Icing effects can reduce the flight safety under certain weather conditions. According to the US
National Transport Safety Board, icing is one of the major causes of flight accidents. Supercooled
water droplets present in clouds impinge on the surface of aircraft structures. They either solidify
totally on impact or partially then creating a thin liquid film runback depending on the flow
temperature and speed hence, creating dry rime ice or glaze wet ice respectively. Designing an
adequate de-icing mechanisms requires full knowledge of the icing phenomenon itself. Icing
experimental study cannot exceed the scope of a handful of simple cases due to complexity and
cost. Consequently the use of computational fluid dynamics is justified.

The icing process is assumed broken up into four steps:
1) single phase air flows around the wing
2) transporting water suspended droplets; droplets impinge into the surface
3) generating a liquid or dry film exchanging energy with the surface
4) accreated to shape the final form during a certain exposure time.

This process is usually assumed to occur on a single step considering that the time scale of the icing
process is very long compared with that of the air flow.

Current Icing simulation codes used by industries are based on over-simplified models. 1) A 2D
inviscid panel methods with an empirical boundary layer method is used for the air flow. Which is
usually followed by 2) a Lagrangian transport of droplets. And finally 3,4) an iterative
thermodynamic model for the liquid film to compute the ice thickness. To generate the final
geometry however, a Lagrangian node displacement is needed. A multi-step icing approach repeats
this process for portions of the required exposure time but still with decoupled time scales.

Maintaining a good grid quality requires a tedious amount of work, since strange irregularities in
iced shapes are difficult to be fully accounted for.
The Level-Set method introduced by Osher and Fedkiw could alleviate such a task. A passive scalar
function is introduced and is put equal to zero at the interface, positively defined outside and
negatively inside; the zero level represents the time evolution of the air/ice interface. To complete
the model, a PDE type thermodynamic model is used for the film, coupled with an external flow
solver.

In the present study a new method of icing simulation is developed. To get the most out of such
model, it is developed in the three-dimensional structured multiblock Navier-Stokes solver NSMB.
For a multi-step icing procedure, the geometry is defined by a passive scalar called the level-set.
This level-set function is set equal to the distance, negative on the inside and positive outside. A
penalized Navier-Stokes equation is solved on the external flow using a simple non-body fitted
mesh, wherein the solid is represented by the negative level-set valued cells. The droplets are
transported using an Eulerian approach using a TVD and a local time stepping schemes. The
impingement rate or what's called the collection efficiency is then fed to a Shallow-Water Icing
Model that evaluates the ice accretion, its height and velocity. The convective heat transfer
coefficient is obtained from the Navier-Stokes solver. Following that the Level-set function is
advected with the icing velocity to predict the new deformed geometry. The process is then repeated
for as many portions of the exposure time as needed.