O estudo de sistemas logísticos envolve a concatenação de elementos estratégicos e operacionais, comumente compondo sistemas com múltiplas facetas, objetivos antagônicos e grande número de alternativas. Nesse contexto, o presente trabalho discute a utilização de análise de decisão multicritério (MCDA), simulação de eventos discretos (SED) e otimização para simulação. A metodologia MCDA captura, mensura e pondera os objetivos e valores dos tomadores de decisão. Por sua vez, a SED representa o sistema estudado com alto nível de detalhamento, permitindo a avaliação de diversas configurações do sistema. Por fim, métodos de otimização para simulação possibilitam a busca e comparação de alternativas mais eficientes. As três metodologias são avaliadas, identificando suas vantagens, desvantagens e complementaridades quando aplicadas a sistemas logísticos. Através da aplicação de um estudo de caso sobre o dimensionamento de um sistema de transporte, constatou-se que: a) a SED incorporou detalhes importantes para a avaliação mais precisa de vários indicadores de desempenho b) a metodologia MCDA possibilitou a captura de vários objetivos e valores, propiciando a realização de tradeoffs robustos; c) um método de busca exaustiva e técnicas de redução de variância permitiram a comparação das alternativas em tempos computacionais reduzidos. Por fim, conclui-se que a metodologia híbrida apresentada expande o potencial de aplicação da SED em sistemas logísticos complexos. / A logistic system study involves strategic and operational elements, commonly composing multi-faceted systems with antagonistic goals and large number of alternatives. In this context, this thesis discusses the use of multi-criteria decision analysis (MCDA), discrete event simulation (DES) and optimization for simulation. The MCDA methodology captures, measures and weighs the goals and values of decision makers. DES is useful for representing systems with high level of detail, allowing the evaluation of several system configurations. Finally, optimization for simulation procedures are useful for searching and comparing more efficient alternatives. These three methodologies are assessed and their advantages, disadvantages, and complementarities are identified for logistics systems applications. Through a case study of a transportation system, we conclude that: a) the SED incorporated important details for more precise evaluation of various performance indicators b) the MCDA methodology was useful to capture several goals and values, so that robust tradeoffs could be carried out c) an exhaustive search routine and variance reduction techniques allowed the comparison of several alternatives in feasible computational times. Finally, we conclude that the presented hybrid methodology expands the application of DES to complex logistics systems.

L'étude présentée dans ce mémoire concerne le diagnostic des machines électriques à l'aide d'une association innovante entre la modélisation à évènements discrets, la Simulation Comparative et Concurrente (SCC) et les Réseaux de Neurones Artificiels (RNAs). Le diagnostic des machines électriques est effectué à partir d'une analyse temporelle des signaux statoriques et rotoriques à l'aide de réseaux de neurones de type Feed-Forward. Afin de comparer les différentes configurations de ces réseaux de neurones, l'approche proposée dans ce document utilise la simulation comparative et concurrente implémentée grâce au formalisme à évènements discrets DEVS (Discrete EVent system Specification). L'intégration des algorithmes de la SCC et des RNAs au sein du formalisme DEVS a été effectuée de manière générique et indépendamment du simulateur en développent des extensions et une librairie de modèles dans l'environnement de modélisation et de simulation à évènements discrets DEVSimPy. L'application de cette nouvelle solution pour le diagnostic des machines électriques permet de détecter les défauts à partir d'une architecture logiciel facilement portable sur des systèmes embarqués de type FPGA. / This thesis deals with the time-domain analysis of the electrical machines fault diagnosis due to early short-circuits detection in both stator and rotor windings. It also introduces to the Discrete EVent system Specification (DEVS) a generic solution to enable concurrent and comparative simulations (CCS). The DEVS-based CCS is an extension introduced using an aspect-oriented programming (AOP) to interact with the classic DEVS simulator. A new DEVS-based artificial neural network (ANN) is also introduced with a separation between learning and calculation models. The DEVS-based CCS is validated on the proposed ANN DEVS library inside the DEVSimPy environment. The concurrent ANN contributes in the time-domains analysis for the electrical machine fault diagnosis. This new method is based on data coming directly from the sensors without any computation but with a new dedicated preprocessing technique. Later, some enhancements are brought to the artificial neural network based on a new multistage architecture reducing the training time and errors compared to the single ANN. The new architecture and techniques has been validated on real data sixteen non-destructive windings faults analysis and localization.

