Académique Documents
Professionnel Documents
Culture Documents
Finite volume method is based on the discretization of the integral forms of the
conservation equations. Each physical equation is integrated into a control volume and
then discretized at each control volume centroid. The control volumes are commonly
called as cells. Each cells face interact with the other cells so, despite the calculation
occurs mainly to their centroids, they are then smoothed to follow contours and other
flow gradient properties such as pressure and velocity.
The conservation equations that were discretized at the software are listed. The
general equations wont have the unsteady term once that the simulation will be
performed assuming permanent regime and incompressible flow (no density variation).
This solving approach is commonly known as RANS (Reynolds Averaged NavierStokes).
The continuity (or mass conservation equation):
(
+
+
)=0
Eq. (1)
=
+
[ (
+
)]
Eq. (2)
Where the variables i and j relate to the coordinates x,y,z generating three equations
of moment. One for each direction.
Accordingly with the RANS methodology (Reynolds,1884), an average process is
applied to the equations (1 and 2) as follows:
(
1
)
=
+
[ (
+
)]
Eq. (3)
The knowledge of the averaged field does not allow the calculation of the advective trip:
(
)
Eq. (4)
Hence, it is necessary the decomposition into two terms: the average () and the
fluctuation () as follows:
= +
Eq. (5)
Eq. (6)
Where is the Reynolds tensor. To model it the Boussinesq hypothesis (1877), that
says the momentum transfer caused by turbulent eddies can be modeled with eddy
viscosity:
2
+
= 2
3
1
= ( + )
Eq. (7)
Eq. (8)
(
+
+
)=0
( )
1
=
+
[( + ) (
+
)]
Eq. (9)
Eq. (10)
The energy equation will not be solved since the flow at this application have negligible
temperature variations hence their solution will not pay off the accuracy reached
comparing it with the increase of the computational cost.
After the discretization of the equations the software ensemble the equations related to
the cells centroids using a matrix, then resulting into a problem of linear equations. The
software solves the matrix iteratively, reaching a final solution after the residual of all
cells calculation converge.
The characteristic length of the airplane is = 4.6220 []. Some authors say that in
order to have a result that does not have any domain interference the domain must have
a diameter of at least 10 times the characteristic length. The next figures show the
different domain sizes simulated to better comprehend and choose the best domain size
for this problem.
Tetrahedral/Pyramidal/Prisms
10 7mm
50
60
21
22
0.2
54 mm
1.2
0.35
Figure 5 Mesh Example (Blue plane Symmetry plane; Grey plane Auxiliary
plane)
The mesh was separated into six regions (Symmetry Plane, Wing, Fuselage,
Canard, Winglet and Far-field). Each of the regions has different boundary conditions.
On the symmetry plane. The solution that is obtained with a symmetry plane boundary
is identical to the solution that would be obtained by mirroring the mesh about the
symmetry plane (in half the resulting domain). The wing, fuselage, winglet and canard
regions were separated so they could have different element sizes on the mesh, allowing the
mesh to better follow the leading edge and wing profile curves improving the quality of the
mesh. Their boundary condition was set to Wall or no-slip condition. The Far-field region or
the outside shell of the semi-spherical domain was set to Velocity Intake condition i.e. the
hall shell had the velocity set to a constant, the cruise speed in the boundary conditions.
The software set up for the study of size and boundary conditions was:
= 200 = 87.45556 /
= 8.745556 /
= 10
Although the meshes built to check the domain influence in the result had the a close
number of cells a further study is required regarding the increase of the cell number once
that more cells would provide better information (calculation) about the smaller scales of
the flow hence have influence at the final result.
Four meshes were made in order to check the number of cells influence. They were
made using the same ICEM set up shown in the table 1. One mesh density was modified
to increase the number of cells closer to the aircraft.
N of Cells
3965144
6138069
7654387
9700551
The simulations set ups used to calculate the influence and the convergence of the final
result were the same used to analyze the domain influence (Table 1). The final mesh to
be used in the simulations for all the aircraft conditions will be chosen based on the
convergence of each mesh and the time taken to finish the calculations.
Given the transient nature of the problem related to the flow instability the solution
has to take into account the influence of the flow variations related to the turbulence.
The main parameter to check if the flow has or not turbulent behavior is the Reynolds
number. It relates the degree of influence of inertial forces over the viscous forces i.e,
whether the flow suffers great or few influence of its viscosity. For the domain, boundary
conditions and refinement study, one flow condition was used in order to standardize the
analysis. The cruise condition Reynolds number is:
=
87.4556 4.622
=
= 2.76 107
1.4604 105
A reference value largely used at the literature for external aerodynamic flow is
that air flows with Reynolds numbers higher than 5 105 can be considered fully
turbulent. Hence, the flow at cruise condition is turbulent and can be modeled as such.
Three turbulence models were compared using the same flight condition (cruise)
and the same software set up only changing the turbulence model to one of the three
shown in the table(4).
N of
Equations
Spalart-Allmaras
(SST) Menter
Each of the three models gives an equation (or more) that models the influence of the
Turbulent viscosity ( ) at the calculation of flow moments given by equation (10).
2
= 1 (1 2 ) + [
(( + )
) + 2
] [ 2 2 ] [ ] + 1 2
Eq. (11)
3 ; = ; 1 = 7.1
3
1
Eq. (12,13,14)
The production term is composed by a constant (1 ), a dumping factor for wall regions
(2 ) and for a modified deformation rate () defined by:
= +
2 ; 2 = 1
; = 0.41; 1 = 0.1355
2
( )
1 + 1
Eq. (15,16,17,18)
The modified deformation rate depends on the distance which is the distance to the
closest wall and is the absolute value of the deformation rate calculated with filtered
variables of the field, ie:
= 2
Eq. (19)
6
6
1 + 3
= ( 6
; = + 2 ( 6 ); 2 2 ; 2 = 0.3 ; 3 = 2.0
)
6
Eq. (20,21,23,24,25)
The model is implemented on STAR-CCM+ and will be the primarily base of comparison
for aerodynamic forces once proven that the calculation can give closest to experimental
results regarding this forces.
Figure 1 General Wall treatment Law for turbulent flow at smooth surfaces.
; + =
Eq. (26)
Where is the averaged flow velocity, y is the wall distance, and are the density and
viscosity of the flow respectively, and is the friction velocity defined as:
Eq. (26)
The software STAR-CCM+ has different wall treatment for each turbulence model. For
two turbulence models ( SST and Spalart-Allmaras) All + wall treatment was used.
For the Realizable the two layer all + was used.
The All y+ Wall Treatment is a hybrid model. It provides a more realistic modeling than
either the low-Re or the high-Re treatments, for when the wall-cell centroid falls in the
buffer region. This wall treatment is meant to give results similar to the low-y+ treatment
as + 0 and to the high-Reynolds treatment for + > 30 This treatment uses damping
functions for the source terms in the transport equation, but the source terms of the model
in the wall cell are suitably modified using the blended wall laws.
The Two-Layer All y+ Wall Treatment is a hybrid approach that seeks to recover the
behaviors of the other two wall treatments in the limit of very fine or very coarse
meshes. It contains a wall boundary condition for that is consistent with the two-layer
formulation. Apart from the specification of the Reynolds stresses in the wall cell. It is a
design goal that this wall treatment should give results similar to the low-y+ treatment
as + 0 and to the high-y+ treatment for + > 30 . It will also give reasonable results
for intermediate meshes where the cell centroid falls in the buffer layer.