L’utilisation des matériaux composites dans les structures assemblées a permis d’atteindre des niveaux de performance très élevés grâce aux propriétés spécifiques de ces matériaux. Cependant, pour garantir ces performances, les industriels s’appuient sur des gammes d’assemblage complexes afin de réduire les incertitudes et ainsi maitriser les chemins d’effort. Par exemple pour les assemblages par fixations, le contreperçage est privilégié pour limiter les défauts de positionnement des alésages et limiter les jeux. Ces gammes d’assemblage, incompatibles avec le principe d’interchangeabilité, augmentent considérablement les coûts de production.Aujourd’hui aucun outil ne permet de justifier, de remettre en cause ou d’optimiser les gammes d’assemblage en terme de compromis coût performance.Ce manuscrit propose donc une démarche de tolérancement fiabiliste des paramètres incertains pilotant le comportement des assemblages métal-composite par fixation. L’approche développée a pour but d’étudier l’effet des variabilités géométriques et matériaux sur la tenue des assemblages et de proposer une démarche pour les tolérancer. Le caractère probabiliste de la démarche, a nécessité le développement d’un modèle simplifié de comportement des assemblages par fixations afin de maitriser les temps de calcul. Ce modèle, construit à partir de surfaces rigides connectées et des éléments plaque 3-D, permet de représenter finement les phénomènes prépondérants dans le comportement des assemblages(jeu alésage fixation, contact, adhérence induit par la précharge des fixations). Il a été validé à la fois numériquement et expérimentalement. La réponse numérique de la propagation des incertitudes sur un critère de tenue mécanique est ensuite formalisée par une loi de densité de probabilité analytique. Les paramètres de cette loi sont identifiés par une stratégie combinant la méthode de Monte Carlo et la recherche des pire-cas avec un Algorithme Génétique. Cette approche permet alors de calculer les tolérances associées à un niveau de fiabilité visé. / The use of composite materials in joined structures allowed reaching highperformance level thanks to its specific material properties. However, to ensure theseperformances, a complex flow-process grid is needed in order to reduce uncertaintiesand to control therefore load distribution between fasteners. For example, with largeaeronautical structures, holes are drilled in a single operation in order to reduce holelocationerrors and clearance which increase considerably manufacturing cost.Today, no robust tool is available to allow justifying or optimizing joining process interms of both cost and joint performance.This thesis proposes therefore a reliability-based approach for the tolerancing ofuncertain parameters which could affect the behavior of fastened metal-compositejoints. The aim through the developed approach is to study the effect of geometricaland material variabilities on the performance of joints and to propose a method toprovide the optimal tolerances. The probabilistic character of the approach requiredthe development of a simplified model of fastened joint in order to reduce calculationtime. This model, based on connected rigid surfaces and continuum shell elements,allows representing finely physical phenomena in multi-material fastened joints(clearance, contact, adherence induced by tightening). This model was validatedexperimentally and numerically. The output distribution of the performance criteria,calculated after running a propagation of uncertainties, is then formalized by ananalytical probability density law. The parameters of this law are identified usingMonte Carlo method and worst-case approach based on a Genetic Algorithm. Thedeveloped approach allows therefore providing the appropriate tolerance associatedto an admissible performance criteria for the required reliability of the joint.

Este trabalho trata das falhas que podem ocorrer em uniões e juntas de materiais compósitos unidas mecanicamente por parafusos. O compósito de fibra de carbono (tecido) embutido em resina epoxy foi estudado neste trabalho devido ao amplo uso em estruturas de vários segmentos da indústria. O trabalho apresenta vários critérios de falha, demonstrando as vantagens e desvantagens de cada um para materiais compósitos. A fabricação dos corpos de provas e os ensaios necessários para obtenção de parâmetros e validação de estruturas são descritos. A resistência da junta mostrou-se muito menor do que a da estrutura de compósito, demonstrando a importância de estudos assim. Criou-se um modelo numérico utilizando critérios de falhas como o critério de Hashin e o de máxima tensão. Os resultados da simulação de elementos finitos tiveram uma boa relação com os ensaios experimentais e o modelo foi então validado e considerado representativo. / This work shows failures that can occur in composite mechanically fastened joints. The composite carbon fiber embedded in epoxy resin, used in this study, was chosen due to it\'s wide use in structures of any segment of the industry. Many failure criteria, showing the advantages and disadvantages for each, regarding composite structures are presented in this work. Test specimens\' manufacturing is described along with required tests for parameter definition and structures validation. The countersunk fastened joint strength is much lower than the composite structure itself, demonstrating the necessity of studies like this. A numerical model using criteria like Hashin and maximum stress was created. The finite elements\' simulation results had a close response to the experimental results and the model was validated and considered representative.

Este trabalho trata das falhas que podem ocorrer em uniões e juntas de materiais compósitos unidas mecanicamente por parafusos. O compósito de fibra de carbono (tecido) embutido em resina epoxy foi estudado neste trabalho devido ao amplo uso em estruturas de vários segmentos da indústria. O trabalho apresenta vários critérios de falha, demonstrando as vantagens e desvantagens de cada um para materiais compósitos. A fabricação dos corpos de provas e os ensaios necessários para obtenção de parâmetros e validação de estruturas são descritos. A resistência da junta mostrou-se muito menor do que a da estrutura de compósito, demonstrando a importância de estudos assim. Criou-se um modelo numérico utilizando critérios de falhas como o critério de Hashin e o de máxima tensão. Os resultados da simulação de elementos finitos tiveram uma boa relação com os ensaios experimentais e o modelo foi então validado e considerado representativo. / This work shows failures that can occur in composite mechanically fastened joints. The composite carbon fiber embedded in epoxy resin, used in this study, was chosen due to it\'s wide use in structures of any segment of the industry. Many failure criteria, showing the advantages and disadvantages for each, regarding composite structures are presented in this work. Test specimens\' manufacturing is described along with required tests for parameter definition and structures validation. The countersunk fastened joint strength is much lower than the composite structure itself, demonstrating the necessity of studies like this. A numerical model using criteria like Hashin and maximum stress was created. The finite elements\' simulation results had a close response to the experimental results and the model was validated and considered representative.