A simulação de eventos discretos é uma das mais poderosas ferramentas disponíveis para modelagem e análise de problemas. Freqüentemente ela é utilizada para o planejamento, projeto e controle de sistemas complexos em diversos segmentos; entre eles, a área de sistemas logísticos, que a tem aplicado extensivamente para analisar as capacidades logísticas, avaliar os impactos de mudanças nos processos produtivos, testar diferentes configurações na alocação de recursos limitados, determinar parâmetros das operações, validar novos projetos, entre outras aplicações. Esta popularização da simulação despertou o interesse das empresas especializadas no desenvolvimento de softwares, resultando em diversas opções com diferentes características e custos. Desta forma, identificou-se a necessidade de um método para seleção de softwares de simulação de eventos discretos voltados à resolução de problemas logísticos, pois os trabalhos disponíveis na Literatura abordam somente a escolha de softwares para fins genéricos. Para preencher essa lacuna, foi proposto um conjunto de critérios que endereçam as necessidades específicas do desenvolvimento de modelos voltados à análise de sistemas logísticos. Esses critérios formaram a base do método proposto, que incorporou as melhores características dos trabalhos identificados na Literatura. Sua primeira fase gerou uma lista com os softwares que suportaram os critérios definidos como essenciais por meio da aplicação de um método de atribuição de pesos e pontuação, cuja eficiência e simplicidade são amplamente reconhecidas; já a segunda fase, utilizou o método de análise hierárquica (AHP), recomendado para a tomada de decisão baseada em múltiplos critérios, por permitir a identificação e correção das inconsistências geradas pelo julgamento humano. Dois problemas testes também foram propostos para o desenvolvimento de aplicações em sistemas logísticos e condução de experimentos durante a segunda fase. Os softwares Arena, ProModel e @Risk, extensivamente utilizados para ensino e pesquisa na Escola Politécnica da Universidade de São Paulo, foram avaliados durante a aplicação prática do método proposto. Os resultados obtidos confirmaram a percepção, adquirida durante a segunda fase do método, de que os softwares Arena e ProModel apresentam características semelhantes, dada a discreta diferença em suas pontuações totais. Os diversos cenários desenvolvidos na análise de sensibilidade também ratificaram essa tendência. / The discrete event simulation is one of the most powerful tools available for modeling and problem analysis. Frequently it is used for planning, designing and controlling complex systems in several segments. The logistic system area has also applied it extensively to analyse the logistic capacities, evaluate the impacts of changes in the productive processes, test different configurations in the allocation of limited resources, determine operation parameters, validate new projects, among other applications. Thus, the growth of the simulation awoke the interest of the companies specialized in software development. This fact resulted in several options with different characteristics and costs. Because of this, it was identified the necessity of a selection method for discrete event simulation software dedicated to the resolution of logistic problems. The available works in Literature only approach the choice of softwares for generic purposes. To fill in this gap, a set of criteria that address the specific necessities of the development of models dedicated to the analysis of logistic systems was proposed. These criteria formed the base of the proposed method. It has incorporated the best characteristics of the works identified in Literature. Its first phase generated a list with softwares with widely recognized efficiency and simplicity that supported the essential criteria through the application of a scoring model. Its second phase used the Analytic Hierarchy Process (AHP) which is recommended for the decision taking based on multiple criteria because it allows the identification and correction of the inconsistencies generated by the human judgment. Two test problems have also been proposed for the development of applications in logistic systems and performance of experiments during the second phase. Arena, ProModel and @Risk, extensively used for education and research in the Polytechnical School of the University of São Paulo, were evaluated during the practical application of the proposed method. The results confirmed the perception acquired during the second phase of the method that Arena and ProModel present similar characteristics. They presented a discrete difference in their total score. The several scenes developed in the sensitivity analysis also ratified this trend.

