3-D computational fluid dynamics simulation of in-flight icing (3-D CFD-icing) has achieved considerable advances in the last decade , and many dynamic OEMs and second tier suppliers are using them to speed icing certification. Yet, others remain on the fence, using technologies from three decades ago.
While the advanced 3-D CFD-aero codes these same OEMs use to analyze and optimize the aerodynamic performance of their aircraft and engines bear little resemblance to what was available 30 years ago, they are content to use pseudo-CFD tools for icing that have literally remained frozen in time (2-D or quasi-3-D, incompressible, panel methods, no viscosity, no turbulence, developed in the 1980s, etc.) because, in their eyes, new CFD-Icing tools lack validation. This unfortunate hesitation to use the same advanced tools for icing, as they already use for their aerodynamics, comes from the fear of delaying the airworthiness granting process.
Some of the pseudo-CFD-icing codes still in use today are based on 2-D panel methods for the airflow and Lagrangian tracking techniques for impingement. Panel methods cannot be considered true CFD, since a crucial characteristic of CFD is that a progressive refinement of the mesh must yield an asymptotic improvement in the results. Panel methods, being based on the method of singularities, instead of converging to an asymptotic solution will literally blow up if the surface mesh is over-refined, thus failing the litmus test of being labeled a CFD method.
In addition, the fact that 2-D computational codes are in the end judged against 2-D icing tunnel ice shapes is an incentive for an ingenious process called “code calibration”. Having identified ice surface roughness as the most important parameter governing the growth of ice shapes, developers extract from hundreds of icing tunnel measurements of airfoil models a “heuristic roughness model that will make the pseudo-CFD-Icing computer code agree with the icing tunnel result each time!” Pseudo-CFD-icing codes are then incorrectly labeled “validated”, based on the fact that they yield similar results to the icing tunnel. In truth, they are not “validated” by any measure, but are simply “calibrated”. In essence, a calibrated code will regurgitate to the user, through a fancy computational procedure, a result that even if not known a priori is nevertheless “preordained”. The analogy would be a video game in which all the possible parametric changes and their results are already stored: a player cannot introduce a new geometry, a new maneuver; the game has not been programmed for that. One can never get a result that has not already been programmed in!
Industry needs CFD-Icing Codes that are Truly Predictive and Not Simply Calibrated
The only way to numerically predict ice shapes for the wide range of aerospace assemblies and components (wings, empennages, engine inlets, nacelles, sensors, probes, all the way up to the complete aircraft), is by developing analytical roughness models that predict ice surface roughness in space (varying all over the 3D body) and in time (roughness is time dependent, increasing asymptotically to a local value as ice accretes). Only in this way can a CFD-Icing code be truly predictive and not simply calibrated.
The new 3-D CFD-Icing tools permit a more efficient and safer icing certification methodology for all types of aircraft by reducing the likelihood of ice-induced hazardous events in service. While dry and icing (wet) tunnel testing, flight testing with artificial ice shapes and flying in natural icing conditions will always play a significant role, the availability of advanced simulation tools can shorten the certification process, fill important gaps, focus or entirely eliminate icing tunnel for ice shapes, predict what will eventually be seen in natural icing testing (calculations and verifications over the entire aircraft with engines, propellers, rotors and turbomachinery stages running, sensors and probes placed), ultimately increasing safety.
Manufacturers could create a dangerous disconnect between the two disciplines that reflects itself in icing incidents and accidents, by allowing a growing chasm between the modern numerical technologies used for aerodynamic design (CFD-Aero) and the ones used in the icing protection systems design and icing certification process (CFD-Icing). In fact, all they would be achieving is delaying the inevitable: 3D CFD-Icing is here to stay. Waiting until it becomes the norm only sets them up to play catchup with the earlier adopters of the technology.
To learn more, check out ANSYS FENSAP-ICE.