Ce travail de thèse constitue le prolongement direct des travaux de thèse Chachuat (2001) sur l'optimisation dynamique et la commande optimale des stations de traitement de petite taille. L'objectif est d'aller plus loin en s'intéressant aux dimensionnement et fonctionnement optimaux des stations de traitement des eaux usées de toute taille. Ainsi, dans une première étape, l'optimisation des stations de traitement de petite taille a été abordée. Contrairement à ce qui a été fait jusqu'à maintenant : (i) l'aération n'est plus alternée, mais continue, (ii) le décanteur n'est plus considéré comme parfait, mais son fonctionnement est modélisé à l'aide d'une série de 10 couches de décantation, (iii) la méthode d'optimisation développée est fondée sur la méthode des sensibilités implémentée au sein du logiciel de simulation et optimisation dynamiques gProms, utilisé dans toute la thèse. L'influence du modèle du décanteur sur la minimisation de l'énergie d'aération a été particulièrement analysée. Dans une deuxième étape, les stations de traitement de grande taille sont considérées. Plus spécifiquement, le modèle benchmark développé par le réseau européen COST a été utilisé pour décrire leur fonctionnement. Un « foreignobject » a été développé pour que la simulation et l'optimisation du fonctionnement de ces stations soient possibles sous gProms. L'optimisation a notamment montré que la consommation d'énergie d'aération pouvait être réduite d'au moins de 30% par rapport au fonctionnement actuel de ces stations. Dans une troisième étape, l'optimisation du dimensionnement des stations de traitement de grande taille a été étudiée. Une superstructure a ainsi été définie avec plusieurs (cinq) réacteurs et un décanteur. Toutes les possibilités de recyclage et de court-circuit entre les réacteurs d'une part et entre les réacteurs et le décanteur d'autre part sont prises en compte. L'objectif était de déterminer la meilleure structure et les valeurs optimales des volumes des réacteurs qui permettent de minimiser le coût total tout en respectant les contraintes réglementaires sur les rejets.Par ailleurs, une optimisation multicritère de la station optimale résultante a été réalisée. Elle a permis de déterminer l'ensemble de Pareto des solutions qui minimisent la consommation énergétique (d'aération et de pompage) et maximisent la qualité de l'effluent. La quatrième et dernière partie de ce travail s'intéresse à la modélisation, simulation et optimisation de la station de traitement de Verulam près de Durban en Afrique du Sud. Des mesures expérimentales ont été réalisées sur cette station et le modèle ASM1 a été utilisé pour décrire son fonctionnement. Une analyse d'estimabilité des paramètres a été d'abord réalisée pour déterminer les paramètres du modèle qui peuvent être estimés à partir des mesures expérimentales disponibles. Les paramètres estimables ont ensuite été identifiés à l'aide de gProms. Le modèle ainsi identifié a été validé et ensuite utilisé pour optimiser le fonctionnement énergétique de cette station / This work is a direct extension of the PhD thesis of Chachuat (2001) on dynamic optimization and optimal control of small size wastewater treatment plants. The objective is to go further by focusing on optimal design and operation of wastewater treatment plants of any size. Thus, in a first part, optimization of small size wastewater treatment plants was studied. Contrary to what has been done so far: (i) the aeration is no longer alternating, but continuous, (ii) the settler is not considered perfect, but its operation is modeled using a series of 10 sedimentation layers, (iii) the optimization approach developed is based on the method of sensitivities implemented wthin the dynamic simulation and optimization software gProms, used throughout this work. The influence of the settler model on the minimization of aeration energy was particularly investigated. In a second part , the large size treatment plants are considered . More specifically, the benchmark model developed by the European network COST was used to describe their operation. A "foreign object" was developed in order to make the simulation and optimization of these plants possible using gProms. The optimisation showed that the aeration energy consumption could be reduced by at least 30 % compared to the current operation of these plants . In a third part, the optimization of the design of the wastewater treatment plant was studied. A superstructure has been defined with several (five) reactors and a settler. All the possibilities of recycling and by-passes between the reactors on the one hand and between the reactors and the settler on the other are considered. The objective was to determine the best structure and the optimal values of the reacter volumes that minimize the net present value while respecting the regulatory constraints. On the other hand, a multi-objective optimization problems of the treatment plant was carried out. It allawed to determine the Pareto set of solutions that minimize the energy consumption (pumping and aeration) and maximize the effluent quality. The fourth and last part of this work focuses on modeling, simulation and optimization of the treatment plant of the city of Verulam in the area of Durban in South Africa. Experimental measurements were carried out on the plant and the ASM1 model was used to describe its operation. An estimability analysis was first performed in order to determine the model parameters that can be estimated from the available experimental measurements . The estimable parameters were then identified using gProms . The identified model was validated and then used to optimize the energy function of this plant

