Abstract : One of the difficulties faced in turbulent flows comes from the fact that the statistics are heavily dependent on phenomena occurring in the vicinity of the walls. The presence of walls is usually accounted for by introducing damping functions in the model equations. This approach make's the model specific to a particular situation and does not favour it's universality. The present thesis aims at developing explicit algebraic models accounting for the wall effects by the elliptic blending approach developed by Manceau and Hanjalic, and at validating them in several flow configurations. The explicit algebraic methodology consists of projecting the anisotropy tensor equation onto a well chosen basis. The theory of an integrity basis enables the determination of the number of basis tensors necessary to avoid singularities in the model. Under this condition many models can be developed. Due to the accounting of the elliptic blending, an new tensor M appears in the equations in addition to the standard S and W tensors which, on the one hand, makes the theoretical framework more complex, but on the other hand, provides numerous possibilities of approximate models of practical interest. The linear and nonlinear models, validated in channel flow for a wide range of Reynolds numbers, in Couette- Poiseuille flow and in a shearless boundary layer, proved capable of correctly reproducing the near-wall anisotropy. The two-component limit is preserved. The extension of these models to 3D showed the possibility to represent correctly the anisotropy in a 3D case by using a 3-tensor truncated basis, provided that the complete set of invariants appearing in 3D are accounted for.