Roughness Implementation in FENSAP-ICE: Model Calibration and Influence on Ice Shapes

Introduction N OT all in- ight icing certiŽ cation conditions can be wind/icetunnel tested,  ight tested, tanker tested,or encounteredin natural icing testing. Numerical simulation, thus, complements such tests and allows the safe exploration of the complete combined  ight/icing envelopes. As such, two-dimensional and quasi-threedimensional in- ight ice accretion simulation codes have been in use by the aerospace industry as an aid to the certiŽ cation process. Numerically predicted ice shapes are manufactured from a light material and attached as disposable proŽ les on an aircraft, to investigate it for stability and control under model icing encounters. Although efŽ cient for calculating ice shapes on simple geometries, these Ž rst-generation simulation codes have limitations for complex truly three-dimensionalgeometries.Current advanced computational  uid dynamics (CFD) technologiesare quicklyovercoming these limitations. To predict ice accretion, accurate convective heat  uxes on smooth and rough surfaces are needed. In most current icing codes, convective heat  uxes are computed by solving integral boundarylayer equations and using an equivalent sand grain parameter to account for the roughness of the iced surface. The Ž nite element Navier–Stokes analysis package for ice (FENSAP-ICE) is a new three-dimensional CFD-based in- ight icing simulation system, built in a modular and interlinkedway to successivelysolve each of  ow, impingement, accretion, heat loads, and performance degradation.The codemodels the  ow based on the Euler/Navier–Stokes equationsfor the clean and degraded  ow and on new partial differential equations (PDEs) for each of the other three icing processes: The collection efŽ ciency is solved with the DROP3D module, a one-shot Eulerian method (as opposed to traditional particle-byparticle Lagrangian techniques), and ice accretion shapes are computed with ICE3D,2 a Ž nite volume method (as opposed to onedimensional control volume approaches).After much investigation of the most appropriate turbulence model for icing situations, the Spalart–Allmaras (S–A) one-equationmodel was selected for its simplicity and for the ease of implementation of roughness calculations. In this Technical Note, we Ž rst present the roughness calibration results for the model and then show the strong in uence of such roughness on ice shape predictions.