Este trabalho aborda um método de otimização dinâmica baseado em modelos bioeconômicos estabelecidos na teoria do controle ótimo, que visa modelar o resultado econômico-financeiro relacionado à atividade de extração dos recursos naturais, de modo que a otimização do resultado financeiro seja controlada por uma extração sustentável desse recurso. Mais especificamente, consideramos a exploração de madeira florestal restrita a uma série de vínculos econômicos e operacionais, bem como à dinâmica de crescimento natural da floresta. Avaliando o uso efetivo dessa metodologia aplicada ao planejamento das concessões florestais e procurando contribuir com o debate a respeito da viabilidade da forma de gestão florestal baseada em concessões florestais no Brasil. / This work addresses a dynamic optimization method based on bioeconomics models established in optimal control theory, which aims to model the economic-financial result related to the activity of extraction of natural resources, so that the optimization of the financial result is controlled by a sustainable extraction of this resources. More specifically, we consider the exploration of forest wood restricted to a series of economic and operational linkages, as well as the dynamics of natural forest growth. Assessing the effective use of this methodology applied to the planning of forest concessions and seeking to contribute to the debate about the viability of forest management form based on forest concessions in Brazil.

Les terminaux à conteneurs utilisent les nouvelles technologies (EDI, RFID et GPS) pour échanger des données avec leurs partenaires, pour localiser les conteneurs et leurs équipements dans le terminal, et pour automatiser des tâches. Dans cette thèse, nous montrons comment ces informations peuvent être utilisées dans la gestion des opérations.La première partie utilise les informations sur les volumes annoncés pour affecter des ressources internes dans le but de minimiser le retard global au terminal. Nous représentons cette problématique à l'aide d'un problème de flot que nous implémentons comme programme linéaire mixte. Une étude de cas est réalisée pour un terminal du Grand Port Maritime de Marseille. En outre, nous combinons le problème d'affectation de ressources avec le dimensionnement d'un système de rendez-vous. Ceci permet de minimiser le retard global.La deuxième partie utilise les informations sur les conteneurs à retirer et leurs emplacements pour optimiser le déstockage. Le but est de retirer tous les conteneurs d'une rangée en minimisant le nombre de repositionnements parasites. Nous améliorons un modèle binaire, proposons une approche exacte de type branch and price - avec un sous-problème binaire et deux variantes d'un sous-problème énumératif - et en dérivons une approche heuristique - avec un sous-problème heuristique. L'approche exacte ne résout que les petites instances ; l'approche heuristique obtient des résultats satisfaisants mais devra être améliorée. Nous nous intéressons aussi à la version dynamique du problème où les informations sur les conteneurs à retirer arrivent petit à petit et comparons différentes stratégies de repositionnement. / Container terminals use intelligent freight technologies (e.g., EDI, RFID and GPS) to exchange data with their partners, to locate containers and equipment within the terminal, and to automate tasks. This thesis illustrated, via two examples, how this data may be used to optimize operations at the terminal.The first part uses information on announced volumes to allocate internal handling equipment. The objective is to minimize overall delays at the terminal. The problem is represented as a network flow problem and implemented as a linear mixed integer programming model. A case study for a terminal at the Grand Port Maritime de Marseille is carried out. We also showed that combining the allocation problem with the dimensioning of a truck appointment system may reduce overall delays at the terminal. The second part uses information on announced container retrievals and container positions to improve retrieval operations. The objective is to retrieve containers from a bay in a given sequence with a minimum number of parasite relocations. We improve an existing binary programming model and introduce an exact branch and price approach - with a binary subproblem and two variants of an enumerative subproblem - and a heuristic branch and price approach - with a heuristic subproblem. The exact approach solves only small instances; the heuristic approach performs well on several instances, but should be improved further. We also deal with a dynamic version of the problem where the retrieval order becomes revealed over time and evaluate different relocation strategies for this case.

