong applied electric field pushes the ion density toward that ofthe electrons through a well - a constrictive phenomenon. Both a transition region, dominated by density gradients, and a diffusion-driven zone are found to move the system toward the plasma interface. The characteristics of this process are influenced by the applied electric field, but the instability of the chemistry-induced stagnation regions precludes numeric convergence. Insufficient dissipation may prevent the stability of the anode fall model as presented. Suggested improvements to the model descriptions include considering the effects of temperature gradients, magnetic fields, three-body recombination, diffusion written in terms of the electric field, multi-dimensionality and/or timedependencies^

No description available.

The purpose of this thesis was to assess the repeatability and reproducibility of two high-speed macrotexture measuring systems. The first portion of the study collected macrotexture measurements using the two high-speed systems on the Virginia Smart Road facility and validated the reproducibility of the mean profile depth (MPD) measurements with reference CT Meter measurements. The various data sets were then compared with each other. The objective was to determine whether the two systems are collecting repeatable and reproducible data. The analysis showed that the two high-speed systems investigated have good repeatability (0.105 mm for the Ames and 0.113 mm for the SCRIM) when measuring the average MPD of the sections investigated. The two systems produce measurements that are highly-correlated (Ames R2 = 0.9591 and SCRIM R2 = 0.9157) with the reference ones obtained with the CT Meter. While the Ames systems, with the data processed using the Virginia Tech filter, measures MPD values that are very close to those of the CT Meter, with a virtually zero systematic bias. The SCRIM obtains slightly lower readings. The differences are thought to be due to the filtering of the raw pavement elevation measurements used by the SCRIM processing software to eliminate dropout and spikes in the laser measurements. / Master of Science / The researched compared two different devices used to record data for a surface property of pavement, macrotexture. The macrotexture is a measure of the spaces in between the particles making up the surface of the asphalt. This property is linked to the level of friction on the roadways affecting safety. The readings from each of the devices were validated through a reference device in order to insure they were obtaining reliable results on the Virginia Smart Road. The repeatability and reproducibility of each of the devices was examined to determine whether the two systems were collecting repeatable and reproducible data. It was determined the two devices had good repeatability and were highly correlated with the reference device. The first device called the Ames system obtained measurements very close to those of the reference device, with a virtually zero systematic bias. The second device called the SCRIM system obtained measurements close to the reference device but slightly lower. Through the comparisons and data analysis, an equation permitting users to convert readings from the SCRIM to Ames system and vice versa was computed. Because one device obtains slightly higher readings than the other, this equation is very important in order to have comparable results.

[pt] A técnica de Perfuração com Gerenciamento de Pressão está em crescente expansão no contexto das operações marítimas. Devido à complexidade dos reservatórios e das novas fronteiras de exploração e produção de petróleo, a perfuração com Gerenciamento de Pressão se apresenta como uma forma de redução de custo e aumento da segurança operacional para um grande volume das reservas de petróleo. Em alguns casos, a tecnologia é utilizada como viabilizadora, sendo a única forma de se perfurar os poços. Neste cenário, o entendimento do comportamento de migração de gás, quando da aplicação da técnica de Pressurized Mud Cap Drilling (PMCD), permite a elaboração do projeto de poço mais econômico além de contribuir para o aumento da segurança operacional. O presente estudo tem por objetivo apresentar um modelo matemático capaz de simular o comportamento da perda de fluido de perfuração para a formação com a ocorrência de influxo simultâneo e da migração de gás para a superfície, durante a perfuração com esta técnica, com robustez. É utilizado o modelo bifásico Drift-Flux associado ao método numérico Advection Upstream Splitting Method (AUSMV) para simular o comportamento descrito anteriormente. Antes da apresentação de estudos de casos de simulação com migração de gás, os resultados de cenários mais simples são comparados com os resultados de um software comercial tido, como referência na indústria para validação do programa. Através dos casos simulados é possível verificar a robustez do modelo matemático proposto, que se mostra capaz de fornecer respostas compatíveis quanto ao comportamento esperado do gás. A análise dos resultados obtidos permite estabelecer procedimentos para o monitoramento do que ocorre no poço de forma a otimizar as operações de bullhead no cenário de PMCD. / [en] The use of the Manage Pressure Drilling (MPD) is spreading in offshore operations. The increasing complexity in the new exploratory frontiers is demanding for new techniques to reduce costs and increase operational safety. MPD appears as an answer for that demand and sometimes it is the only viable way to drill some of the challenging wells. In that way, understanding the gas migration behavior while drilling in PMCD mode allows an optimized well design concerning cost and operational safety. The present study validates a mathematical model capable of simulating a scenario where loss of drilling fluid in the bottom of the well is present while having an influx from the same reservoir and observing gas migration to the surface in a PMCD operation. A Drift Flux Two Phase Flow Model is used in association with the Advection Upstream Splitting Method (AUSMV). Before the presentation and discussion of the complete PMCD scenario, two simple cases were simulated and the results were compared to the ones from a computer application considered as a reference to the industry, validating the proposed model. The results of the simulations can be used as a base for the elaboration of operational procedures to monitor gas behavior and optimize bullhead in PMCD scenarios.

