Professional Documents
Culture Documents
A01-31146
AIAA 2001-2566
FENSAP-ICE: A Comprehensive 3D
Simulation System for In-flight Icing
F. Morency and H. Beaugendre
CFD Lab, McGill University
AIAA 2001-2566
Each of the four systems of PDEs Fig. 1: Module interactions within ICE3D
(airflow/impingement/accretion/heat-load) is
solved independently, with selected variables
passed between modules, when required.
Dt
- FENSAP, DROP3D and CHT3D use a weak-
Galerkin finite element method (FEM), on
structured, unstructured and on hybrid meshes.
(1)
In the ice accretion module, ICE3D, a finite
volume method (FVM) is found more where d is the distance from the wall, 5 the vorticity at
appropriate. The distortion of the solid a point, c6p cb^, c^, K, a, the closure coefficients and
surfaces caused by ice growth is automatically
f^ the closure function. This equation has shown to be
and transparently accounted for by an
Arbitrary Lagrangian Eulerian (ALE) scheme. particularly attractive for unstructured meshes.
The aim of this paper is to demonstrate, using the The tripping functions, ftl and //2, permit user
ONERA-M6 wing as an example, the 3D ice prediction control over the extent of the laminar region before
capability of FENSAP-ICE. Emphasis is put on the ice imposing transition. The extension for rough-wall
accretion module rather than on the flow solver, treatment9, essential in icing, has been included in the
impingement and heat transfer modules, all of which model. For a roughness coefficient ^, the conventional
have already been demonstrated. For completion, Nikuradse sandgrain roughness, the distance from the
however, a short description of the turbulent flow wall is increased as follows:
solver FENSAP is followed by a presentation of the
droplet flow solver DROP3D and the ice accretion (2)
module ICE3D. Finally, 3D ice accretion results for the
ONERA-M6 wing geometry, using a coarse grid are Spatial discretization is again carried out by a weak-
shown. These are only meant to illustrate FENSAP- Galerkin FEM and the equation is linearized by a
ICE's 3D capabilities and have not yet been calculated Newton method. To advance the solution in time, an
with precision in mind, a task that will certainly follow. implicit scheme is used, along with an iterative
GMRES procedure to solve the resulting matrix system.
FENSAP Heat fluxes at the wall are post-processed via a
consistent FEM approach10.
An accurate and consistent turbulent heat flux
evaluation is of particular importance when dealing DROP3D
with the ice accretion process. FENSAP-ICE uses one
of two turbulence models: the two-equation k-e model Lagrangian particle-tracking techniques are still widely
or the one-equation Spalart-Allmaras8 model. In the used to compute the droplet impingement on aircraft
latter case, the transport equation for the dimensionless components. DROP3D, on the other hand, is a fully 3D
turbulent viscosity V is (not quasi-3D) Eulerian alternative for airflows
containing water droplets. Suitable variables (water velocities are set to zero at all walls at the start of the
volume fraction and droplet velocity) are then calculations. No specific boundary conditions are
computed at the same nodes of the discrete domain needed at outlet.
where the airflow variables are known, so no particles
have to be tracked as they go through the mesh as in the A weak Galerkin formulation FEM is used to discretize
Lagrangian approach. On complex 2D or 3D the equations, with a streamline upwinding Petrov-
geometries, the traditional difficulties encountered by a Galerkin (SUPG) term11 added. Details of the numerical
Lagrangian method (long calculations times, difficulty method and validations can be found in Morency et al.12
near walls, difficulty in determining impingement
limits, sparsely seeded particles, etc.) do not surface in
an Eulerian approach. Validation
(6)
dt
where a(x, t) and ud(x, t) are mean values of the water Using the droplet's Reynolds number definition:
volume fraction and of the droplet velocity,
respectively, over a fluid element at location x at time t.
The first term on the right-hand-side of Eq. (4)
represents the air drag force on droplets, while the
second term represents the buoyancy and gravity forces. Total collection efficiency around a cylinder
The non-dimensional air velocity, ua, is obtained by
using any Navier-Stokes or Euler code, here FENS AP.
Once the friction force and the heat fluxes are known
from the airflow solution (FENSAP) and the mass rate
of water caught is known from the impingement
module (DROP3D), the ice accretion can be assessed
byICE3D.
One of the rare 3D cases available for validation is the The resulting system of partial differential equations is
impingement of droplets around a sphere. In figure 3, the following:
the collection efficiency distribution is presented. An
inviscid airflow solution is first calculated, and then
Mass Conservation
DROP3D is used to calculate collection efficiency for
an 18-micron mono-disperse droplet and an 18-micron
MVD distribution. The two collections efficiency -m
curves are inside the experimental range of error.