Este trabalho aborda um método de otimização dinâmica baseado em modelos bioeconômicos estabelecidos na teoria do controle ótimo, que visa modelar o resultado econômico-financeiro relacionado à atividade de extração dos recursos naturais, de modo que a otimização do resultado financeiro seja controlada por uma extração sustentável desse recurso. Mais especificamente, consideramos a exploração de madeira florestal restrita a uma série de vínculos econômicos e operacionais, bem como à dinâmica de crescimento natural da floresta. Avaliando o uso efetivo dessa metodologia aplicada ao planejamento das concessões florestais e procurando contribuir com o debate a respeito da viabilidade da forma de gestão florestal baseada em concessões florestais no Brasil. / This work addresses a dynamic optimization method based on bioeconomics models established in optimal control theory, which aims to model the economic-financial result related to the activity of extraction of natural resources, so that the optimization of the financial result is controlled by a sustainable extraction of this resources. More specifically, we consider the exploration of forest wood restricted to a series of economic and operational linkages, as well as the dynamics of natural forest growth. Assessing the effective use of this methodology applied to the planning of forest concessions and seeking to contribute to the debate about the viability of forest management form based on forest concessions in Brazil.

Problèmes de tournées de véhicules statiques et déterministes ne peuvent pas être utilisés dans de nombreux systémes de la vieréelle, du fait que les données d’entrée ne sont pas fiables et sont révélées au fil du temps. Dans cette thèse, nous étudions un problème de ramassage et de livraison avec fenêtres de temps et un maximum de temps de trajet - le problème dial-a-ride - dans sa variante statique et dynamique, et nous faisons des propositions spécifiques sur les modèles d’optimisation robustes pour résoudre ce problème. Pour résoudre le modèle statique, nous développons une approche branch-and-price qui gère toutes les contraintes detemps dans le processus de création d’itinéraires de véhicules. Notre travail est axé sur les techniques de résolution du sous-problème et d’accélération pour l’approche branch-and-price. Nos résultats numériques montrent que la méthode est compétitive par rapport aux approches existantes qui sont basées sur le branch-and-cut. Dans le contexte dynamique, où certaines données d’entrée sont révélées dynamiquement ou modifiées au fil du temps, nous appliquons notre algorithme branch-and-price pour la ré-optimisation dans une approche sur horizon glissant. / Static and deterministic vehicle routing problems cannot be used in many real-life systems, as input data are not reliable and revealedover time. In this thesis, we study a pickup and delivery problem with time windows accounting for maximum ride time constraints – the so-called diala- ride problem – in its static and dynamic variant, and we make specific proposal on robust optimization models for this problem. To solve the static model, we develop a branch-and-price approach that handles ride time constraints in the process of generating feasible vehicle routes in the course of the optimization procedure. Our work is focussed on the pricing problem solver and acceleration techniques for the branch-and-price approach. Our numerical results show that the method is competitive compared to existing approaches that are based on branch-and-cut. In the dynamic context, where some input data are revealed or modified over time, we apply our branchand- price algorithm for re-optimization in a rolling horizon approach.

