Due to difficulties associated with formulating complex models and obtaining reliable estimates of unknown model parameters, modellers often use simplified models (SMs) that are structurally imperfect and that contain a smaller number of parameters. The objectives of this research are: 1) to develop practical and easy-to-use strategies to help modellers select the best SM from a set of candidate models, and 2) to assist modellers in deciding which parameters in complex models should be estimated, and which should be fixed at initial values. The aim is to select models and parameters so that the best possible predictions can be obtained using the available data and the modeller’s engineering and scientific knowledge. This research summarizes the extensive qualitative and quantitative results in the statistics literature regarding the use of SMs. Mean-squared error (MSE) is used to judge the quality of model predictions obtained from different candidate models, and a confidence-interval approach is developed to assess the uncertainties associated with whether a SM or the corresponding extended model will give better predictions. Nine commonly-applied model-selection criteria (MSC) are reviewed and analyzed for their propensities of preferring SMs. It is shown that there exist preferential orderings for many MSC that are independent of model structure and the particular data set. A new MSE-based MSC is developed using univariate linear statistical models. The effectiveness of this criterion for selecting dynamic nonlinear multivariate models is demonstrated both theoretically and empirically. The proposed criterion is then applied for determining the optimal number of parameters to estimate in complex models, based on ranked parameter lists obtained from estimability analysis. This approach makes use of the modeller’s prior knowledge about precision of initial parameter values and is less computationally expensive than comparable methods in the literature. / Thesis (Ph.D, Chemical Engineering) -- Queen's University, 2009-12-23 09:48:45.423

O comportamento estrutural de veículos de transporte coletivo, quando submetido a eventos de impacto, é de grande relevância na engenharia automobilística, pois principalmente o ônibus, vem se tornando um dos meios de transporte de elevada importância, sendo que seu uso vem aumentando a cada ano. No entanto, as estruturas de ônibus que circulam nas estradas brasileiras não são capazes de resistir, sem causar danos aos passageiros, a eventos de impacto em acidentes, o que pode ser verificado em pesquisas divulgadas pela Agência Nacional de Transportes Terrestres (ANTT) em 2007, que mostra a evolução de acidentes e vítimas envolvendo ônibus, que cada ano aumentam consideravelmente. Neste contexto, pretende-se desenvolver um absorvedor de impacto que aumente a capacidade de absorção da energia de impacto e evite a invasão da região de sobrevivência dos passageiros, e desta forma diminuindo o número de vítimas em acidentes. Serão abordados neste trabalho eventos de impacto semi-frontal, justificado pela gravidade dos acidentes desse tipo nas estradas, sendo este o tipo de acidente que deixa um maior número de vítimas, pois geralmente parte da estrutura lateral do ônibus é removida, expondo os passageiros no período de impacto. Esse efeito é conhecido como “Efeito Abridor de Latas”. O modelo numérico deste trabalho será formado por elementos unifilares, sendo a estrutura do ônibus formada por vigas flexíveis e/ ou rígidas, unidas através de juntas não lineares rotacionais e translacionais. A rigidez de cada junta não linear é obtida a partir de métodos analíticos que descrevem o comportamento do tubo de parede fina quando submetido a impacto. / The structural behavior of transportation vehicles when subjected to events of impact is of great relevance in automotive engineering, mainly because the bus has become a highly important alternative means of transportation, and its use is increasing every year. However, the structures of the buses circulating on Brazilian roads are not able to withstand strong impact, without causing passengers’ injury in events where accidents occur, which can be shown in surveys released by the National Transport Agency (ANTT) in 2007, presenting that the great incidence of accidents with victims involving buses is increasing considerably every year. In this context, the development of an impact absorber that had the capability to increase the impact energy absorption and prevent the invasion of part of bus on the region where passengers are located (reducing the number of fatalities in accidents) are the objectives of this work. In this work, a semi-frontal impact will be presented and studied due to the severity of such accident, as the side of the bus structure is removed, exposing the passengers during the impact. This is the effect known as “Can opener effect”. The numerical model of this study will consist of a single-wire element, with a bus structure formed by flexible and/or rigid beams joined by nonlinear rotational and translational joints. The rigidity of each nonlinear joint is obtained by analytical methods describing the behavior of a thin-walled tube subjected to an impact force.

