Thesis (PhD)--Stellenbosch University, 2012. / ENGLISH ABSTRACT: Road Tra c legislation prescribes constraints on the forces/loads which vehicles may exert on the road at contact surfaces or load points. These constraints have bearing on speci c contact surfaces or groups of contact surfaces. The constraints are in the form of maximum force values and maximum ratios, which may exist between forces. The forces exerted by a vehicle on the road is determined by a combination of the vehicle's own mass, dimensional characteristics, and the mass and position of its cargo. The mass and position of the cargo can be modelled by only considering its centroid. The centroid's position is linked to a certain cargo arrangement. It is necessary, in terms of legal constraints, to distinguish between legally permissable- and non-permissable arrangements. Methods are proposed in this dissertation to determine the borders for the position of the centroid according to cargo-mass, for a variety of vehicle con gurations. The determination of borders for the position of the centroid is the result of a three part process. Firstly, the legal limits are translated to mathematical inequalities that must be satis ed for the forces at contact surfaces. Secondly, conservative parametric estimates for the forces, in terms of the position and the mass of the centroid of a given vehicle, must be derived. Finally these estimates are incorporated into the load constraints and an enclosed permissable area for the centroid is found. These position constraints are suitable inputs to solution methods of various cargo arrangement problems. Linear programmes which can solve certain general arrangement problems by utilising the position constraints on the centroid, were successfully developed. Solutions are presented for the arrangement of pallet-freight, as well as the arrangement of cargo units with asymetrical centroids over the length or width of a vehicle's cargo area. / AFRIKAANSE OPSOMMING: Padvervoerwetgewing skryf grense op kragte/laste voor wat voertuie by kontakoppervlakke of laspunte, op die pad mag uitoefen. Hierdie voorskrifte het betrekking op spesi eke laspunte en groepe van laspunte. Die voorskrifte is in die vorm van maksimum laswaardes by laspunte en maksimum ratio's wat tussen sekere laste mag bestaan. Die laste wat 'n voertuig op die pad uitvoer word bepaal deur 'n kombinasie van die voertuig se eie massa, dimensionele eienskappe, sowel as sy vrag se massa en posisie. Die vrag se massa en posisie kan gemodelleer word deur slegs die massamiddelpunt daarvan te beskou. Die massamiddelpunt se posisie word gekoppel aan 'n bepaalde vragrangskikking. Dit is nodig om, in terme van wetlike voorskrifte, te kan onderskei tussen wetlik toelaatbare en -ontoelaatbare rangskikkings. Metodes word in hierdie proefskrif voorgestel waarmee daar grense vir die posisie van die vragmassamiddelpunt bepaal kan word na gelang van die vragmassa, vir 'n verskeidenheid van voertuigkon gurasies. Die bepaling van grense vir die posisie van die vragmassamiddelpunt is die resultaat van 'n drieledige proses. Eerstens word die wetlike beperkings herlei na wiskundige ongelykhede waarbinne laste by kontakoppervlakke moet val. Tweedens word konserwatiewe parametriese beramings vir die laste verkry in terme van die ligging en massa van die vragmassamiddelpunt vir 'n gegewe voertuig. Laastens word hierdie beramings in die lasbeperkings opgeneem om 'n afgebakende toelaatbare gebied vir die vragmassamiddelpunt se posisie te verkry. Hierdie posisiebeperkings is geskik om as inset te dien in die oplossingmetodes van 'n verskeidenheid van vragrangskikkingsprobleme. Line^ere programme wat sekere algemene rangskikkingsprobleme kan oplos deur die aanwending van posisiebeperkings op die vragmassamiddelpunt, is suksesvol ontwikkel. Oplossings word aangebied vir die rangskikking van paletvrag, asook die rangskikking van vrageenhede met asimmetriese massamiddelpunte oor die lengte of breedte van 'n voertuig se vragarea.

