Ultra High Performance Fiber Reinforced Concrete (UHPFRC) are characterized by a compressive strength much higher than Very High Performance Concrete (VHPC) currently considered by standardisation, an excellent durability and the use of relatively high content of fibers. In particular, their tensile strength is quite important. Nevertheless, depending on fibers ratio and fibers types forecasted in the initial mix design and the real orientation of fibers in the structure compared with the main tensile directions, UHPFRC can exhibit either strain-softening or strain-hardening tensile behaviour. Each considered behaviour needs specific approaches in order to ensure a safe design. In a first time, characterization methods of UHPFRC tensile behaviour have been developed in order to determine which type of behaviour will occur considering a given UHPFRC and structure. These methods are based on the four point bending test. An inverse analysis of the results of this experimental method permits to deduce the “stress – strain” relationship (in the case of hardening UHPFRC) or “stress – crack opening” relationship (in the case of softening UHPFRC). The results depend on assumptions assumed during the inverse analysis. Thus, we have developed analysis methods which minimize the number of hypothesis in order to predict the most realistic behaviour law. The four point bending test configuration can involve artefacts. A comparison with direct tensile test is then necessary. In order to conduct this comparison, a direct tensile test method has been developed. It permits to use specimens with the same cross-section for direct traction and for the four point bending configuration. The results obtained from four point bending tests associated with the inverse analysis have been compared to those obtained with direct tensile tests. This comparison has been achieved using results of an experimental campaign considering different specimens sizes and two UHPFRC. Such comparison allows to highlight the effectiveness of the proposed method and particularly, its capability to deduce a strain-hardening or strain-softening behaviour of the material from observed crack patterns. In a second time, calculation methods adapted for « stress – crack opening » or « stress – strain » approaches have been tested or developed in order to predict the ultimate capacity or behaviour of UHPFRC beams submitted to a coupled shear and bending loading. Indeed, for this loading configuration, the tensile behaviour of the material is a main parameter. In order to validate the proposed calculation methods, eleven beams made of reinforced or prestressed UHPFRC, with or without stirrups and with or without fibers (metalics organics) have been tested in bending conducting to shear failure. The concomitant characterization of the UHPFRC mechanical behaviour at the “material scale”, taking into account the real orientation of fibers within the beams, constitutes an originality of this program. It has been useful to analyze the interaction between material, structure configuration and casting method on the orientation of fibers. Moreover, developed analysis methods of four point bending tests have been used to evaluate the influence of the structure (real orientation of fibres, influence of an eventual prestress or the structure configuration) on the parameters characterizing the tensile behaviour of the UHPFRC, in particular the strain corresponding to the localization of a critical crack. The conditions of additional contribution of UHPFRC and eventual stirrups in the shear capacity of the beam have been described. In order to extend the analysis, the approach based on the “stress – crack opening” relationship has been tested in order to predict the shear capacity of beams. The approach based on “stress – strain” relationship has also been applied, participating to the development and the validation of elastoplastic methods adapted to UHPFRC