O comportamento estrutural de veículos de transporte coletivo, quando submetido a eventos de impacto, é de grande relevância na engenharia automobilística, pois principalmente o ônibus, vem se tornando um dos meios de transporte de elevada importância, sendo que seu uso vem aumentando a cada ano. No entanto, as estruturas de ônibus que circulam nas estradas brasileiras não são capazes de resistir, sem causar danos aos passageiros, a eventos de impacto em acidentes, o que pode ser verificado em pesquisas divulgadas pela Agência Nacional de Transportes Terrestres (ANTT) em 2007, que mostra a evolução de acidentes e vítimas envolvendo ônibus, que cada ano aumentam consideravelmente. Neste contexto, pretende-se desenvolver um absorvedor de impacto que aumente a capacidade de absorção da energia de impacto e evite a invasão da região de sobrevivência dos passageiros, e desta forma diminuindo o número de vítimas em acidentes. Serão abordados neste trabalho eventos de impacto semi-frontal, justificado pela gravidade dos acidentes desse tipo nas estradas, sendo este o tipo de acidente que deixa um maior número de vítimas, pois geralmente parte da estrutura lateral do ônibus é removida, expondo os passageiros no período de impacto. Esse efeito é conhecido como “Efeito Abridor de Latas”. O modelo numérico deste trabalho será formado por elementos unifilares, sendo a estrutura do ônibus formada por vigas flexíveis e/ ou rígidas, unidas através de juntas não lineares rotacionais e translacionais. A rigidez de cada junta não linear é obtida a partir de métodos analíticos que descrevem o comportamento do tubo de parede fina quando submetido a impacto. / The structural behavior of transportation vehicles when subjected to events of impact is of great relevance in automotive engineering, mainly because the bus has become a highly important alternative means of transportation, and its use is increasing every year. However, the structures of the buses circulating on Brazilian roads are not able to withstand strong impact, without causing passengers’ injury in events where accidents occur, which can be shown in surveys released by the National Transport Agency (ANTT) in 2007, presenting that the great incidence of accidents with victims involving buses is increasing considerably every year. In this context, the development of an impact absorber that had the capability to increase the impact energy absorption and prevent the invasion of part of bus on the region where passengers are located (reducing the number of fatalities in accidents) are the objectives of this work. In this work, a semi-frontal impact will be presented and studied due to the severity of such accident, as the side of the bus structure is removed, exposing the passengers during the impact. This is the effect known as “Can opener effect”. The numerical model of this study will consist of a single-wire element, with a bus structure formed by flexible and/or rigid beams joined by nonlinear rotational and translational joints. The rigidity of each nonlinear joint is obtained by analytical methods describing the behavior of a thin-walled tube subjected to an impact force.

O comportamento estrutural de veículos de transporte coletivo, quando submetido a eventos de impacto, é de grande relevância na engenharia automobilística, pois principalmente o ônibus, vem se tornando um dos meios de transporte de elevada importância, sendo que seu uso vem aumentando a cada ano. No entanto, as estruturas de ônibus que circulam nas estradas brasileiras não são capazes de resistir, sem causar danos aos passageiros, a eventos de impacto em acidentes, o que pode ser verificado em pesquisas divulgadas pela Agência Nacional de Transportes Terrestres (ANTT) em 2007, que mostra a evolução de acidentes e vítimas envolvendo ônibus, que cada ano aumentam consideravelmente. Neste contexto, pretende-se desenvolver um absorvedor de impacto que aumente a capacidade de absorção da energia de impacto e evite a invasão da região de sobrevivência dos passageiros, e desta forma diminuindo o número de vítimas em acidentes. Serão abordados neste trabalho eventos de impacto semi-frontal, justificado pela gravidade dos acidentes desse tipo nas estradas, sendo este o tipo de acidente que deixa um maior número de vítimas, pois geralmente parte da estrutura lateral do ônibus é removida, expondo os passageiros no período de impacto. Esse efeito é conhecido como “Efeito Abridor de Latas”. O modelo numérico deste trabalho será formado por elementos unifilares, sendo a estrutura do ônibus formada por vigas flexíveis e/ ou rígidas, unidas através de juntas não lineares rotacionais e translacionais. A rigidez de cada junta não linear é obtida a partir de métodos analíticos que descrevem o comportamento do tubo de parede fina quando submetido a impacto. / The structural behavior of transportation vehicles when subjected to events of impact is of great relevance in automotive engineering, mainly because the bus has become a highly important alternative means of transportation, and its use is increasing every year. However, the structures of the buses circulating on Brazilian roads are not able to withstand strong impact, without causing passengers’ injury in events where accidents occur, which can be shown in surveys released by the National Transport Agency (ANTT) in 2007, presenting that the great incidence of accidents with victims involving buses is increasing considerably every year. In this context, the development of an impact absorber that had the capability to increase the impact energy absorption and prevent the invasion of part of bus on the region where passengers are located (reducing the number of fatalities in accidents) are the objectives of this work. In this work, a semi-frontal impact will be presented and studied due to the severity of such accident, as the side of the bus structure is removed, exposing the passengers during the impact. This is the effect known as “Can opener effect”. The numerical model of this study will consist of a single-wire element, with a bus structure formed by flexible and/or rigid beams joined by nonlinear rotational and translational joints. The rigidity of each nonlinear joint is obtained by analytical methods describing the behavior of a thin-walled tube subjected to an impact force.