Shipment Consolidation (SCL) is a logistics strategy that combines two or more orders or shipments so that a larger quantity can be dispatched on the same vehicle to the same market region. This dissertation aims to emphasize the importance and substantial cost saving opportunities that come with SCL in a logistics supply chain, by offering new models or by improving on the current body of literature. Our research revolves around "three main axes" in SCL: Single-Item Shipment Consolidation (SISCL), Multi-Item Shipment Consolidation (MISCL), and Pricing and Shipment Consolidation. We investigate those topics by employing various Operations Research concepts or techniques such as renewal theory, dynamic optimization, and simulation. In SISCL, we focus on analytical models, when the orders arrive randomly. First, we examine the conditions under which an SCL program enables positive savings. Then, in addition to the current SCL policies used in practice and studied in the literature, i.e. Quantity-Policy (Q-P), Time-Policy (T-P) and Hybrid Policy (H-P), we introduce a new one that we call the Controlled Dispatch Policy (CD-P). Moreover, we provide a cost-based comparison of those policies. We show that the Q-P yields the lowest cost per order amongst the others, yet with the highest randomness in dispatch times. On the other hand, we also show that, between the service-level dependent policies (i.e. the CD-P, H-P and T-P), H-P provides the lowest cost per order, while CD-P turns out to be more flexible and responsive to dispatch times, again with a lower cost than the T-P. In MISCL, we construct dispatch decision rules. We employ a myopic analysis, and show that it is optimal, when costs and the order-arrival processes are dependent on the type of items. In a dynamic setting, we apply the concept of time-varying probability to integrate the dispatching and load planning decisions. For the most common dispatch objectives such as cost per order, cost per unit time or cost per unit weight, we use simulation and observe that the variabilities in both cost and the optimal consolidation cycle are smaller for the objective of cost per unit weight. Finally on our third axis, we study the joint optimization of pricing and time-based SCL policy. We do this for a price- and time-sensitive logistics market, both for common carriage (transport by a public, for-hire trucking company) and private carriage (employing one's own fleet of trucks). The main motivation for introducing pricing in SCL decisions stems from the fact that transportation is a service, and naturally demand is affected by price. Suitable pricing decisions may influence the order-arrival rates, enabling extra savings. Those savings emanate from two sources: Scale economies (in private carriage) or discount economies (in common carriage) that come with SCL, and additional revenue generated by employing an appropriate pricing scheme. Throughout the dissertation, we offer numerical examples and as many managerial insights as possible. Suggestions for future research are offered.

Shipment Consolidation (SCL) is a logistics strategy that combines two or more orders or shipments so that a larger quantity can be dispatched on the same vehicle to the same market region. This dissertation aims to emphasize the importance and substantial cost saving opportunities that come with SCL in a logistics supply chain, by offering new models or by improving on the current body of literature. Our research revolves around "three main axes" in SCL: Single-Item Shipment Consolidation (SISCL), Multi-Item Shipment Consolidation (MISCL), and Pricing and Shipment Consolidation. We investigate those topics by employing various Operations Research concepts or techniques such as renewal theory, dynamic optimization, and simulation. In SISCL, we focus on analytical models, when the orders arrive randomly. First, we examine the conditions under which an SCL program enables positive savings. Then, in addition to the current SCL policies used in practice and studied in the literature, i.e. Quantity-Policy (Q-P), Time-Policy (T-P) and Hybrid Policy (H-P), we introduce a new one that we call the Controlled Dispatch Policy (CD-P). Moreover, we provide a cost-based comparison of those policies. We show that the Q-P yields the lowest cost per order amongst the others, yet with the highest randomness in dispatch times. On the other hand, we also show that, between the service-level dependent policies (i.e. the CD-P, H-P and T-P), H-P provides the lowest cost per order, while CD-P turns out to be more flexible and responsive to dispatch times, again with a lower cost than the T-P. In MISCL, we construct dispatch decision rules. We employ a myopic analysis, and show that it is optimal, when costs and the order-arrival processes are dependent on the type of items. In a dynamic setting, we apply the concept of time-varying probability to integrate the dispatching and load planning decisions. For the most common dispatch objectives such as cost per order, cost per unit time or cost per unit weight, we use simulation and observe that the variabilities in both cost and the optimal consolidation cycle are smaller for the objective of cost per unit weight. Finally on our third axis, we study the joint optimization of pricing and time-based SCL policy. We do this for a price- and time-sensitive logistics market, both for common carriage (transport by a public, for-hire trucking company) and private carriage (employing one's own fleet of trucks). The main motivation for introducing pricing in SCL decisions stems from the fact that transportation is a service, and naturally demand is affected by price. Suitable pricing decisions may influence the order-arrival rates, enabling extra savings. Those savings emanate from two sources: Scale economies (in private carriage) or discount economies (in common carriage) that come with SCL, and additional revenue generated by employing an appropriate pricing scheme. Throughout the dissertation, we offer numerical examples and as many managerial insights as possible. Suggestions for future research are offered.

