In a fire situation, the decrease of the material properties of a structure can significantly modify its overall behaviour. Hence, during fire tests or real fires, very large deflections can be observed on a floor without any global collapse. This highlights the activation of a large-displacement plastic upper bound mechanism called membrane action. Thus, in spite of the property loss of concrete, reinforcement steel and constructional steel of the beams connected to a reinforced concrete or composite slab, the load bearing capacity of this slab is defined as an increasing function of its vertical deflection. In practice, the behaviour of composite steel and concrete floors can be assessed with simplified or advanced models, depending on the expected level of precision. For instance, the analytical method named FRACOF enables to study a whole floor at elevated temperatures, on the basis of the Eurocodes simplified models for the behaviour of steel and concrete. With this method, the load bearing capacity of a slab can then be estimated taking account of steel profiles connected to the slab and tensile membrane action in large displacements. This analytical method has been validated against finite elements models as well as results from full scale fire tests. It applies to hot-rolled steel profiles spanning up to 20 m. However, such spans require sections with a great moment of area to limit the floor deflection in serviceability state. In order to limit the amount of steel required, cellular beams can be utilized as a practical and aesthetical solution. A finite element model for steel and composite steel and concrete cellular beams is proposed in the scope of the PhD thesis. The thermo-mechanical behaviour of steel cellular beams is modelled under Cast3M code. Composite beams are modelled combining a heat transfer calculation under Cast3M to a mechanical analysis under ANSYS. The steel beams and the reinforced or composite slab are modelled with shell elements. The shear studs are modelled with beam elements. Besides, this 3D model takes into account both material and geometrical nonlinearities. It is compared with tests results at both normal and elevated temperatures. Once validated, the model is compared to an existing analytical method in order to check the precision and the level of conservatism of the latter. Then, cellular beams are studied as part of composite steel and concrete floors in a fire situation. A full-scale natural fire test puts into evidence tensile membrane action with unprotected cellular beams, without any overall collapse. The test results are used for calibrating a 3D finite element model. This calibration relies on the temperature distribution in the different parts of the floor components, the fire resistance degree, the deformed shape and the failure modes. The model, which can reproduce the thermo-mechanical behaviour of a composite floor, is then utilized for assessing an extension proposal of the FRACOF method to composite floors made of cellular beams.
