The literature dedicated to the wind-milling performance of fans reports that fair accuracy is at the reach of low-order modelling, at least for the prediction of the freewindmilling regime. However the different studies of the literature where conducted for Reynolds numbers high enough to promote an early transition of the boundary layer, and avoid massive separations on the pressure side of the rotor blades. The present paper describes an experimental analysis of the free-windmilling regime, on a low dimension fan designed for cooling applications in aircrafts. The WilLow test facility gives access to this regime for different mass-flows, i.e., different Reynolds numbers. For the highest Reynolds number configuration, it is demonstrated that the flow structure is in accordance with the theoretical expectations, and a linear evolution is observed in the loadingto-flow coefficient map. However, the low Reynolds-number configuration induces strong non-linearities, for which massive separation on the rotor blades is suspected. The flow structure is then completely recast, and the span-wise work distribution is affected. The linearity of the loading-to-flow coefficient is then lost, which results in a different value of φ p , which is no more predictable by the available low-order methods. Finally, the ability of more resolved numerical simulations to keep track of this delicate physics effect is checked.
KEYWORDSWindmilling, Fan flows, Reynolds effect NOMENCLATURE H blade height h t total enthalpy k, κ n model constants M Mach number r, r h , r s , r p radius, radius at hub, radius at shroud, zero-work radius Re Reynolds number V , W & U absolute, relative & blade velocities α flow angle in the absolute frame of reference β flow angle in the relative frame of reference ψ loading coefficient φ flow coefficient φp free-windmilling flow coefficient