A three-dimensional icing model has been developed at ONERA to calcula
te ice accretion shapes for aerodynamic components that can not be pre
dicted using conventional two-dimensional codes. It is described, emph
asizing the original parts with respect to the two-dimensional existin
g models. The model includes Euler inviscid flow calculation. Droplet
trajectories are calculated in a three-dimensional grid. The remesh on
the Leading edge is adapted to follow aerodynamics singularities. The
boundary layer is calculated using a mixing length formulation to mod
el the wall roughness influence on convective heat transfer. Runback p
aths are integrated. The heat balance is calculated in a grid created
along the runback paths. The domain of validity of the three-dimension
al icing code is described; compared with the two-dimensional model th
is domain is wider, especially for high speeds. The three-dimensional
model is shown to simulate well a uniform ice deposit on a three-dimen
sional rotor blade tip. Then, a comparison of the three- and two-dimen
sional codes on an infinite swept wing shows that the corrected two-di
mensional code predicts the catch efficiency but not the ice shape. Fi
nally, it is shown that the continuum flux hypothesis prevents the thr
ee-dimensional model from simulating correctly the ''lobster tail'' ic
e shape (nonuniform ice deposit).