This project has been initiated in cooperation with a Swedish manufacturing company. Due to increased demand and competition, the company wants to streamline its production process, increase the degree of automation and visualize specific workflows. By creating a frame of reference and a literature review, a theoretical basis for methods and concepts which has been utilized throughout the project has been obtained. With the help of the identified methods and methodologies, a current state analysis was performed. Using traditional Lean tools such as Genchi genbutsu, Ishikawa diagram and a 5-why analysis, in combination with time studies and interviews, the current state of the studied system was successfully mapped and analysed. With the help of the current state analysis, which served as a conceptual model, a simulation model of the current state was created in order to handle the large variety and the complexity of the system. The simulation model was validated and verified in order to ensure that it was “good enough” for the purpose of this project in the depiction of the real world system. During the experimental design, several improvement suggestions were created by utilizing methods such as brainstorming, Ishikawa diagram and a 5-why analysis. In a Kaizen event, onsite personnel had the opportunity to decide which suggestions that was fit for experimentation using simulation. With the result of the Kaizen event, experiments were performed in order to evaluate the proposed improvement suggestion. As a result, several new insights regarding improvements could be obtained, which provided several suggestions for an improved future state. Including a proposed automated cell. The analysis of the results did not entirely satisfy the aim of the project since certain factors could not be analysed, therefore the authors recommend that further studies are needed if proposed improvement suggestions are to be implemented.

Realizar a análise econômica de uma produção agropecuária não é algo trivial, seja pela enorme heterogeneidade entre cada unidade produtiva, seja porque utiliza muitos recursos naturais, alguns de difícil mensuração. Há diversos métodos disponíveis para o cálculo do custo de produção, que é o indicador chave para a análise da viabilidade de um empreendimento. O desafio é permitir que a projeção da atividade no horizonte produtivo seja a mais real e dinâmica possível. A simulação computacional é atualmente uma das mais poderosas ferramentas de análise disponível para o planejamento, projeto e controle de sistemas complexos, e vem sendo cada vez mais utilizada e difundida. Simular compreende a elaboração de um método de experimentação que, através da construção de modelos de um sistema real, procura descrever comportamentos, construir teorias ou hipóteses por meio do observado e predizer comportamentos futuros. O uso de modelos de simulação, que incorporem o risco e a probabilidade dentro produção animal, pode ser uma alternativa tanto técnica - auxiliando na tomada de decisão, gestão e planejamento pecuário, quanto científica - permitindo a avaliação de efeitos de resultados da pesquisa e identificação de limitantes que podem incentivar o desenvolvimento de pesquisas futuras. O modelo de simulação híbrido - baseado em simulação de eventos discretos e baseado em agentes, aqui proposto - visou identificar coeficientes zootécnicos e os critérios de manejo que mais impactam a produção de ovinos de corte. Esse modelo de simulação computacional híbrido possui caráter dinâmico e probabilístico, com eventos marcados no tempo (estação de monta, gestação, parição, desmame, engorda, abate, entre outras) e complexos o suficiente para que seus agentes sofram alterações tanto no tempo quanto em resposta a ocorrência ou não das variáveis ligadas a eles. Os resultados dos experimentos e dos cenários estudados revelaram que dentre os índices zootécnicos das matrizes, a ocorrência de aborto gera maior impacto na taxa de abate e financeiramente na margem líquida operacional. Contudo, analisando as variáveis tanto para matrizes como para cordeiros, a mortalidade neonatal, até os cinco dias de vida dos cordeiros, demonstrou ser o ponto fundamental para a lucratividade da atividade. A estabilização do rebanho foi mais afetada pela presença de matrizes adultas no plantel, que culminaram no aumento no número de cordeiros por matriz. Os resultados da análise dos fluxos de caixa de 30 anos, a uma taxa mínima de atratividade de 6,17% aa, permitiram constatar que iniciar a atividade com número muito reduzido de matrizes é o cenário menos interessantes, com taxa interna de retorno negativa para o período de análise. O estudo do fluxo de caixa permitiu atribuir a 200 matrizes como o tamanho de rebanho inicial mais interessante (TIR = 3,30% aa). Assim, o uso de simuladores híbridos baseados em simulação de eventos discretos e baseados em agentes, para estudos na pecuária nacional, apresentou-se como ferramenta com grande potencial de contribuição, no sentido de permitir conhecer os resultados possíveis das