Experimental and numerical multi-scale approach for Sheet-Molding-Compound composites fatigue prediction based on fiber-matrix interface cyclic damage

•This work principally aims to identify the mathematical form of the evolution of the proposed fiber-matrix interface failure criterion parameters during fatigue.•To this end, an original inverse engineering methodology is proposed based on an accurate and complete experimental description of the interfacial damage, coupled to a micromechanical predictive damage model adapted to cyclic loading.•It is shown that the identified local fatigue damage criterion can be simply represented by an exponential increasing shape as a function of the gap between the current local stresses (i.e. for the considered number of cycles) and the local stresses undergone during the first applied cycle.•This expression should be considered as a representative of the fiber-matrix strain hardening fatigue effect in short fiber reinforced composites.•Thus, the identification of the criterion should be based on a simple reverse engineering using only macroscopic stiffness reduction during tensile fatigue.•Hence, there is no more need to perform a complete local damage experimental analysis. In this paper, a multi-scale approach is proposed to predict the stiffness reduction of a Sheet-Molding-Compound (SMC) composite submitted to low cycle fatigue (until 2.105 cycles). Strain-controlled tensile fatigue tests (R = 0.1) are carried out at various strain ranges. Damage is investigated at both macroscopic and microscopic scales through the evolutions of Young’s modulus and SEM observations, after interrupted fatigue tests at different lifetime periods. The results show that the fatigue degradation of the composite is mainly controlled by fiber-matrix interface debonding. A quantitative analysis allows determining the threshold and kinetics of the fiber-matrix interface damage during cyclic loading as a function of the orientation of fibers. Moreover, a fiber-matrix interface damage criterion, taking into account the local cyclic normal and shear stresses at the interface, is introduced in the Mori and Tanaka approach in order to predict the loss of stiffness. The parameters of this local criterion are identified by reverse engineering on the basis of the experimental results described above. Finally, the predicted loss of stiffness is very consistent with the experimental results.