Submitted by Geandra Rodrigues (geandrar@gmail.com) on 2018-03-27T15:52:15Z No. of bitstreams: 1 joaogabrielrochasilva.pdf: 4504745 bytes, checksum: 5d371d913693d1bf3a9425d887ab2ed1 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2018-03-27T17:49:33Z (GMT) No. of bitstreams: 1 joaogabrielrochasilva.pdf: 4504745 bytes, checksum: 5d371d913693d1bf3a9425d887ab2ed1 (MD5) / Made available in DSpace on 2018-03-27T17:49:33Z (GMT). No. of bitstreams: 1 joaogabrielrochasilva.pdf: 4504745 bytes, checksum: 5d371d913693d1bf3a9425d887ab2ed1 (MD5) Previous issue date: 2018-02-23 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / A simulação da atividade eletromecânica do coração é uma ferramenta relevante para a interpretação e estudos de medidas fisiológicas e diversos fenômenos cardíacos. Entretanto, modelos computacionais para este propósito podem ser computacionalmente custosos. Assim, são propostos neste trabalho três modelos simplificados, a nível celular, que foram capazes de reproduzir de forma quantitativa o fenômeno da contração de miócitos cardíacos. Para obtenção destes modelos um ajuste de parâmetros foi realizado via algoritmos genéticos. Os modelos propostos com parâmetros ajustados apresentaram resultados satisfatórios para reprodução da força ativa do coração com a vantagem de serem baseados em apenas duas equações diferenciais ordinárias. Além disso, o modelo final foi validado utilizando simulações envolvendo extra-sístoles, sendo capaz de reproduzir o fenômeno de alternância na força ativa. / The simulation of the heart electromechanical activity is a relevant tool for the interpretation and studies of physiological measures and various cardiac phenomena. However, computational models for this purpose may be computationally costly. Thus, three simplified models which were able to quantitatively reproduce the phenomenon of cardiac myocyte contraction were proposed in this work. At the cellular level, they were able to quantitatively reproduce the phenomenon of cardiac myocyte contraction. A parameter adjustment via genetic algorithm was performed to obtain these models. The proposed models with adjusted parameters presented satisfactory results for the reproduction of the active force of the heart with the advantage of being based on only two ordinary differential equations. In addition, the final model was validated using simulations involving extra-systoles, being able to reproduce the phenomenon of alternation in the active stress.