Energy Conservation ice mice. Compatibility relations are needed to close the
system. One way to write them is the following:
r \*U^LWCP hf>0
r (12)
(13)
Pw\ Jf
[+div(ufhfCj}\ mke>0 (14)
hff>0 (15)
where the first three terms on the right hand side model,
respectively, the heat transfer caused by the super- (16)
cooled water droplets impingement, the evaporation
and the ice accretion. The last two terms represent the
radiative and convective heat transfer. The discretization of the equations is by FVM. The hull
of the three-dimensional mesh at the air-structure/ice
The coefficients Pw' ^w' ..**
*^ice»^evap» ^fusion shape interface is called the surface mesh. From the
represent physical properties of water, while T rf/00 , U^, surface mesh a dual surface mesh is obtained by
connecting the ban-centers of the surface mesh cells to
LWC, a, and T^ are airflow and droplet parameters
the mid-edges of the cells.
specified by the user. The ambient icing conditions
completely determine those values. The tilde symbol Validation
over 7, i.e. 7\ stands for the temperature in Celcius,
otherwise temperature is in Kelvin. Ice accretion computation is carried out through a full
interaction between FENSAP, DROP3D and ICE3D. A
The Eulerian droplet module provides local values for turbulent airflow solution is first calculated on a clean
the collection efficiency /J and the droplet impact (non-iced) geometry, then DROP3D computes the
velocity U^. The flow solver provides the local wall droplet impingement and ICE3D the nodal
displacement due to ice accretion. A new solution and
shear stress twall and the convective heat flux Qh. The mesh are then automatically generated by FENSAP
evaporative mass flux is recovered from the convective with the ALE feature.
heat flux using a parametric model18. There remain
three unknowns: the film thickness hr, the equilibrium Figures 5, 6, and 7 present validations results on a
NACA 0012 airfoil for rime and glaze ice conditions.
temperature T within the air/water film/ice/wall
interface, and the instantaneous mass accumulation of
impingement evap/subl
\
Fig. 4: Heat and mass balance in a thin Him Fig. 5: Ice accretion on a NACA 0012 airfoil:
comparison with numerical results,
LEWICE Run 404
0 Experimental
2 - NACA 001 2
ICE3d i »...i~>
0 Experimental
Uwfc»7mn • ICE3D
1.5
——— NACA 0012
0.5
-0.5
Fig. 6: Ice accretion on a NACA 0012 airfoil: Fig. 7: Ice Accretion a NACA 0012 airfoil
comparison with experimental results, comparison with experiment,
LEWICE Run 404 LEWICE run 308
Re 4.14 106
MVD 20|im
The only purpose here is to demonstrate that 3D ice Fig. 8: 3D mesh around ONERA-M6 wing
accretion calculations around a wing are possible with
FENSAP-ICE, and thus, a very coarse grid hexahedral also shows the LWC distribution around the wing. Most
mesh of around 200 000 nodes is used. That size of grid of the water mass is caught in the first 10% of the wing
enables a viscous turbulent air solution in reasonable chord, around the leading edge.
times. The upper part of the computational domain is
shown in figure 8, with cutting planes at y=0 and z=0. The resulting ice shape for the studied ambient
conditions is plotted in figures 11 and 12, which clearly
DROP3D uses the turbulent air solution to get the mass show that the ice shape is substantially bigger near the
of water caught by the wing. Although DROP3D does tip of the wing. Because of the cold temperature
not need a fine grid to achieve convergence, it is clear selected, no horns are captured, as one small amount of
that the quality of the droplet solution depends largely water runback gives a rounded shape to the ice
on the quality of the airflow solution. In the present accretion.
case, the size of the boundary layer is exaggerated by
the flow solver, due to the coarseness of the grid nearThe resulting ice shape is quite smooth because it was
the wing surface. calculated using a single airflow solution. This means
that heat transfer coefficients and collection efficiencies
The ratio of tip to root chords is 0.5 for the ONERA- were calculated on the clean wing and then assumed to
M6 wing. 2D cuts along wingspan, shown in figure 9 prevail for the next 6 minutes of ice accretion. This is
as a function of jc, confirm that maximum collection the fastest way to do the calculations, but it probably
efficiency increases toward the tip. This is to be prevents some 3D effects from appearing.
expected since total collection efficiency is usually
higher on small bodies than on large ones. At the tip, Figure 13 shows a closer view of the tip of the wing.
3D vortical effects also dramatically enhance the The ice shape thickness varies from 1.5% of the chord
amount of water caught. It is not clear how a 2D or at the root of the wing to 6.6% at the tip. At the tip, ice
quasi-3D approach could capture such effects, crucial accretes normally to the leading edge but also laterally,
for stability and control and for anti- or de-icing. creating a spherical shape.
Collection efficiency values below 0.05 are probably The ice accretion calculations on the ONERA-M6 wing
inaccurate in the present case because of the coarseness are a good example of the modular interactivity
of the grid. capability of FENSAP-ICE to predict 3D ice shapes.
