Résumé : The loss of non-chemically linked water from concrete has a significant effect on its structural behaviour through a modification of its strength, Young modulus, fracture properties, creep and volume stability. The knowledge of the hydric state of concrete is therefore of upmost importance regarding the mechanical behaviour and durability potential of concrete structures. We use a non-linear diffusion equation for predicting the drying kinetics, including processes of the diffusion of water vapour and the permeation of liquid water. The multiple parameters of this model are then identified based on an in-house optimization algorithm and on a thorough experimental setup including measurements of mass loss and located relative humidity evolutions on various sample geometries. This coupled experimental and numerical approach allows identifying robust and meaningful material parameters of concrete used for drying predictions. This approach allows also identifying the mechanisms in concrete drying phenomena and processes by inverse analysis.