O trabalho trata da formulação e análise da resposta numérica de um modelo de danificação e plastificação localizadas em zonas previamente definidas ao longo de elementos estruturais de barra, estendendo-se os conceitos dos modelos clássicos de plasticidade concentrada. Inicialmente é feita uma breve revisão bibliográfica sobre modelos de dano e plasticidade, destacando-se alguns que contém propostas voltadas para a simplificação de suas formulações ou então da etapa de implementação numérica adotando-se, neste caso, a técnica de discretização em estratos dos elementos de barra. Em seguida, já no âmbito dos modelos ditos simplificados e estendendo-se a formulação de um modelo para estruturas de barras proposto na bibliografia, deduz-se a matriz de rigidez para o caso de barra em regime elástico com dano. Na dedução, considera-se que os processos de danificação, associados à flexão e restritos às extremidades do elemento de viga, são dependentes entre si e afetam diretamente os coeficientes de transmissão de esforços ao longo do elemento. Nesta etapa, através de dois exemplos numéricos, avalia-se a resposta do modelo quando implementado em um programa de elementos finitos. Apresenta-se, na seqüência, um estudo sobre a viabilidade da aplicação do modelo estudado às estruturas em concreto armado, incluindo-se casos em que a distribuição da armadura longitudinal é assimétrica. Como resultado desse estudo, sugere-se uma modificação da função critério de danificação, originalmente empregada, para melhor representar o processo de dissipação de energia relacionada à danificação. A partir dos confrontos entre respostas experimentais de vigas e pórtico em concreto armado e resultados numéricos, conclui-se pelo bom desempenho do modelo modificado. Finalmente, reúnem-se sugestões para a continuidade dos estudos. / This work is related to formulation and numerical analysis of a damage and plasticity model which considers such effects localised on zones previously defined along the structural beam elements, extending the concepts of the classics lumped plasticity models. First of all, a brief bibliography revision on damage and plasticity based models is given, enlightening the ones which propose some kind of simplification at the level of the formulation or only at the stage of the numerical implementation, for instance using a layered technique in a beam element discretization. Afterwards, on the field of the so called simplified models, a stiffness matrix of an elastic damage beam element is deduced, extending the formulation of a framed structures model proposed in the bibliography. The main characteristic of the new element is that the damage processes is localized on hinges at the extremities of the beam element, being the associated damage variables dependent on each other and reducing progressively the transmission efforts factor along the element. Two basic numerical examples show the performance of the model when implemented in a finite element code. Next, a study on the feasibility of the model to analyse reinforced concrete structures is presented. In order to enlarge the application field, cases where longitudinal reinforcement is asymmetrically distributed in the cross section are also considered. As a result of the study, a change in the damage criteria function originally used is suggested, aiming to improve the valuation of the dissipated energy related to the damage process. The response of the modified resulting model is valuated by a confront between experimental and numerical results of beams and frame reinforced concrete structures. The results show a very good performance of the modified model. Finally, some topics for further research on the theme explored in this work are suggested.

O trabalho trata da formulação e análise da resposta numérica de um modelo de danificação e plastificação localizadas em zonas previamente definidas ao longo de elementos estruturais de barra, estendendo-se os conceitos dos modelos clássicos de plasticidade concentrada. Inicialmente é feita uma breve revisão bibliográfica sobre modelos de dano e plasticidade, destacando-se alguns que contém propostas voltadas para a simplificação de suas formulações ou então da etapa de implementação numérica adotando-se, neste caso, a técnica de discretização em estratos dos elementos de barra. Em seguida, já no âmbito dos modelos ditos simplificados e estendendo-se a formulação de um modelo para estruturas de barras proposto na bibliografia, deduz-se a matriz de rigidez para o caso de barra em regime elástico com dano. Na dedução, considera-se que os processos de danificação, associados à flexão e restritos às extremidades do elemento de viga, são dependentes entre si e afetam diretamente os coeficientes de transmissão de esforços ao longo do elemento. Nesta etapa, através de dois exemplos numéricos, avalia-se a resposta do modelo quando implementado em um programa de elementos finitos. Apresenta-se, na seqüência, um estudo sobre a viabilidade da aplicação do modelo estudado às estruturas em concreto armado, incluindo-se casos em que a distribuição da armadura longitudinal é assimétrica. Como resultado desse estudo, sugere-se uma modificação da função critério de danificação, originalmente empregada, para melhor representar o processo de dissipação de energia relacionada à danificação. A partir dos confrontos entre respostas experimentais de vigas e pórtico em concreto armado e resultados numéricos, conclui-se pelo bom desempenho do modelo modificado. Finalmente, reúnem-se sugestões para a continuidade dos estudos. / This work is related to formulation and numerical analysis of a damage and plasticity model which considers such effects localised on zones previously defined along the structural beam elements, extending the concepts of the classics lumped plasticity models. First of all, a brief bibliography revision on damage and plasticity based models is given, enlightening the ones which propose some kind of simplification at the level of the formulation or only at the stage of the numerical implementation, for instance using a layered technique in a beam element discretization. Afterwards, on the field of the so called simplified models, a stiffness matrix of an elastic damage beam element is deduced, extending the formulation of a framed structures model proposed in the bibliography. The main characteristic of the new element is that the damage processes is localized on hinges at the extremities of the beam element, being the associated damage variables dependent on each other and reducing progressively the transmission efforts factor along the element. Two basic numerical examples show the performance of the model when implemented in a finite element code. Next, a study on the feasibility of the model to analyse reinforced concrete structures is presented. In order to enlarge the application field, cases where longitudinal reinforcement is asymmetrically distributed in the cross section are also considered. As a result of the study, a change in the damage criteria function originally used is suggested, aiming to improve the valuation of the dissipated energy related to the damage process. The response of the modified resulting model is valuated by a confront between experimental and numerical results of beams and frame reinforced concrete structures. The results show a very good performance of the modified model. Finally, some topics for further research on the theme explored in this work are suggested.