The dissertation addresses the research topics of machine learning outlined below. We developed the theory about traditional first-order algorithms from convex opti- mization and provide new insights in nonconvex objective functions from machine learning. Based on the theory analysis, we designed and developed new algorithms to overcome the difficulty of nonconvex objective and to accelerate the speed to obtain the desired result. In this thesis, we answer the two questions: (1) How to design a step size for gradient descent with random initialization? (2) Can we accelerate the current convex optimization algorithms and improve them into nonconvex objective? For application, we apply the optimization algorithms in sparse subspace clustering. A new algorithm, CoCoSSC, is proposed to improve the current sample complexity under the condition of the existence of noise and missing entries. Gradient-based optimization methods have been increasingly modeled and inter- preted by ordinary differential equations (ODEs). Existing ODEs in the literature are, however, inadequate to distinguish between two fundamentally different meth- ods, Nesterov’s acceleration gradient method for strongly convex functions (NAG-SC) and Polyak’s heavy-ball method. In this paper, we derive high-resolution ODEs as more accurate surrogates for the two methods in addition to Nesterov’s acceleration gradient method for general convex functions (NAG-C), respectively. These novel ODEs can be integrated into a general framework that allows for a fine-grained anal- ysis of the discrete optimization algorithms through translating properties of the amenable ODEs into those of their discrete counterparts. As a first application of this framework, we identify the effect of a term referred to as gradient correction in NAG-SC but not in the heavy-ball method, shedding deep insight into why the for- mer achieves acceleration while the latter does not. Moreover, in this high-resolution ODE framework, NAG-C is shown to boost the squared gradient norm minimization at the inverse cubic rate, which is the sharpest known rate concerning NAG-C itself. Finally, by modifying the high-resolution ODE of NAG-C, we obtain a family of new optimization methods that are shown to maintain the accelerated convergence rates as NAG-C for minimizing convex functions.

The thesis covers studies conducted during 1976-79 under a Science Research Council contract to examine the uses of reliability information in decision-making in maintenance in the process industries. After a discussion of the ideal data system, four practical studies of process plants are described involving both Pareto and distribution analysis. In two of these studies the maintenance policy was changed and the effect on failure modes and frequency observed. Hyper-exponentially distributed failure intervals were found to be common and were explained after observation of maintenance work practices and development of theory as being due to poor workmanship and parts. The fallacy that constant failure rate necessarily implies the optimality of maintenance only at failure is discussed. Two models for the optimisation of inspection intervals are developed; both assume items give detectable warning of impending failure. The first is based upon constant risk of failure between successive inspections 'and Weibull base failure distribution~ Results show that an inspection/on-condition maintenance regime can be cost effective even when the failure rate is falling and may be better than periodiC renewals for an increasing failure situation. The second model is first-order Markov. Transition rate matrices are developed and solved to compare continuous monitoring with inspections/on-condition maintenance an a cost basis. The models incorporate planning delay in starting maintenance after impending failure is detected. The relationships between plant output and maintenance policy as affected by the presence of redundancy and/or storage between stages are examined, mainly through the literature but with some original theoretical proposals. It is concluded that reliability techniques have many applications in the improvement of plant maintenance policy. Techniques abound, but few firms are willing to take the step of faith to set up, even temporarily, the data-collection facilities required to apply them. There are over 350 references, many of which are reviewed in the text, divided into chapter-related sectionso Appendices include a review of Reliability Engineering Theory, based on the author's draft for BS 5760(2) a discussion of the 'bath-tub curves' applicability to maintained systems and the theory connecting hyper-exponentially distributed failures with poor maintenance practices.

Time series forecasting and modeling are challenging problems during the past decades, because of its plenty of properties and underlying correlated relationships. As a result, researchers proposed a lot of models to deal with the time series. However, the proposed models such as Autoregressive integrated moving average (ARIMA) and artificial neural networks (ANNs) only describe part of the properties of time series. In this thesis, we introduce a new hybrid model integrated filter structure to improve the prediction accuracy. Case studies with real data from University of Kentucky HealthCare are carried out to examine the superiority of our model. Also, we applied our model to operating room (OR) to reduce the inefficiency cost. The experiment results indicate that our model always outperforms compared with other models in different conditions.