diferentes combinações tecnológicas disponíveis. O modelo ainda permite ser utilizado como ferramenta de estudo e análise para a cadeia produtiva, contribuindo na orientação aos cientistas, auxiliando no direcionamento de seus esforços no desenvolvimento de futuras pesquisas / Conduct an economic analysis of agricultural production is not trivial, either by its enormous heterogeneity between each production unit, or because it uses many natural resources, some of which are difficult to measure. There are several methods available for calculating the cost of production, which is the key indicator for assessing the feasibility of a project. The challenge is to allow the projection of activity in the productive horizon in the most real and dynamic form as possible. The computer simulation is currently one of the most powerful analysis tools available for planning, design and control of complex systems and is being increasingly used and disseminated. Simulate includes the development of a method of testing by building models of a real system, that seeks to describe behaviors, build theories or hypotheses through noted and predicted future behaviors. The use of simulation models that incorporate uncertainty and probability in animal production can be both an alternative technique - assisting in decision-making, management and livestock planning; as scientific - allowing the evaluation of the research results of effects and identification of limiting that may encourage the development of future research. The model of hybrid simulation - based on discrete event simulation and on agent-based, proposed here - aimed to identify factors husbandry and management criteria that most affect the production of meat sheep. The model of hybrid computer simulation have dynamic and probabilistic characteristics, with events scheduled in time (breeding season, pregnancy, parturition, weaning, fattening, slaughter, etc.) and enough complexity that its agents be adversely affected both in time and in response to the occurrence or not of variables linked to them. The results of the experiments and the scenarios studied showed that among the zootechnical indexes of the sheep, the occurrence of abortion generates a greater impact on the slaughter rate and financially in the net operating margin. However, analyzing the variables for both sheep and lambs, neonatal mortality, up to the lambs\' five days of life, proved to be the fundamental point for the profitability of the activity. The stabilization of the herd was more affected by the presence of adult sheep in the herd, which culminated increasing the number of lambs per sheep. The results of the analysis of the 30-year cash flows, at a minimum attractiveness rate of 6.17% per year, showed that starting the activity with a very small number of sheep was the least interesting scenario, with a negative internal rate of return for the analysis period. The study of the cash flow allowed to assign to 200 sheep as the most interesting initial herd size (IRR = 3.30% per year). Thus, the use of hybrid simulators, based on discrete event simulation and agent-based for studies in national livestock, is presented as a tool with great potential to contribute, to allow knowing the possible outcomes of different combinations of available technology. The model also allows to be used as a study tool and assessment of different technological combinations for the production chain, contributing to the guidance of scientists, assisting their efforts in the development of future research

The goal of the thesis is to develop tools for improving shipbuilding workshops productivity. These tools have been tested on workshops of Aker Yards France shipyard, in Saint-Nazaire. The main chapter concerns the modelling of a workshop with the discrete-event simulation methodology. Production simulation is strongly used in some fields as automobile industry but is less used in shipbuilding. Indeed, the difficulty is that almost each piece to produce is unique. The objective is to show the importance of production simulation for such workshops. The modelled workshop is linked to a genetic algorithm to improve its productivity by optimizing its production sequence. Secondly a tool of creation and optimization of PERT networks has been created. The tool contains a graphical interface to easily handle networks and an optimization algorithm. This last one can minimize total cost of any project for a given total time. The tool has been linked to the modelled workshop. Consequently we get a PERT diagram of the workshop that can highlight critical activities. Finally, the last chapter explains the development of a tool to solve space allocation problems. Indeed space is often a critical point in many shipyards. Thanks to a user-friendly interface, the developed tool facilitates scheduling of these workshops. Furthermore it contains an optimization algorithm based on a heuristic approach to automate the planning process. In conclusion a set of efficient tools has been carried out with concrete applications