[pt] Para acesso aos reservatórios portadores de hidrocarbonetos é necessária a perfuração de poços de petróleo. Entre as várias funções do fluido de perfuração podem se destacar o transporte do cascalho gerado pela broca, a função de primeira barreira para inibição de influxo das formações expostas no poço e a manutenção da pressão do poço dentre os limites de resistência das formações atravessadas (janela operacional). Alguns novos campos petrolíferos, bem como campos depletados sofrem restrições operacionais devido a pequenas janelas operacionais. A variação da pressão no poço aberto devido a perdas por atrito no anular, em alguns casos é maior do que a janela operacional, impedindo a perfuração do poço. Managed Pressure Drilling (MPD) é uma tecnologia alternativa à perfuração convencional. O anular do poço é vedado, possibilitando a retenção de pressão no anular. O ajuste da pressão no anular do poço possibilita compensar a variação da perda de atrito e perfurar poços com janelas operacionais estreitas. Para manutenção da pressão da zona objetivo dentro da janela operacional é necessário o controle adequado da pressão na superfície. A definição do valor de pressão necessário é obtida através do conhecimento do comportamento da dinâmica do fluido de perfuração no poço. Este trabalho desenvolveu um modelo de um sistema de poço de petróleo com MPD, levando em conta os efeitos estáticos e dinâmicos envolvidos na perfuração MPD. Sobre o modelo desenvolvido foram propostas e aplicadas diferentes técnicas de controle linear e não-linear para avaliação do desempenho dos controladores e identificação de pontos-chave na aplicação da técnica MPD. Os controles nãolineares propostos foram bem sucedidos em compensar os efeitos altamente nãolineares do sistema, melhorando sua resposta em tarefas típicas. / [en] For access of hydrocarbon reservoirs, drilling oil wells is required.. Among the various functions the drilling fluid may provide are cutting transportation generated by the bit, a barrier function for inhibiting influx of exposed formations in the well and maintaining well pressure within the resistance limits of the formations exposed (operational window). Some new oil fields as well as some depleted fields are experiencing severe operational restrictions due to small operational windows. The open hole pressure variation due to friction losses on the annular in some cases is greater than the operational window, inhibiting the perforation of the well. Managed Pressure Drilling (MPD) is an alternative technology to conventional drilling. The annular of the well is sealed while pressure is contained within it. By adjusting annular pressure it is possible to compensate for friction loss variation on the annular and drill well sections with close operational windows. For the proper pressure maintenance on the target zone within the operating window, it is required adequate pressure control on the surface. The definition of the necessary pressure value is obtained through knowledge of the dynamic behavior of the drilling fluid in the well. This work has developed a model of an oil well system with MPD, taking into account the static and dynamic effects involved in MPD. To the developed model different linear and non-linear control techniques were proposed and applied for evaluating the performance of controllers and identifying key points in implementing the MPD technique. The proposed non-linear control techniques were able to compensate for the highly non-linear nature of the system, improving its response during typical tasks.

[pt] Caracterização Reológica de Fluidos para Perfuração com Gerenciamento de Pressão. Forças Hidráulicas desempenham uma função importante em muitas operações de campo de petróleo, incluindo perfuração, completação, fraturamento, acidificação, workover e produção. Em aplicações de Perfuração com Gerenciamento de Pressão (Managed Pressure Drilling - MPD), onde as estimativas de perdas de pressão são críticas para controlar o poço dentro da janela de operacional, é necessário utilizar a reologia correta para a modelagem matemática precisa do comportamento do fluido. Os métodos API (American Petroleum Institute) empregam para os cálculos de hidráulica os modelos reológicos de Herschel-Bulkley (H-B), Power Law (PL) ou plástico de Bingham. Este trabalho resume os resultados de um estudo aprofundado sobre as questões e os aspectos relevantes relacionados com o equipamento e os métodos utilizados para caracterizar os fluidos de perfuração para aplicações MPD, bem como as implicações operacionais que divergem das práticas convencionais. Uma comparação da caracterização reológica de fluídos é feita usando reômetros de alta precisão contra métodos convencionais tais como o viscosímetro FANN35. Subsequentemente, é apresentada uma comparação da seleção do modelo reológico proposto por API 13B em contrapartida com o método de Regressão Não Linear (NLR). Investigações detalhadas das faixas de taxas de cisalhamento são apresentadas para geometrias de um poço anular MPD típico, calculadas através de Dinamica de Fluidos Computacional (CFD) e comparadas com as fórmulas sugeridas na API RP 13D. Para concluir, é apresentada uma discussão sobre os efeitos das medições, do tratamento de dados (Curve Fit) e do meio ambiente (observações de laboratório em comparação com experiências de campo) na precisão da obtenção da reologia do fluido e as consequências na estimativa das perdas de carga no anular. / [en] Enhanced fluid rheology characterization for Manage Pressure Drilling. Hydraulics play an important role in many oil field operations including drilling, completion, fracturing, acidizing, workover and production. In Managed Pressure Drilling (MPD) applications, where pressure losses become critical to accurately estimate and control the well within the operational window, it is necessary to use the correct rheology for a precise mathematical modelling of fluid behavior. The standard API methods for drilling fluid hydraulics employ Herschel-Bulkley (H-B), Power Law (PL) or Bingham plastic as rheological models. This work summarizes the results of an extensive study on issues and relevant aspects related to the equipment and methods used to characterize the drilling fluids for MPD applications, as well as the operational implications that diverge from conventional practices. A comparison of fluid rheology characterization is made using high precision rheometers versus conventional FANN35 methods. Subsequently, a comparison of rheology model selection proposed by API 13B and by Non Linear Regression (NLR) is presented. Further investigation of shear rate ranges is presented in a MPD typical annular geometry. Results obtained via Computational Fluid Dynamics (CFD), and with the formulas suggested in API RP 13D are compared. To conclude, the effects of measurements, data treatment (Curve Fit), and environment (laboratory observations versus field experiences) in the accuracy of fluid rheology characterization and annulus pressure loss estimation are presented and discussed.

