[en] In the frame of radioactive waste management, this work aims to study the flow around a heating wall-mounted cylinder in crossflow in URANS approach. Well-known limitations of first order turbulence models lead us to consider second order turbulence modelling. In that frame, a heat transfer model is developed and validated on academic test cases. To begin with, when mixed convection regime is dominant, these simulations, completed by an isotherm one, all performed with low-Reynolds k-w SST model, give prominence to several eddy structures registered by the bibliography. One simulation is also performed with the high-Reynolds Rij-epsilon SSG model. With the k-w SST model, the heat transfer is correctly reproduced compared to the VALIDA experiment lead by the CEA, though with the Rij-epsilon SSG model, it is strongly under-estimated. It is supposed that it comes from the use of wall functions. Subsequently, when natural convection is predominant, flow topology becomes completely different and the heat transfer becomes less accurate to the VALIDA experiment. Following Durbin's approach, the Elliptic Blending-Renolds Stress Model EBRSM, consists in accounting for wall effects, and in wall blockage in particular. Following this formalism, an Elliptic Blending-Algebraic Flux Model is developed, the EBAFM. With this model, a priori tests in the three convection regimes and then simulations on the same test cases show major improvements in flow predictions. This leads to an interesting perspective to an intermediate model between SGDH and transport equations. (author)