With increasing growing the number of batteries, namely accumulators, Consequently, their take-away and a need of recyclation is increasing too. It is thus necessary to find a suitable way to their optimalization that can lead to the establishment of new waste collection machineries. This study deals with the data analysis by ECOBAT s.r.o. company, data processing and a model construction (in five different versions). The output of the diploma thesis is a model of the batteries collection enabling closer look on take-away routes, The model is based on the distances of take-away routes based on a principle of the utility areas and a role of a transport. The data are obtained from program ArcGIS 10.5 and subsequently handled in program Microsoft Excel. The outputs are the map schemes issuing from five different versions. The first one is showing the current situation, The other one is showing the version of treating units in three largest towns of the Czechia, the third one treating units in all the towns with above 100 000 inhabitants. The forth on eis showing a possibility to have a collection point in each of the county seats. The last one has the same location, but the threating units have higher capacities. The model is general, there is therefore a possibility of using it in modeling for other...

Esta dissertação apresenta um problema de otimização relacionado com a designação de rotas de carga completa para frota própria ou dedicada, visando a minimização dos custos de transporte numa rede de distribuição formada por fábricas uni-produto, centros de distribuição (consolidação) e clientes. Essas rotas são conjugadas formando ciclos fechados (viagens) para garantir a otimização do tempo através do movimento contínuo desta frota dedicada. A metodologia é aplicada em uma rede de distribuição de um fabricante de linha branca no Brasil. Além dos resultados econômicos favoráveis, outras contribuições para o tema de conjugação de rotas foram encontradas nesta dissertação, tais como, a regra de formação de caminhos, o limite para conjugação de rotas numa rede de distribuição e o desenvolvimento de uma estrutura para custear esses caminhos conjugados. O modelo de programação linear inteira desenvolvido mostrou-se apto a resolver problemas de tamanho real em tempo factível, mesmo com recursos computacionais comuns. / This thesis presents an optimization problem regarding the assignment of truckload lanes to a private or dedicated fleet to minimize transportation costs in a distribution network formed by single-product plants, distribution (consolidation) centers and clients. These lanes are conjugated in order to form closed cycles (trips) to guarantee time optimization through continuous movement of this dedicated fleet. This methodology is applied to a distribution network of a white goods manufacturer in Brazil. More than good economic results, there are others contributions for the theme of conjugated lanes in this thesis, such as, the rule of formation trips, the limit to conjugated lanes to a distribution network and the development of a structure to get the conjugated lanes\' costs. The model of integer linear program that was development is capable to solve the real problems in a reasonable time, even if with regular.

Production planning problems are of paramount importance within supply chain planning, supporting decisions on the transformation of raw materials into finished products. Lot sizing in production planning refers to the tactical/operational decisions related to the size and timing of production orders to satisfy a demand. The objectives of lot-sizing problems are generally economical-related, such as saving costs or increasing profits, though other aspects may be taken into account such as quality of the customer service and reduction of inventory levels. Lot-sizing problems are very common in production activities and an efficient planning of such activities gives the company a clear advantage over concurrent organizations. To that end it is required the consideration of realistic features of the industrial environment and product characteristics. By means of mathematical modelling, such considerations are crucial, though their inclusion results in more complex formulations. Although lot-sizing problems are well-known and largely studied, there is a lack of research in some real-world aspects. This thesis addresses two main characteristics at the lot-sizing context: (a) setup crossover; and (b) perishable products. The former allows the setup state of production line to be carried over between consecutive periods, even if the line is not yet ready for processing production orders. The latter characteristic considers that some products have fixed shelf-life and may spoil within the planning horizon, which clearly affects the production planning. Furthermore, two types of perishable products are considered, according to the duration of their lifetime: medium-term and short-term shelf-lives. The latter case is tighter than the former, implying more constrained production plans, even requiring an integration with other supply chain processes such as distribution planning. Research on stronger mathematical formulations and solution approaches for lot-sizing problems provides valuable tools for production planners. This thesis focuses