Professional Documents
Culture Documents
Aerodynamic CFD Analysis On Experimental Airplane: Mechanical Engineering 'Fluid Dynamics' Karel Lammers (s1005138)
Aerodynamic CFD Analysis On Experimental Airplane: Mechanical Engineering 'Fluid Dynamics' Karel Lammers (s1005138)
experimental airplane
The choice for a CFD analysis follows from my background in uid dynamics. The aspect that I
like in this eld is the computational methods. I wanted to gain some experience in the way these
methods are applied in practical problems, therefore I chose for this subject
This internship is established with help of the following people, and I would like to thank them:
Prof.Dr.Ir. H.W.M. Hoeijmakers, for arranging the contacting for my internship and putting
trust in me.
Assoc.Prof. C. Bil, for guiding my through the whole internship. Putting trust in me and
answering all my questions
Mrs. A. Brabender, for helping me with the visa application and helping me with the necessary
arrangements
Dr.Ir. D. van de Belt, for guiding me through the overall internship process.
1
2 | Abstract
In nowadays society airplanes are a common form of transport. Airplanes are used for international
transportation and also for domestic ights.
But for domestic ights in general small airports are used. A lot of these small airports have strict
noise restrictions and that limits the range of airplanes that can make use of these airports. So
airplanes that produce small amounts of noise are required. That is the focus of the study of Cees
Bil. The study is about a mid size passenger airplane with a low noise prole.
In order to keep this airplane commercially attractive and with an eye on the environment, it is
necessary to keep the fuel consumption as low as possible. A way to keep this consumption low is
to reduce the drag of the airplane.
The airplane is still in the design stage and in order to get a quick approximation of the basic
aerodynamic properties of the airplane a CFD analysis has to be done, which is my assignment.
Therefore this article is about a CFD analysis on an experimental airplane. The required model
was already present and made in CATIA V5. The engine is modeled with help of an actuator disk.
The meshing was done in ANSYS MESHING and the analysis is done in ANSYS FLUENT.
Due to this analysis it is clear that the geometry of the airplane is not yet optimal in terms of
drag. However the new conguration with the twin engine at the end of the fuselage in combination
with the empennage seems plausible and did not introduce supersonic ow regimes or strange ow
behaviour.
2
Contents
1 Preface 1
2 Abstract 2
3 Table of symbols and denitions 4
4 Introduction 6
5 Analysis 7
5.1 Denition of the problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
5.2 Work plan and used research method . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
5.2.1 Work plan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
5.2.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5.2.3 Deadlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5.2.4 Meetings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
5.2.5 Predecessor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.3 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.3.1 Governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
5.3.2 Turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.3.3 Actuator Disk . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.3.4 Solvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.3.5 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.4 Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.4.1 Choice of method and elements . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4.2 Deleted features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4.3 Boundary layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.4.4 Local mesh sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.4.5 Final mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.5.1 Main model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.5.2 Analysis without engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.6 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.6.1 Grid renement study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.6.2 Comparison with expected data . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.8 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.8.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
5.8.2 Airplane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Bibliography 25
Appendix 26
I Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3
3 | Table of symbols and denitions
Denitions
CFD Computational uid dynamics
Drag Aerodynamic resistance force caused by the ow, parallel to the ow
Lift Aerodynamic force caused by the ow, perpendicular to the ow
Empennage Horizontal stabilizer (rear wing)
Meshing Generating a structure of small elements or volumes, which is used to
perform nite element or nite volume analyses.
Grid The network of elements on which an nite element or nite element volume analysis
can be performed.
Symbols
x x-position
y y -position
z z -position
t Time
h Height
S Wing area
b Width of the airplane
L Length of the airplane
Wplane Weight of the airplane
Fengine Trust of one engine
Aengine Surface area of the inlet of the engine
rengine inner radius of the engine
T Absolute temperature
T∞ Free stream absolute temperature
p Static pressure
p∞ Free stream static pressure
ρ Density
ρ∞ Free stream density
a Speed of sound
E Total energy
µ Dynamic viscosity
v Speed
v∞ Speed of the plane
g Gravity
M∞ Free stream Mach number
Re Reynolds number
Re∞ Free stream Reynolds number
τ̄¯ Stress tensor
k Turbulence kinetic energy
ω Specic dissipation rate
ṁ Mass ow
4
4 | Introduction
As described earlier, airplanes with a low noise prole are needed for domestic ights in order to
make use of airports with strict noise restrictions.
Therefore an experimental airplane has been developed, which can bee seen in gure 4.1. This new
airplane conguration is believed to have a low noise prole because the empennage (rear wing)
functions as a noise shield for the engines.
This project is still in the conceptual design stage and therefore a lot of details are not fully known.