The coarse grid used to solve the airflow is, however,
In figure 10, the collection efficiency distribution not propitious for some of the 3D effects to occur. Heat
around the wing is plotted. A 2D cut in the x-y plane transfer distribution and shear stress on the wing
Conclusions
FENSAP-ICE is a comprehensive icing code for in-
flight ice simulations in 3D, as well as in 2D. It is built
in a modular fashion with four modules (three are
described in this paper), each having a specific task but
with all being linked. The first module, FENSAP, takes
care of the airflow computation by either Euler or
Navier-Stokes models. DROP3D uses the calculated
airflow for droplet impingement calculations, with no
particle tracking. Finally, ICE3D uses shear forces and
heat fluxes from the airflow calculation of FENSAP
and the water catch from DROP3D to yield the 2D
shape of the ice on the 3D surface. FENSAP-ICE has
been successfully used to predict collection efficiencies
on 2D and 3D bodies. The ice calculation module has
been validated in 2D and is in the final phases of its
Fig. 9: Collection efficiency distribution as
validation process in 3D, with quite promising results.
function of x: 2D cuts along the wing
Fig. 10: Collection efficiency on Fig. 12:2D cuts of ice shape along
ONERA-M6 wing ONERA-M6 wing span
Fig. 11: Top view of the ice shape on Fig. 13: Ice shape: zoom on the tip of
the ONERA-M6 wing the ONERA-M6 wing
10
Gresho, P.M., Lee R.L., Sani R.L., Maslanik, M.K.
and Eaton, B.E. "The consistent Galerkin FEM for
Computing Derived Boundary Quantities in Thermal
References and/or Fluid Problems" International Journal for
Numerical Methods in Fluids, 7, pp.371-394,1987.
11
Hughes, TJ.R and Brooks, A., "A Theoretical
Abbott I. and von Doenhoff A.E., "Theory of Framework for Petrov-Galerkin Methods with
Wing Sections", Dover Publications, New York, Discontinuous Weighting Functions: Application to
1949. the Streamline-Upwind Procedure", Finite Elements
in Fluids, Volume 4, Chapter 3, Edited by R.H.
Lynch F., Valarezo W. and McGhee R., "The Gallagher, D.H. Norrie, J.T. Oden and O.C.
Adverse Aerodynamic Impact of very Small Zienkiewicz, 1982 John Wiley & Sons.
Leading-Edge Ice (Roughness) Buildups on
Wings and Tails", Effects of Adverse Weather 12
Morency, F., Beaugendre, H. and Habashi, W.G.,
on Aerodynamics, AGARD CP-496,p. 12.1-12.8. "FENSAP-ICE: a Navier-Stokes Eulerian Droplet
Impingement Approach for High-Lift Devices", CASI
Riley J.T., Anderson D., Rios M.A., Dumont conference, May 2001.
C.J., "A Study of Intercycle, Residual and Pre-
activation Ice" AIAA Paper 2001-0089, January 13
Boutanios, Z., Bourgault, Y., Habashi, W.G., Isaac,
2001. G.A. and Cober, S.G., "3-D droplets impingement
analysis around an aircraft's nose and cockpit using
Baruzzi, G.S., Habashi, W.G., Guevremont, G. FENSAP-ICE". AIAA Paper 98-0200, January 1998.
and Hafez, M.M., "A Second Order Finite
Element Method for the Solution of the 14
Brun, R.J., Lewis, W, Perkins, P.L. and Serafm, J.S.,
Transonic Euler and Navier-Stokes Equations", "Impingement of Cloud Droplets on a Cylinder and
International Journal for Numerical Methods in Procedure for Measuring Liquid-Water Content and
Fluids, Vol. 20, 1995, pp. 671-693. Droplet Sizes in Supercooled Clouds by Rotating
Multicylinder Method", NACA Report 1215,1955.
Bourgault, Y., Habashi, W.G., Dompierre, J. and
Baruzzi, G.S., "A Finite Element Method Study 15
Brahimi, M.T., Tran, P., Chocron, D., Tezok, F. and
of Eulerian Droplets Impingement Models", Paraschivoiu, I., "Effect of supercooled large droplets
International Journal for Numerical Methods in on ice accretion characteristics". AIAA Paper 97-
Fluids, 29:429-449,1999. 0306.
Beaugendre, H., Morency, F. and Habashi, 16
Bourgault, Y., Beaugendre, H. and Habashi, W.G.,
W.G., "ICE3D, FENSAP-ICE's In-flight Ice "Development of a Shallow Water Icing Model in
Accretion Module", CASI conference, May 2001. FENSAP-ICE". Journal of Aircraft, 37:640-646,
July-August 2000.
Croce G., Habashi W.G., Guevremont G. and
Tezok F., "3D Thermal Analysis of an Anti-Icing 17
Morency, F., Tezok, F. and Paraschivoiu, I., "Anti-
Device Using FENSAP-ICE", AIAA paper 98- Icing System Simulation Using CANICE", Journal of
0193, 1998. Aircraft, 36:999-1006, November-December 1999.
Spalart, Ph.R. and Allmaras, S.R., "A One- 18
Tran, P., Brahimi, M.T. and Paraschivoiu, I., "Ice
Equation Turbulence Model for Aerodynamic Accretion on Aircraft Wings with Thermodynamic
Flows" AIAA Paper 92-0439. Effects", AIAA Paper 94-0605.
Spalart, Ph.R., "Trends in Turbulence 19
Wright, W.B. and Rutkowski, A., "Validation Results
Treatments" AIAA Paper 2000-2306. for LEWICE 2.0" and CDROM.
10