on the development of mixed-integer linear programming (MILP) formulations for the lot-sizing problems considering the aforementioned features. Novel modelling techniques are introduced, such as the proposal of a disaggregated setup variable and the consideration of lot-sizing instead of batching decisions in the joint production and distribution planning problem. These formulations are subjected to computational experiments in state-of-the-art MILP-solvers. However, the inherent complexity of these problems may require problemdriven solution approaches. In this thesis, heuristic, metaheuristic and matheuristic (hybrid exact and heuristic) procedures are proposed. A lagrangean heuristic addresses the capacitated lot-sizing problem with setup carryover and perishable products. A novel dynamic programming procedure is used to achieve the optimal solution of the uncapacitated single-item lot-sizing problem with setup carryover and perishable item. A heuristic, a fix-and-optimize procedure and an adaptive large neighbourhood search approach are proposed for the operational integrated production and distribution planning. Computational results on generated set of instances based on the literature show that the proposed methods yields competitive performances against other literature approaches. / Problemas de planejamento da produção são de suma importância no planejamento da cadeia de suprimentos, dando suporte às decisões da transformação de matérias-primas em produtos acabados. O dimensionamento de lotes em planejamento de produção é definido pelas decisões tático-operacionais relacionadas com o tamanho das ordens de produção e quando fabricá-las para satisfazer a demanda. Os objetivos destes problemas são geralmente de cunho econômico, tais como a redução de custos ou o aumento de lucros, embora outros aspectos possam ser considerados, tais como a qualidade do serviço ao cliente e a redução dos níveis de estoque. Problemas de dimensionamento de lotes são muito comuns em atividades de produção e um planejamento eficaz de tais atividades, estabelece uma clara vantagem à empresa em relação à concorrência. Para este objetivo, é necessária a consideração de características realistas do ambiente industrial e do produto. Para a modelagem matemática do problema, estas considerações são cruciais, embora sua inclusão resulte em formulações mais complexas. Embora os problemas de dimensionamento de lotes sejam bem conhecidos e amplamente estudados, várias características reais importantes não foram estudadas. Esta tese aborda, no contexto de dimensionamento de lotes, duas características muito relevantes: (a) preservação da preparação total e parcial; e (b) produtos perecíveis. A primeira permite que o estado de preparação de uma linha de produção seja mantido entre dois períodos consecutivos, mesmo que a linha de produção ainda não esteja totalmente pronta para o processamento de ordens de produção. A ultima característica determina que alguns produtos tem prazo de validade fixo, menor ou igual do que o horizonte de planejamento, o que afeta o planejamento da produção. Além disso, de acordo com a duração de sua vida útil, foram considerados dois tipos de produtos perecíveis: produtos com tempo de vida de médio e curto prazo. O ultimo caso resulta em um problema mais apertado do que o anterior, o que implica em planos de produção mais restritos. Isto pode exigir uma integração com outros processos da cadeia de suprimentos, tais como o planejamento de distribuição dos produtos acabados. Pesquisas sobre formulações matemáticas mais fortes e abordagens de solução para problemas de dimensionamento de lotes fornecem ferramentas valiosas para os planejadores de produção. O foco da tese reside no desenvolvimento de formulações de programação linear inteiro-mistas (MILP) para os problemas de dimensionamento de lotes, considerando as características mencionadas anteriormente. Novas técnicas de modelagem foram introduzidas, como a proposta de variáveis de preparação desagregadas e a consideração de decisões de dimensionamento de lotes ao invés de decisões de agrupamento de ordens de produção no problema integrado de planejamento de produção e distribuição. Estas formulações foram submetidas a experimentos computacionais em MILP-solvers de ponta. No entanto, a complexidade inerente destes problemas pode exigir abordagens de solução orientadas ao problema. Nesta tese, abordagens heurísticas, metaheurísticas e matheurísticas (híbrido de métodos exatos e heurísticos) foram propostas para os problemas discutidos. Uma heurística lagrangeana aborda o problema de dimensionamento de lotes com restrições de capacidade, preservação da preparação total e produtos perecíveis. Um novo procedimento de programação dinâmica e utilizado para encontrar a solução ótima do problema de dimensionamento de lotes de um único produto perecível, sem restrições de capacidade e preservação da preparação total. Uma heurística, um procedimento x-and-optimize e uma abordagem por buscas adaptativas em grande vizinhanças são propostas para o problema integrado de planejamento de produção e distribuição. Resultados computacionais em conjuntos de instâncias geradas com base na literatura mostram que os métodos propostos obtiveram performances competitivas com relação a outras abordagens da literatura.