Container terminals use intelligent freight technologies (e.g., EDI, RFID and GPS) to exchange data with their partners, to locate containers and equipment within the terminal, and to automate tasks. This thesis illustrated, via two examples, how this data may be used to optimize operations at the terminal.The first part uses information on announced volumes to allocate internal handling equipment. The objective is to minimize overall delays at the terminal. The problem is represented as a network flow problem and implemented as a linear mixed integer programming model. A case study for a terminal at the Grand Port Maritime de Marseille is carried out. We also showed that combining the allocation problem with the dimensioning of a truck appointment system may reduce overall delays at the terminal. The second part uses information on announced container retrievals and container positions to improve retrieval operations. The objective is to retrieve containers from a bay in a given sequence with a minimum number of parasite relocations. We improve an existing binary programming model and introduce an exact branch and price approach - with a binary subproblem and two variants of an enumerative subproblem - and a heuristic branch and price approach - with a heuristic subproblem. The exact approach solves only small instances; the heuristic approach performs well on several instances, but should be improved further. We also deal with a dynamic version of the problem where the retrieval order becomes revealed over time and evaluate different relocation strategies for this case.

Orientador: Rubens Maciel Filho / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Quimica / Made available in DSpace on 2018-08-11T18:20:16Z (GMT). No. of bitstreams: 1 Pontes_KarenValverde_D.pdf: 6046859 bytes, checksum: 07c62254e493125c08f79bb23d284453 (MD5) Previous issue date: 2008 / Resumo: Para o desenvolvimento de novas resinas poliméricas ou para a melhoria da qualidade de resinas já existentes, usualmente realizam-se experimentos em escala industrial ou em planta piloto. Entretanto, tais práticas são altamente imprecisas, demoradas, caras e levam à formação de produtos fora de especificação. Ferramentas baseadas em modelos matemáticos são alternativas atraentes para contornar essas desvantagens e, portanto, o incentivo para o desenvolvimento de políticas operacionais ótimas para os reatores de polimerização tem crescido muito nos últimos anos. Neste escopo, o presente estudo aborda o projeto de resinas de polietileno com propriedades feitas sob medida, as quais compreendem não apenas as médias da distribuição de peso molecular (DPM), mas também toda a DPM, já que algumas propriedades de uso final são melhores correlacionadas com certas frações da DPM. O caso de estudo é a polimerização do eteno em solução com catalisador Ziegler-Natta, que ocorre em uma série de reatores agitados e tubulares contínuos. Devido à presença de dois tipos de reatores, diversas resinas podem ser produzidas. Inicialmente, são realizados planejamentos fatoriais tipo Plackett-Burman para determinar as variáveis mais importantes, e portanto que devem ser consideradas como variáveis de decisão na futura otimização. Além disso, planejamentos fatoriais completos e superfícies de resposta permitem uma fácil identificação de comportamentos não lineares e um melhor entendimento do processo. Dois métodos diferentes de otimização são utilizados: algoritmos baseados em SQP (sequential quadratic programming) e algoritmo genético. No primeiro método, o processo é modelado como um sistema multi-estágio no estado estacionário, de modo que métodos de otimização dinâmica ou controle ótimo são adequados para resolver o problema ao se considerar a coordenada axial do reator tubular como variável independente ao invés do tempo. Ambos os métodos de otimização mostram ótima predição das propriedades desejadas do polímero, sendo que a performance do algoritmo genético melhora quando a solução da otimização baseada em SQP é incluída na população inicial. / Abstract: In order to design new polymer grades or to improve the quality of existing polymer resins, usually pilot plant or industrial scale experiments are carried out. Such practices, though, are highly imprecise, time delayed, result in high costs and lead to off-spec products. Model based tools present an attractive alternative to overcome such disadvantages and therefore, in