Etude des effets thermiques dans des joints collés. Application à des structures renforcées par patchs composites

Context Composite patches are often used to repair or to reinforce metallic aircraft structures. Those two materials however exhibit very di erent Coefficients of Thermal Expansions (CTE). Shear stress peaks may then appear in a bonded joint submitted to wide temperature variations during its service life. Those stress peaks may lead to the debonding or to the failure of the joint. It is then necessary to precisely evaluate the stress distribution in the bonded joints to avoid any failure. Phenomena occurring in the structure have been studied using di erent models. In addition, temperature variations cause modi cations in the assembled material properties, especially in the adhesive. Hence, a precise characterization of the mechanical behaviour of the adhesive used in patch bonding has been performed. Analytical bidirectional model Several unidirectional models have been used within the framework of linear elasticity in order to evaluate thermal stresses in a bonded joint. A bidirectional analytical model of a rectangular bonded joint previously developed for mechanical loadings has been improved to account for thermal loadings. The obtained analytical model, validated with a Finite Element calculation, has been used to compare the e ciency and strength of di erent patches made of carbon bres, bore bres or GLARE. A bidirectional e ect due to the di erence between the Poisson's ratios of the assembled materials has been clearly highlighted for thermal loadings, thus proving the relevance of a bidirectional model compared with a unidirectional one. Thermo-viscoelastic calculation In the previous models, all materials were supposed to exhibit a linear elastic response. The adhesive's shear modulus is however known to decrease with time and temperature. The adhesive has therefore been tested to characterize its thermo-viscoelastic behaviour. The resulting model has then vi Abstract been introduced in a speci c calculation model. This model has been used in various con gurations. Two main e ects have been characterized with the obtained results: the rst one is related to temperature and the other one to loading history. These e ects may have di erent consequences and it is necessary to take both of them into account for a correct bonded joint design. This calculation method is in fact an extension of some classical linear elastic models. It gives therefore the same results if the thermo-viscoelastic behaviour law is limited to its linear elastic component. Experimental testing of reinforced specimens Experimental tests have been carried out on patched specimens in order to con rm the models developed in the rst part of the work. Thermal loadings were applied to the specimens through thermal resistances while both thermal and displacement elds were measured. The displacements were obtained using the grid method: successive images of a grid bonded on the upper surface of the patch were analyzed with a speci c procedure which enabled us to calculate the displacements and strains in the patch. These tests led to the experimental evidence of the phenomena predicted by the model. Thermo-mechanical tests have nally been carried out. A mechanical loading has been applied to the specimen at di erent temperatures between 20 C and 70 C. The expected viscoelastic e ects have been observed from a qualitative point of view.Des patchs en matériaux composites sont classiquement utilisés dans l'aéronautique pour réparer ou renforcer des structures métalliques. Ces deux types de matériaux présentent cependant des coefficients de dilatation très différents. De ce fait, leur assemblage, soumis à de grandes variations de température en cours de fonctionnement, est le siège de contraintes pouvant provoquer la rupture ou le décollement du patch. Afin d'évaluer précisement les distributions de contraintes dans l'assemblage collé, différents modèles de calcul ont été proposés : d'abord un modèle analytique bidimensionnel mettant en évidence des effets liés à la géométrie plane de la structure, puis un modèle thermo-viscoélastique intégrant le comportement non-linéaire de la colle en fonction du temps et de la température. Ces modèles ont été ensuite partiellement validés expérimentalement sur des éprouvettes patchées en utilisant des méthodes de mesure de champs