on real workshops./Lobjectif de la thèse est de développer une série doutils qui permettent doptimiser la production dateliers de construction navale. Ces outils sont testés sur des ateliers du Chantier Naval Aker Yards France à Saint-Nazaire. Le chapitre principal traite de la modélisation dun atelier dit de PréPréFabrication sur le principe de la simulation de production à évènements discrets. La simulation de production est fortement utilisée dans certains domaines industrie automobile par exemple mais encore peu souvent en construction navale. En effet, la difficulté réside dans la diversité des pièces à produire. Lobjectif est de démontrer lintérêt de la simulation de production même dans ces ateliers où pratiquement chaque pièce à construire est unique. Latelier modélisé a été couplé avec un algorithme génétique pour améliorer sa productivité en optimisant la séquence de production. En second lieu, un outil de création et doptimisation des réseaux PERT a été créé. Loutil contient une interface graphique permettant une manipulation aisée des réseaux et un algorithme doptimisation. Ce dernier permet de minimiser le coût de réalisation dun projet pour un temps total fixé. En étant couplé avec le logiciel de simulation développé, une création automatique du diagramme PERT de latelier est créée, permettant de mettre clairement en évidence ses activités critiques. Finalement, le dernier chapitre de la thèse porte sur la création dun logiciel permettant de résoudre les problèmes de gestion de surface et de planification rencontrés dans certains chantiers. En effet, lespace au sol est un élément critique pour beaucoup dateliers de construction navale. Le logiciel développé, grâce à une interface conviviale et pratique, permet de faciliter grandement la planification. De plus, il contient un algorithme doptimisation basé sur des heuristiques afin dautomatiser le processus de planification. En conclusion, un ensemble doutils efficaces a été généré, avec des applications concrètes sur des ateliers réels.

The objective of this dissertation is to introduce a unified framework for modeling and simulating discrete event logistics systems (DELS) by using a formal language, the System Modeling Language (SysML), for conceptual modeling and a corresponding methodology for translating the conceptual model into a simulation model. There are three parts in this research: plant modeling, control modeling, and simulation generation. Part 1:Plant Modeling of Discrete Event Logistics Systems. Contemporary DELS are complex and challenging to design. One challenge is to describe the system in a formal language. We propose a unified framework for modeling DELS using SysML. A SysML subset for plant modeling is identified in this research. We show that any system can be described by using the proposed subset if the system can be modeled using finite state machines or finite state automata. Furthermore, the system modeled by the proposed subset can avoid the state explosion problem, i.e., the number of the system states grows exponentially when the number of the components increases. We also compare this approach to other existing modeling languages. Part 2:Control Modeling of Discrete Event Logistics Systems. The development of contemporary manufacturing control systems is an extremely complex process. One approach for modeling control systems uses activity diagrams from SysML, providing a standard object-oriented graphical notation and enhancing reusability. However, SysML activity diagrams do not directly support the kind of analysis needed to verify the control model, such as might be available with a Petri net (PN) model. We show that a control model represented by UML/SysML activity diagrams can be transformed into an equivalent PN, so the analysis capability of PN can be used and the results applied back in the activity diagram model. We define a formal mathematical notation for activity diagrams, show the mapping rules between PN and activity diagrams, and propose a formal transformation algorithm. Part 3:Discrete Event Simulation Generation. The challenge of cost-effectively creating discrete event simulation models is well-known. One approach to alleviate this issue is to describe a system using a descriptive modeling language and then transform the system model to a simulation model. Some researchers have tried to realize this idea using a transformation script. However, most of the transformation approaches depend on a domain specific language, so extending the domain specific language may require modifying the transformation script. We propose a transformation approach from SysML to a simulation language. We show that a transformation script can be independent of the associated domain specific language if the domain specific language is implemented as domain libraries using a proposed SysML subset. In this case, both the domain library and the system model can be transformed to a target simulation language. We demonstrate a proof-of-concept example using AnyLogic as the target simulation language.