Esta dissertação trata do problema da distribuição de combustíveis com caminhões-tanque para realizar a entrega de derivados de petróleo para diversos postos de abastecimento a partir de uma base de distribuição. O problema consiste da determinação de rotas para veículos de uma frota heterogênea, visando minimizar o custo total de distribuição dos veículos envolvidos sujeitos a restrições de capacidade dos compartimentos de cada veículos. O objetivo é garantir que cada entrega seja alocada a exatamente um veículo e que todos os veículos sejam adequadamente seqüenciados. Deve-se notar que cada caminhão pode ter até seis compartimentos com diferentes capacidades. Além disso, são consideradas restrições que impedem que um veículo atenda determinado cliente. As restrições relacionadas a essa alocação de pedidos aos compartimentos dos veículos fazem esse problema tornar-se muito diferente de outros problemas de roteirização de veículos. Para ilustrar isso, uma entrega de 5.000 litros para um cliente apenas pode ser alocada em um compartimento de exatamente 5.000 litros, mas não a um compartimento maior preenchido parcialmente. Adicionalmente, caminhões do mesmo tamanho e capacidade (e.g. 30.000 litros) podem possuir diferentes números de compartimentos, inclusive de diferentes tamanhos (e.g. um caminhão de 30.000 litros pode ter 6 compartimentos de 5.000 litros ou 2 compartimentos de 10.000 litros e 2 compartimentos de 5.000 litros), tornando o problema aindamais complexo. Propõe-se inicialmente uma modelagem matemática inédita para o problema. Dada a dificuldade de resolver instâncias de tamanhos reais utilizando ferramentas comerciais de otimização como o ILOG CPLEX 9.0, foi também proposto um algoritmo heurístico que pode alcançar boas soluções em tempos curtos de processamento. Este algoritmo é inspirado em algumas idéias do GRASP. ) Ele se baseia em um método heurístico rápido de construção, que é repetidamente aplicado, baseado em um algoritmo de controle que, repedida e aleatoriamente, remove alguns pedidos da solução corrente, e então reconstrói uma nova solução a partir dos pedidos não-alocados restantes. Também são relatados resultados computacionais com diversos problemas de teste que foram gerados, considerando diferentes tamanhos de problema, bem como diferentes níveis de dificuldade de alocação de pedidos aos caminhões. / This Master\'s dissertation deals with the problem of distributing fuels by petroleum tank trucks in the context of the delivery of petroleum products to gas stations originating at a single distribution base. The problem comprises determining the vehicle delivery routes for a heterogeneous fleet, aiming to minimize the total distribution and fixed costs of the vehicles involved subject to capacity constraints for the tank compartments of each vehicle. The objective is to ensure that each delivery is assigned to exactly one truck and all trucks are properly sequenced. It should be noticed that each truck may have one to six tank compartments with different capacities eventually. In addition, there may be restrictions on which vehicles can service each client. The constraints related to the assignment of deliveries to truck compartments makes this problem much different from other vehicle routing problems, thus preventing the traditional routing approaches and formulations to be applied in this case. To illustrate this, a delivery of 5,000 liters to a single client can only be assigned to a compartment of exactly 5,000 liters, but not to a larger compartment which is not entirely filled up. In addition, trucks of the same size and capacity (e.g. 30,000 liters) may have different numbers of compartments and even different sizes (e.g. a 30,000 liters truck may have 6 compartments of 5,000 liters or 2 compartments of 10,000 liters and 2 compartments of 5,000 liters), making the problem even more complicated. We initially propose a novel mathematical IP formulation for this problem. Given the difficulty to solve instances of the same size as found in practice using off-the-shelf cutting-edge optimization tools like ILOG CPLEX 9.0, we also propose a heuristic algorithm that can reach good solutions in very short CPU times. This algorithm is inspired on some ideas of GRASP. ) It relies on a fast constructive heuristic, which is repeatedly applied, based on a control algorithm that repeatedly and randomly remove some deliveries from the current solution, and then rebuilds a new solution from the remaining unassigned and unrouted deliveries. We also report the computational results with several test problems that we have generated, considering different problem sizes, as well as different levels of difficulty related to assignment of orders to trucks.