recent years, there is an increasing incentive to develop optimal operating policies for polymeric reactors. Whithin this scope, the present study approaches the design of polyethylene resins with tailored properties that comprehend not only average properties, but also the whole molecular weight distribution (MWD), since some end-use properties are better correlated with certain fractions of the MWD. The case study is the ethylene polymerization in solution with Ziegler-Natta catalyst, which takes place is a serie of continuous stirred and tubular reactors. Due to the presence of two types of reactors, a broad range of polymer resins can be produced. Initially, Plackett-Burman designs of experiments is carried out in order to ascertain the most important variables that should be considered as degrees of freedom for the future optimization. In addition to that, complete factorial designs and surface responses allow for an easy identification of nonlinear behaviors and a better understanding of the process. Two different methods are employed for the optimization: SQP based algorithms and genetic algorithm. For the former, the process is modeled as a multi-stage system at the steady state, in such a manner that optimal control tools are suitable to solve the problem if the axial coordinate of the tubular reactor is the independent variables, replacing time. Both methods present good prediction of the desired polymer properties and the performance of the GA improves when the solution of the SQP optimization is included in the initial population. / Doutorado / Desenvolvimento de Processos Químicos / Doutor em Engenharia Química

En 2014, dans le cadre du Plan Climat, les pays membres de l’Union Européenne, se sont engagés à réduire de près de 27% leur consommation d’énergie. L’un des axes d’études concerne l’augmentation de l’efficacité énergétique des transports urbains. Cette thèse a pour objectif de proposer une méthodologie afin de réduire la consommation énergétique de lignes de métro automatique tout en intégrant les perturbations de trafic qui se produisent dans des conditions normales d’exploitation. Le principe retenu dans ces travaux est de maximiser la réutilisation de l’énergie générée lors du freinage des trains, par les autres trains présents sur la ligne. Une première partie est dédiée à la modélisation électrique d’une ligne de métro automatique et à la présentation de méthodes permettant de calculer les flux de puissances entre les trains et les sous-stations d’alimentation. Ensuite, des algorithmes d’optimisation sont introduits pour effectuer l’optimisation des paramètres d’exploitation les plus influents dans une configuration idéale n’intégrant pas les aléas de trafic. Enfin, une méthodologie basée sur un apprentissage des données de simulation est développée dans le but de réaliser l’optimisation énergétique de la consommation en temps réel et en intégrant les perturbations de trafic. Cette dernière partie aura ainsi pour objectif de fournir une aide à la décision dans le choix des temps d’arrêts que doivent effectuer chaque train en station afin de maximiser la récupération de l’énergie issue du freinage. / In 2014, as part of the Climate Plan, EU member countries have committed to reduce by 27% their energy consumption. One of the main focal areas consists in increasing the energy efficiency of urban transports. This thesis aims to propose a methodology to reduce the energy consumption of automatic metro lines while integrating traffic disruptions that occur under normal operating conditions. The principle adopted in this work is to maximize the reuse of electrical energy generated during braking of the train, by other trains running on the line. First part is dedicated to the electrical modeling of an automatic metro line and development of methods to calculate power flows between trains and power substations. Then, optimization algorithms are introduced to perform optimization of the most influential operating parameters in an ideal configuration ignoring traffic fluctuations. Finally, a methodology based on learning simulation data is developed in order to achieve optimization of energy consumption integrating traffic disruptions in real time. This last part will thus purchase the objective to provide a decision support to determine optimal dwell times to be carried out by trains in each station, so as to maximize braking energy recovery.