Development of a tool for discrete event simulations in forest technology, dependent onspatial components, has successfully been initialized in this thesis project. These simulations may be used to optimize the way the forest is used and to evaluate new machine concepts in forestry. The Python library for discrete event simulation, SimPy, was chosen as the foundationfor the tool. The developed tool can handle spatial objects such as moving machines, trees and boulders. Support for continuous linear movements was also added, which has resulted in a model that partially overlaps continuous and discrete event simulations without any additional computational costs. The result is simulations of a forest with machines operating in it. Two pilot simulations,one of a thinning machine and one of a planting machine, were performed with useful results. The new simulation tool shows promising properties. Limitations and improvements arediscussed, with the conclusion that continued development is recommended. / I det här examensarbetet har ett verktyg för diskret event simuleringar inom skogsteknikutvecklats. Sådana simuleringar kan användas för att optimera arbete inom skogsbruk samtför att utvärdera nya maskintekniska koncept. Programmeringsspråket Pythons bibliotek för diskret event simulering, SimPy, valdessom grund för verktyget. Det utvecklade verktyget kan hantera spatiala objekt såsom träd,stenar och stubbar. Stöd för kontinuerliga rörelser har även utvecklats, vilket resulterat i enmodell som delvis överlappar kontinuerliga simuleringar och diskret event simuleringar.Resultatet är kvalitativa simuleringar av skog och skogsmaskiner. Två lyckade simule-ringsstudier har gjorts för Sveriges Lantbruksuniversitet, en simulering av en gallringsmaskinoch en simulering av en planteringsmaskin. Det nya simuleringsverktyget uppvisar lovande egenskaper, även om det har sina be-gränsningar. Fortsatt utveckling av verktyget rekommenderas starkt.

In the present fierce global competition, poor responsiveness, low flexibility to meet the uncertainty of demand, and the low efficiency of traditional assembly lines are adequate motives to persuade manufacturers to adopt highly flexible production tools such as cross-trained workers who move along the assembly line while carrying out their planned jobs at different stations [1]. Cross-trained workers can be applied in various models in assembly lines. A novel model which taken into consideration in many industries nowadays is called the linear walking worker assembly line and employs workers who travel along the line and fully assemble the product from beginning to end [2]. However, these flexible assembly lines consistently endure imbalance in their stations which causes a significant loss in the efficiency of the lines. The operational time variability is one of the main sources of this imbalance [3] and is the focus of this study which investigated the possibility of decreasing the mentioned loss by arranging workers with different variability in a special order in walking worker assembly lines. The problem motivation comes from the literature of unbalanced lines which is focused on bowl phenomenon. Hillier and Boling [4] indicated that unbalancing a line in a bowl shape could reach the optimal production rate and called it bowl phenomenon.  This study chose a conceptual design proposed by a local automotive company as a case study and a discrete event simulation study as the research method to inspect the questions and hypotheses of this research.  The results showed an improvement of about 2.4% in the throughput due to arranging workers in a specific order, which is significant compared to the fixed line one which had 1 to 2 percent improvement. In addition, analysis of the results concluded that having the most improvement requires grouping all low skill workers together. However, the pattern of imbalance is significantly effective in this improvement concerning validity and magnitude.

Active queue management (AQM) techniques for congestion control in Internet Protocol (IP) networks have been designed using both heuristic and analytical methods. But so far, there has been found no AQM scheme designed in the realm of stochastic optimization. Of the many options available in this arena, the gradient-based stochastic approximation method using Infintesimal Perturbation Analysis (IPA) gradient estimators within the Stochastic Fluid Model (SFM) framework is very promising. The research outlined in this thesis provides the theoretical basis and foundational layer for the development of IPA-based AQM schemes. Algorithms for computing the IPA gradient estimators for loss volume and queue workload were derived for the following cases: a single-stage queue with instantaneous, additive loss-feedback, a single-stage queue with instantaneous, additive loss-feedback and an unresponsive competing flow, a single-stage queue with delayed, additive loss-feedback, and a multi-stage tandem network of $m$ queues with instantaneous, additive loss-feedback. For all cases, the IPA gradient estimators were derived with the control parameter, $ heta$, being the buffer-limits of the queue(s). For the single-stage case and the multi-stage case with instantaneous, additive loss-feedback, the IPA gradient estimators for when the control parameter, $ heta$, is the loss-feedback constant, were also derived. Sensitivity analyses and optimizations were performed with control parameter, $ heta$, being the buffer-limits of the queue(s), as well as the loss-feedback constant.