Le dimensionnement des inserts pour les structures sandwich se fait usuellement par les méthodes proposées par le « Insert design handbook » de l'ESA, ou le « Military handbook 23A ». Ces méthodes analytiques basées sur la formulation développée par Ericksen en 1953 mènent à des erreurs de la charge admissible à l’arrachement d’environ ±20 % pour certains cas. Dans cette thèse, les principaux modes de rupture d’inserts sont étudiés ; l’endommagement du nid d’abeille Nomex® en cisaillement et la rupture du potting. Une analyse fine du postflambement en cisaillement du nid d’abeille Nomex® est effectuée qui permet de proposer un modèle d'endommagement à 2 paramètres.Puis, les résultats obtenus sont utilisés pour développer un modèle virtuel d’insert qui est validé par comparaison à des essais d'arrachement puis utilisé pour tracer des cartographies des modes de rupture. En utilisant cette méthode, la charge admissible à l’arrachement peut être estimée plus précisément. Cette méthode peut être une alternative face aux modèles analytiques pour le dimensionnement des inserts. Les graphiques obtenus peuvent être fournis aux ingénieurs pour le dimensionnement des inserts comme un outil qui peut réduire le temps de conception-validation. / The insert sizing for sandwich structures is usually made using the methods proposed in the “Insert design handbook” of the ESA and the “Military handbook 23A”. These analytical methods based in the in the research carried by Ericksen in 1953 could lead to errors of the pull-out allowable load prediction in the range of ±20 % for some cases.In this thesis, the principal failure modes of inserts are investigated; the core shear damage of the Nomex honeycomb core, and the potting failure. A detailed analysis of the shear postbuckling of the Nomex honeycomb core is made, allowing to propose a two variables damage model. Then, the obtained results are used to develop an insert virtual model that is validated through comparison with pull-out tests, and then used to draw failure mode maps of inserts. Using this method, the admissible pull-out load of inserts can be estimated more precisely. This method could be an alternative to using the analytical methods for the insert sizing. The resulting charts could be given to engineers as a tool for the insert sizing which could reduce the insert’s design-validation time.

Failure modes for sandwich beams of GFRP laminate skins and Nomex honeycomb core are investigated. Theoretical models using honeycomb mechanics and classical beam theory are described. A failure mode map for loading under 3-point bending, is constructed, showing the dependence of failure mode and load on the ratio of skin thickness to span length and honeycomb relative density. Beam specimens are tested in 3-point bending. The effect of honeycomb direction is also examined. The experimental data agree satisfactorily with the theoretical predictions. The results reveal the important role of core shear in a sandwich beam's bending behaviour and the need for a better understanding of indentation failure mechanism. High order sandwich beam theory (HOSBT) is implemented to extract useful information about the way that sandwich beams respond to localised loads under 3-point bending. 'High-order' or localised effects relate to the non-linear patterns of the in-plane and vertical displacements fields of the core through its height resulting from the unequal deformations in the loaded and unloaded skins. The localised effects are examined experimentally by Surface Displacement Analysis of video images recorded during 3-point bending tests. A new parameter based on the intrinsic material and geometric properties of a sandwich beam is introduced to characterise its susceptibility to localised effects. Skin flexural rigidity is shown to play a key role in determining the way that the top skin allows the external load to pass over the core. Furthermore, the contact stress distribution in the interface between the central roller and the top skin, and its importance to an indentation stress analysis, are investigated. To better model the failure in the core under the vicinity of localised loads, an Arcan- type test rig is used to test honeycomb cores under simultaneous compression and shear loading. The experimental measurements show a linear relationship between the out-of-plane compression and shear in honeycomb cores. This is used to derive a failure criterion for applied shear and compression, which is combined with the high order sandwich beam theory to predict failure caused by localised loads in sandwich beams made of GFRP laminate skins and Nomex honeycomb under 3-point bending loading. Short beam tests with three different indenter's size are performed on appropriately prepared specimens. Experiments validate the theoretical approach and reveal the nature of pre- and post-failure behaviour of these sandwich beams. HOSBT is used as a compact computational tool to reconstruct failure mode maps for sandwich panels. Superposition of weight and stiffness contours on these failure maps provide carpet plots for design optimisation procedures.