O uso de novas tecnologias, do computador e da internet na educação viabiliza o estudo de uma nova maneira de transmissão de conhecimento. Um novo modelo de ensino-aprendizagem está se impondo no mundo inteiro. Ele é conhecido como Educação a Distância que se propõe em estender os sistemas educativos formais nos seus diferentes níveis aos setores da população que não possuem acesso aos mesmos. Com a necessidade desenvolveu-se uma pesquisa com o objetivo de criar uma ferramenta de ensino a distância (protótipo) para ensinar engenharia, mas poderia ser utilizada por cursos de qualquer área. Como exemplo e caso prático, utilizou-se uma disciplina na área de Engenharia de Produção \"Técnicas de Seqüenciamento e Programação de Operações em Máquinas\" para auxiliar no aprendizado e na aplicação das técnicas de programação da produção podendo ser aplicado na área acadêmica ou nas indústrias surgindo assim o PONet. / The use of new technologies, of the computer and internet on education make possible the study of a new way of knowledge transmission. A new model of teach-learning have been imposed worldwide. It is known as Distance Education that it considers in extending the formal educative systems in different levels to the sectors of the population that do not possess access to the same ones. With the necessity it was developed a research with the objective to create a distance education tool (archetype) to teach engineering but it could be used by courses of any area. As practical example and case it was used a discipline in the area of production engineering \"Techniques of Sequencing and Scheduling in Machines\" to assist in the learning and in the application of the programming techniques in the production, being able to be applied in the academic area or in the industries thus appearing the PONet.

O processo de tomada de decisão, em diferentes ambientes gerenciais, enfrenta um momento de mudança no contexto organizacional. Nesse sentido, Business Analytics pode ser visto como uma área que permite alavancar o valor dos dados, contendo ferramentas importantes para o processo de tomada de decisão. No entanto, a presença de dados em diferentes formatos representa um desafio. Nesse contexto de variabilidade, os dados de texto têm atraído a atenção das organizações, já que milhares de pessoas se expressam diariamente neste formato, em muitas aplicações e ferramentas disponíveis. Embora diversas técnicas tenham sido desenvolvidas pela comunidade de ciência da computação, há amplo espaço para melhorar a utilização organizacional de tais dados de texto, especialmente quando se volta para o suporte à tomada de decisões. No entanto, apesar da importância e disponibilidade de dados em formato textual para apoiar decisões, seu uso não é comum devido à dificuldade de análise e interpretação que o volume e o formato de dados em texto apresentam. Assim, o objetivo desta tese é desenvolver e avaliar um framework voltado ao uso de dados de texto em processos decisórios, apoiando-se em diversas técnicas de processamento de linguagem natural (PNL). Os resultados apresentam a validade do framework, usando como instância de demonstração de sua aplicabilidade o setor de turismo através da plataforma TripAdvisor, bem como a validação interna de performance e a aceitação por parte dos gestores da área consultados. / The decision-making process, in different management environments, faces a moment of change in the organizational context. In this sense, Business Analytics can be seen as an area that leverages the value of data, containing important tools for the decision-making process. However, the presence of data in different formats poses a challenge. In this context of variability, text data has attracted the attention of organizations, as thousands of people express themselves daily in this format in many applications and tools available. Although several techniques have been developed by the computer science community, there is ample scope to improve the organizational use of such text data, especially when it comes to decision-making support. However, despite the importance and availability of textual data to support decisions, its use is not common because of the analysis and interpretation challenge that the volume and the unstructured format of text data presents. Thus, the aim of this dissertation is to develop and evaluate a framework to contribute with the expansion and development of text analytics in decision-making processes, based on several natural language processing (NLP) techniques. The results presents the validity of the framework, using as a demonstration of its applicability the tourism sector through the TripAdvisor platform, as well as the internal validation of performance and the acceptance by managers.