[pt] As operações das empresas de óleo e gás são naturalmente perigosas e suscetíveis a ocorrência de acidentes. As perdas financeiras associadas a acidentes podem ser elevadas. Para evitar esse risco, é comum que as empresas adquiram seguros. No entanto, determinar seus parâmetros adequados requer estimativas de exposição ao risco, o que ainda é uma tarefa difícil. Para lidar com essa questão, alguns autores sugerem uma caracterização de incerteza baseada em barreiras de segurança. Essa abordagem facilita a definição das consequências e também atua de forma mais preditiva quando comparada aos modelos baseados apenas em dados históricos. Um modelo de otimização é sugerido, utilizando os resultados obtidos com o método de caracterização de incerteza mencionado. Como as funções de perdas não são completamente conhecidas, de forma a resolver o problema estocástico, uma abordagem de Sample Average Approximation (SAA) é usada. Os resultados obtidos foram comparados à situação sem seguro e a outras duas opções de contrato de seguros. O modelo de otimização proposto foi o que conferiu maior previsibilidade dos valores de perdas, apresentando o menor desvio-padrão. Ressalta-se que a segunda melhor opção obteve um desvio-padrão 102 por cento a mais do que o obtido com o seguro otimizado. Além disso, o modelo também proporcionou maior proteção contra os eventos extremos, característica representada pelos menores valores de VaR e CVaR, com a segunda melhor opção apresentando um CVaR 41 por cento maior do que o obtido com o seguro otimizado. / [en] Operations in oil and gas companies are naturally dangerous and susceptible to the occurrence of accidents. The financial losses due to accident damages can be elevated. To avoid the risk of high expenses, it is usual for firms to acquire insurance. However, setting the right parameters for an insurance contract requires estimating the firm s risk exposure, which is still a hard task. To handle this issue, some authors suggest uncertainty characterization models based on safety barriers and precursor information. This approach facilitates the definition of consequences and also acts in a more predictive way when compared to usual models based only on historical data. Then, an optimization model is suggested, using the results obtained with the uncertainty characterization method mentioned as one of its inputs. As loss functions are not fully known, in order to solve the stochastic problem, a Sample Average Approximation (SAA) approach is used. The results obtained were compared to the situation where the company does not acquire insurance and to other two insurance contract options. The optimization model proposed was the one that granted greater predictability to the loss values, presenting the smallest standard deviation. The second best option presented a standard deviation 102 percent greater than the one obtained with the optimized insurance. Also, the model provided greater protection against extreme events, characteristic shown by smaller VaR and CVaR values, with the second best option presenting a CVaR 41 percent greater than the optimized model s CVaR.

Le renforcement de structures ou d'éléments de structure par collage de plats PRF (polymères renforcés de fibres) est une technique actuellement reconnue et utilisée dans le monde entier. Il permet d'augmenter la durée de vie des structures existantes ce qui est très intéressant du point de vue développement durable et est souvent plus intéressant d'un point de vue économique. La première partie de ce travail s'intéresse au renforcement de poutres béton armé par des plats PRF. En effet, ce type de renforcement peut engendrer une rupture prématurée de type peeling-off. Ce mode de ruine très fragile résulte du décollement du béton d'enrobage qui reste collé au matériau de renforcement. Pour une conception optimale d'un renforcement en flexion par collage, il est important d'être en mesure de prévoir ce type de rupture et d'en tenir compte dans le dimensionnement. Pour cela, un modèle numérique fiable de type élasto-plastique est dans un premier temps présenté qui permet de prévoir la rupture de type peeling-off. Ce modèle est validé à l'aide de résultats d'essais expérimentaux. Les paramètres principaux affectant l'efficacité du renforcement sont ensuite mis en évidence dans le cadre d'une étude paramétrique. Les résultats de cette étude sont mis en parallèle avec des résultats d'essais de la littérature prouvant ainsi l'efficacité du modèle proposé. Enfin, plusieurs mesures sont proposées pour améliorer la performance du renforcement et éviter la rupture prématurée de peeling-off. La deuxième partie de ce travail s'attache quant à elle à l'étude de renforcement de tabliers de ponts soumis aux efforts éventuels d'impact d'un véhicule sur une barrière de sécurité. Une campagne expérimentale composée de différentes configurations de dalles est d'abord réalisée. Un modèle numérique s'inspirant du modèle proposé précédemment est ensuite présenté. La confrontation des résultats expérimentaux et numériques montre une concordance encourageante avant la fissuration majeure de la dalle. Enfin, les résultats mettent en relief l'efficacité du renforcement par des plats PRF dans le cas de glissières de sécurité / Strengthening of structures by bonding FRP plates (fiber reinforced polymer) is a technique currently recognized and used worldwide. This method is a viable solution to costly replacement of deteriorating structures and increases the life of reinforced structures. The first part of this doctoral work focuses on the strengthening of reinforced concrete beams with FRP plates and more precisely on a premature failure caused by this type of reinforcement called peeling-off or concrete cover separation. This brittle failure mode which prevents the strengthened RC beams from attaining their ultimate flexural capacity involves the tearing-off of the concrete cover along the level of tension steel reinforcement starting from a plate end. The first step for a successful, safe and economic design of flexural strengthening using FRP composite at the bottom of the beam is then to predict such failure and to take it into account in design. A reliable numerical model analysis which is validated by test results is first presented to predict ultimate loading capacity and the failure mode of RC beams in a four-point bending setup. The main parameters affecting the efficiency of the reinforcement are then highlighted in a parametric study. The results of this study are compared with test results in the literature demonstrating the efficiency of the proposed model. Finally, several measures are proposed to improve the performance of the strengthening and in order to avoid the premature rupture of peeling-off. The second part of this work is concerned with the strengthening study of a bridge deck subject to eventual loads generated by a car crash into a safety barrier. A series of equivalent impact tests is first performed on deck slabs. A numerical model inspired by the previously proposed model for RC beams is then presented. Comparisons between the predictions of the numerical model and test results show a good agreement before the major cracking of the slab. Finally, the results highlight the efficiency of FRP plates in the case of safety guardrails