For aesthetic reasons and due to an increased demand for cost-effective and environmentally friendly civil engineering structures, there is a trend in designing light and slender structures. Consequently, many modern footbridges are susceptible to excessive vibrations caused by human-induced loads. To counteract this, today's design guidelines for footbridges generally require verification of the comfort criteria for footbridges with natural frequencies in the range of pedestrian step frequencies. To ensure that a certain acceleration limit is not exceeded, the guidelines provide simplified methodologies for vibration serviceability assessment. However, shortcomings of these methodologies have been identified. First, for certain footbridges, human-structure interaction (HSI) effects might have a significant impact on the dynamic response. One such effect is that the modal properties of the bridge change in the presence of a crowd; most importantly, the damping of the bridge is increased. If this effect is neglected, predicted acceleration levels might be overestimated. Second, as a running person induces a force of greater amplitude than a walking person, a single runner might cause a footbridge to vibrate excessively. Hence, the running load case is highly relevant. These two aspects have in common that they are disregarded in existing design guidelines. For the stated reasons, the demand for improvements of the guidelines is currently high and, prospectively, it might be necessary to require the consideration of both the HSI effect and running loads. Therefore, this licentiate thesis aims at deepening the understanding of these subjects, with the main focus being placed on the HSI effect and, more precisely, on how it can be accounted for in an efficient way. A numerical investigation of the HSI effect and its impact on the vertical acceleration response of a footbridge was performed. The results show that the HSI effect reduces the peak acceleration and that the greatest reduction is obtained for a crowd to bridge frequency ratio close to unity and a high crowd to bridge mass ratio. Furthermore, the performance of two simplified modelling approaches for consideration of the HSI effect was evaluated. Both simplified models can be easily implemented and proved the ability to predict the change in modal properties as well as the structural response of the bridge. Besides that, the computational cost was reduced, compared to more advanced models. Moreover, a case study comprising field tests and simulations was performed to investigate the effect of runners on footbridges. The acceleration limit given in the design guideline was exceeded for one single person running across the bridge while a group of seven people walking across the bridge did not cause exceedance of the limit. Hence, it was concluded that running loads require consideration in the design of a footbridge. / På grund av estetiska skäl och en ökad efterfrågan på kostnadseffektiva och miljövänliga konstruktioner är merparten av de gångbroar som konstrueras idag förhållandevis lätta och slanka. Med anledning av detta ökar risken för att stora svängningar uppstår på grund av dynamisk belastning från människor på bron. För att motverka att detta inträffar kräver dagens normer att komforten verifieras för gångbroar med egenfrekvenser inom området för människans stegfrekvens. Komforten verifieras genom att säkerställa att ett visst accelerationskriterium inte överskrids. För detta ändamål finns handböcker som tillhandahåller förenklade beräkningsmetoder för uppskattning av accelerationsnivåer. Brister i dessa beräkningsmetoder har emellertid identifierats. För det första kan olika typer av människa-bro-interaktion (HSI) ha en betydande inverkan på responsen hos vissa broar. Exempel på en HSI-effekt är att brons modala egenskaper förändras när människor befinner sig på bron; i huvudsak sker en ökning av brons dämpning. Om denna effekt inte tas i beaktande föreligger stor risk att överskatta förväntade accelerationsnivåer. För det andra är kraften från en löpare större än kraften från en gående person vilket gör att en ensam löpare på en gångbro kan ge upphov till accelerationsnivåer som överskrider gränsvärdena för komfort. Löpande personer är därför ett mycket relevant lastfall. Befintliga normer uttrycker inte explicit att någon av dessa aspekter bör tas i beaktande. Behovet av förbättrade riktlinjer för hur