The windmilling regime of a turbofan corresponds to a freewheeling mode of the fan rotor, driven by the ram pressure at the inlet. Early in the design process, determination of the windmilling rotational speed of the fan can be critical in the design of the supporting structure of the engine. Therefore, prediction of key parameters in windmilling is an important part of engine design. In particular, given the very high bypass ratio obtained at windmill (typically around 50), the flow in the fan stage and bypass duct is of prime interest, as it drives the establishment of the rotational speed of the low pressure spool and the overall drag. Classical CFD simulations have been shown to provide an adequate representation of the flow, but extensive parametric studies can be needed, which underlines the need for reduced-cost modeling of the flow in the engine. In this context, a body force modeling (BFM) approach to windmilling simulations is examined in the present contribution. The main objective is to assess the capability of the BFM approach to reproduce the aerodynamics of the flow in the fan rotor of a turbofan at windmill, and to propose a method to predict the rotational speed of the fan. The test case considered is a high-bypass ratio geared turbofan (the DGEN 380), which has been tested in an experimental facility designed to reproduce ground level windmilling conditions. The available global and local experimental data are used to validate the model. Furthermore, classical RANS simulations are also provided as reference simulations to assess the accuracy of the BFM results.
It is found that the overall performance of the fan is well predicted by the BFM simulations, in particular at the low rotational regime associated to windmilling. In terms of local validation, radial profiles are also found to be in good agreement, except close to the shroud. Analysis of the CFD results shows this can be traced back to massive flow separation in the rotor tip area. In terms of cost, a BFM simulation is about 80 times faster than the baseline CFD computation, making this approach very efficient in term of accuracy-to-cost ratio. Finally, assuming zero-work exchange across the rotor, a transient equation for the rotational speed is derived and included in the time-marching process to the steady state. As a result, the rotational speed of the fan becomes an output of the simulations. The rotational speed predicted by the present model shows good agreement with engine experimental data. However, as only the rotor is modeled, the internal losses are not fully accounted for, and the massflow has to be specified from the experimental data. Further improvement of the approach will consist in modeling the stator and the complete secondary duct so that the loss, and therefore the massflow, can be predicted.
Thermodynamic Properties of La1-xSmxCoO3