In this thesis the industrial design studio has been investigated with particular reference to studio thinking and learning and project-based activity. This investigation has been set in the context of a final-year, degree program in industrial design that includes a substantial research and development project. From a critical review of the relevant literature the characteristics of studio culture have been identified, together with its role in the teaching of both creative and systematic endeavour. In addition, the history and context of the role of the industrial/product designer is reviewed in order to understand the nature and the required skills of the discipline. In this thesis, an initial study surveyed academics involved in teaching industrial design in Australia, and overseas. The study sought to determine the approach of students, in various industrial design degree programs, to their final-year projects and the extent to which design process and design methods were incorporated in their project reports. The findings revealed a number of operational needs associated with studio-based learning, particularly those associated with final-year, project-based activity. These findings, together with teachings from the literature concerning how students go about design in the studio and the needs associated with project activity, led to the proposal of a generic model, entitled the Major Project Development Model "MPD Model". The operational criteria in the MPD Model guided the development of a computer-integrated system of design methods allocated to the respective phases of the process. This system, called the "MPD System", is designed to support and enhance student design work in major projects. A second study was conducted that analysed: student performance in their project reports; the extent to which their design research conformed to the MPD Model; and the extent to which design methods were used in their final-year projects. Criteria and guidelines for the successful conduct and evaluation of such projects have been proposed and set up as part of the experimental programme. The experimental work, reported in this thesis, is based on an in-depth, comparative investigation of a range of major project reports, firstly those produced in the year 2003 during which final-year students did not have access to or knowledge of the MPD System and secondly, those produced in 2004 where students were provided with the MPD System, hence providing two cohorts for comparative purposes. The theoretical and experimental work have been related, with appropriate results and conclusions, to the following issues: Design theory � an MPD Model has been proposed and applied in keeping with a set of operational criteria; design methods - a model reflecting a range of methods aligned to phases of the MPD Model have been established in keeping with needs of designers in their execution of phases of the process; brainbased learning theory � a model of the integration of the MPD System as a means of linking systematic and creative thinking in the studio process is proposed; academic performance � the academic performance of students has been studied and data have been derived which provide valuable information for the design educational process. The results of this research will encourage use of a more structured teaching and learning approach and the employment of design methods in major projects. This comprehensive research thesis provides a framework for further research and recommendations for further research.

The present elemental analysis workhorse worldwide is Inductively Coupled Plasma (ICP) Atomic Emission Spectrometry (AES) and Mass Spectrometry (MS). Due to the high power requirements, large gas consumption and the more obvious attribute, size, the ICP is tethered to the lab. Usually, samples must be collected, bottled, sometimes stabilized by acids and then shipped back to the lab for analysis (hours to days to weeks turnover rate). Due to the demand for a portable analyzer this thesis will focus on further development of microplasma devices (MPDs) for portable on-site analysis, in (near) real-time. Mini-In-Torch Vapourization (mini-ITV) is the sample introduction method for MPDs which removes the need for sample preparation (further necessitates portability). Mini-ITV introduces the sample into the MPD via electrothermal vapourization of a dry (water-free) nano- to micro- volume sample. Pneumatic nebulization, the commercially available ICP sample introduction method would extinguish the microplasma. Microplasma stability is the first issue addressed by confining the microplasma to a quartz tube (“wall-stabilized”) in hopes of a more stable MPD background emission. Once stabilized MPD conditions were found key microplasma parameters were studied including MPD power, HVac frequency, operating mode, inter-electrode distance (IED) and observation location, in hopes of improved MPD analytical performance. Microplasma excitation mechanism and maximum energy available in the microplasma for analyte are discussed. Some fundamental characteristics such as excitation temperature (Texc) and changes in atom/ion population with variation in some of the key MPD parameters were also determined.