normerna bör tillämpas är därför mycket stort och i framtiden kan det bli nödvändigt att kräva att både HSI-effekter och löparlaster tas i beaktande. Därför syftar denna licentiatavhandling till att bidra till en fördjupad förståelse inom dessa två ämnen, med huvudfokus på ovan nämnda HSI-effekt i allmänhet och hur den kan beaktas på ett enkelt, noggrant och tidseffektivt sätt i synnerhet. En numerisk undersökning av HSI-effekten och dess inverkan på den vertikala responsen hos en gångbro genomfördes. Resultaten visar att HSI-effekten reducerar den maximala accelerationen och att störst reduktion erhålls då folksamlingen och bron har ungefär samma egenfrekvens och då folksamlingens massa är stor i förhållande till brons massa. Vidare utvärderades två förenklade metoder för beaktande av HSI-effekten vilka kan implementeras av konstruktörer med grundläggande kunskaper inom strukturdynamik. Det konstaterades att båda metoderna uppskattar HSI-effekten såväl som brons respons förhållandevis väl samtidigt som de reducerar beräkningstiden något jämfört med mer avancerade metoder. Effekten av löpare på gångbroar studerades genom en fallstudie med fältmätningar. Utifrån resultaten från dessa fältmätningar kunde det konstateras att accelerationsgränsen som anges i normerna överskreds när en ensam löpare sprang över bron men inte när en grupp på sju personer gick i takt över samma bro. Därför drogs slutsatsen att löparlaster bör tas i beaktande vid dimensionering av en gångbro. / <p>QC 20180320</p>

Dans l’étude de tout nouveau réacteur nucléaire, la conception de son cœur est une étape décisive. Or il s’agit d’un problème complexe, qui couple fortement la neutronique, la thermomécanique du combustible et la thermo-hydraulique. Actuellement cette conception se fait par longues itérations successives entre les différentes spécialités. Afin d’optimiser de façon plus globale et complète la conception d’un cœur, une nouvelle démarche appelée FARM (FAst Reactor Methodology) a été développée dans le cadre de la thèse. Elle consiste à établir des modèles simplifiés de neutronique, mécanique et thermo-hydraulique, sous forme analytique ou d’interpolation de calculs de codes de référence, puis à les coupler, de manière à pré-dimensionner automatiquement un cœur à partir de variables d’optimisation. Une fois ce modèle établi, on peut explorer et optimiser directement de nombreux cœurs, à partir d’algorithmes génétiques de façon à améliorer leurs performances (inventaire Plutonium en cycle, …) et leur sûreté (estimateurs de sûreté pour accidents protégés et non-protégés). Une réflexion a également due être menée pour déterminer les performances d’un cœur, ainsi que la façon de prendre en compte la sûreté. Cette nouvelle approche a été utilisée pour optimiser la conception de trois concepts de cœurs de Réacteur à Neutrons Rapides refroidi au Gaz (RNR-G). Tout d’abord, la conception du RNR-G à combustible carbure et à aiguilles en SiC a pu être optimisée. Les résultats ont permis d’une part de démontrer que le cœur de référence issu de la méthode itérative était optimal (c'est-à-dire sur le front de Pareto). D’autre part, l’optimisation a également permis de proposer de nombreux autres cœurs, où en dégradant un estimateur de sûreté ou une performance (sur lesquels des marges étaient disponibles), on améliore les autres performances. Une évolution de ce concept utilisant la nouvelle technologie du buffer, a également été modélisée dans FARM et optimisée. FARM a ainsi permis de proposer les premières images de cœur GFR carbure gainé en SiC utilisant la technologie buffer, et d’estimer leurs performances. Les résultats obtenus montrent que cette innovation permet d’atteindre des cœurs beaucoup plus performants et/ou beaucoup plus « sûrs » (plusieurs profils de cœurs étant proposés). Une troisième application de FARM a été réalisée sur un concept de GFR carbure gainé en Vanadium, où là aussi FARM a proposé les premières images de cœur. Toutefois les grandes incertitudes en jeu ne permettent pas véritablement de conclure sur les performances de ce concept, qui semble prometteur.Ainsi, la faisabilité d’une optimisation globale, couplant les différentes physiques d’un cœur de réacteur nucléaire a été démontrée. Si la méthode ainsi obtenue (FARM) est moins précise que la méthode classique, elle permet d’explorer et d’optimiser beaucoup plus rapidement (en quelques semaines au lieu de quelques mois) un grand nombre de cœurs et est parfaitement adaptée pour l’étape de préconception des cœurs de réacteurs ; d’autres études détaillées permettant ensuite d’affiner l’image de cœur retenue. / In the study of any new nuclear reactor, the design of the core is an important step. However designing and optimising a reactor core is quite complex as it involves neutronics, thermal-hydraulics and fuel thermomechanics and usually design of such a system is achieved through an iterative process, involving several different disciplines. In order to solve quickly such a multi-disciplinary system, while observing the appropriate constraints, a new approach has been developed to optimise both the core performance (in-cycle Pu inventory, fuel burn-up, etc…) and the core safety characteristics (safety estimators) of a Fast Neutron Reactor. This new approach, called FARM (FAst Reactor Methodology) uses analytical models and interpolations (Metamodels) from CEA reference codes for neutronics, thermal-hydraulics and fuel behaviour, which are coupled to automatically design a core based on several optimization variables. This global core model is then linked to a genetic algorithm and used to explore and optimise new core designs with improved performance. Consideration has also been given to which parameters can be best used to define the core performance and how safety can be taken into account.This new approach has been used to optimize the design of three concepts of Gas cooled Fast Reactor (GFR). For the first one, using a SiC/SiCf-cladded carbide-fuelled helium-bonded pin, the results demonstrate that the CEA reference core obtained with the traditional iterative method was an optimal core, but among many other possibilities (that is to say on the Pareto front). The optimization also found several other cores which exhibit some improved features at the expense of other safety or performance estimators. An evolution of this concept using a “buffer”, a new technology being developed at CEA, has hence been introduced in FARM. The FARM optimisation produced several core designs using this technology, and estimated their performance. The results obtained show that this innovative feature leads to much higher performing and/or safer cores. The FARM approach has also been applied to a GFR concept using a vanadium cladding. However the large uncertainties involved do not really enable one to evaluate the performance of this promising concept.In summary, the feasibility of a global multi-disciplinary optimization has been demonstrated. Although the resulting method (FARM) is less accurate than the conventional method, it allows fast optimization and permits a large number of cores to be explored quickly, and is ideally suited for the preliminary designs studies before further refinement of the core design.

Commercial fuels are mixtures with large numbers of components. Continuous thermodynamics is a technique for modelling fuel mixtures using a probability density function rather than dealing with each discreet component. The mean and standard deviation of the distribution are then used to model the chemical reactions of the mixture. This thesis develops the necessary theory to apply the technique of continuous thermodynamics to the oxidation reactions of hydrocarbon fuels. The theory is applied to three simplified models of hydrocarbon oxidation: a global one-step reaction, a two-step reaction with CO as the intermediate product, and the four-step reaction of Müller et al. (1992), which contains a high- and a low-temperature branch. These are all greatly simplified models of the complex reaction kinetics of hydrocarbons, and in this thesis they are applied specifically to n-paraffin hydrocarbons in the range from n-heptane to n-hexadecane. The model is tested numerically using a simple constant pressure homogeneous ignition problem using Cantera and compared to simplified and detailed mechanisms for n-heptane. The continuous thermodynamics models are able not only to predict ignition delay times and the development of temperature and species concentrations with time, but also changes in the mixture composition as reaction proceeds as represented by the mean and standard deviation of the distribution function. Continuous thermodynamics is therefore shown to be a useful tool for reactions of multicomponent mixtures, and an alternative to the "surrogate fuel" approach often used at present.