Nevertheless it is necessary to get an indication of the basic aerodynamic properties. It is not
desirable to obtain these properties by a wind tunnel experiment in this stage. In the design stage
the geometry of the plane will be changed constantly and so it would be very time consuming if
after each small adaption a new model must be made. Therefore there is chosen to perform the
analysis by making use of a CFD analysis. The basic aerodynamic properties to be investigated are
the lift and the drag.
6
5 | Analysis
5.1 Denition of the problem
The new airplane is based on the Bombardier CRJ700 series, which is a mid size passenger plane.
The operating conditions and the details of the airplane can be found at table 5.1.
Table 5.1:
Operating conditions Details plane
h = 10668 m Fengine = 12454 N
v∞ = 243 m/s S = 43.85
T∞ = 218 K b = 25.8 m
p∞ = 23842.3 Pa L = 31 m
ρ∞ = 0.379597 kg/m3 Wplane = 29474 kg
a = 296.535 m/s
µ = 1.44446e−5 Pa · s
7
5.2.2 Methods
Two types of software were recommended by Cees Bil and available at RMIT, ANSYS FLUENT
and ANSYS CFX. From literature study follows that there is not a big dierence between the two
types of software. ANSYS FLUENT has the advantage of coupled solvers, which can contribute to
a better convergence for pressure based solvers. Therefore will the analysis be performed in ANSYS
FLUENT.
The meshing can be done in ANSYS MESHING, but also in ICEM CFD. ICEM CFD is capable
of generating a structured mesh, which is an orientated mesh and therefore of good quality for
complex geometries. While ANSYS FLUENT has the benet that only one type of software will be
used which will reduce the amount of time to get familiar with the software. Because of the very
complex geometry a structured mesh would lead to a mesh with a lot of dierent orientated groups
of elements which must be orientated manually and is very time consuming to create. So therefore
an unstructured is used (see section 5.4.1), which makes it unnecessary to make use of ICEM CFD,
so ANSYS MESHING is used for the meshing
There is already a model of the airplane present, which is made in CATIA V5. This model contains
a highly detailed representation of the engine, which is not very interesting for this type of analysis.
Therefore will this detailed representation be simplied to an 'actuator disk'. An actuator disk is
a mathematical model mainly used for propellers. An actuator disk is an innitely thin disk where
the uid can ow through while experiencing a pressure jump (discontinuity). In this way we can
represent the eects of the engine on the surrounding ow in an easier way, which is more suited
for this analysis.
5.2.3 Deadlines
RMIT stated only one ocial deadline, the actual end date of my internship. But for the sake of
the project, some milestones are introduced which are purely for my own planning. The overview
of deadlines and milestones can be seen in table 5.2. The working days are from 09:00-17:00 from
Monday to Friday.
5.2.4 Meetings
In order to keep track of my progress, some meetings are planned. In the rst 5 weeks no meetings
are planned because in this time I will be doing literature study, gathering information, installing
software and getting familiar with the software. During these rst weeks Cees Bil will however be
available for questions and he will keep an eye on my progress.
During the time planned for validation and analysis more meetings are planned because the expertise
of Cees Bil plays a big role in this part.
8
5.2.5 Predecessor
There is no predecessor who was already occupied with building a CFD model of this airplane. But
the airplane itself is designed by a group of students as a project assignment of RMIT. The project
group consists of the following students: Wan Muhammad Ilyas Yusof, Shahril Asyraf Radzali,
Napatsorn Prakobboon Abdul Majid Nazli, Peter Jarrod Woolley, Matthew Yu, Amir Izzuwan
Adnan, Ibrahim Al-Saad and as supervisor Assec. Prof. Cees Bil. The purpose and the process of
designing this aircraft is described in a nal report [7].
5.3 Theory
5.3.1 Governing equations
For a ow of a uid which can be considered to be continuum and which is uniform, homogenous
and non-chemically-reacting the Navier-Stokes equations describe the properties of this type of ow.
This set of equations consists of the following equations:
According to the theory guide [3], ANSYS FLUENT solves the rst two equations for every ow.
The last equation and additional equations can be added manually. In order to determine which
equations for this model have to be added some characteristics of the ow have to be evaluated.
This can be done with the free stream Mach number (M∞ ) and the free stream Reynolds number
(Re∞ ), which are both dimensionless numbers and can be found at equation 5.1.
v∞ 243
M∞ = = = 0.82
a 296
ρv∞ L 0.379597 · 243 · 31 (5.1)
Re∞ = = ≈ 200e6
µ 1.44446e−5
A ow is generally considered to be incompressible when M∞ < 0.3, therefore the ow for this
model will be considered to be a subsonic compressible ow.
As mentioned earlier the ow is considered to be compressible. The density will be computed by
the ideal gas law, given as:
p∞ + p
ρ= R
(5.4)
Mw T
9
Where Mw is the molecular weight. The temperature T will be obtained by solving the equation of
conservation of energy.
Where kef f is the eective conductivity, J~j is the diusion ux and τ̄¯ is the eective stress tensor.
Sh is an optional heat source but it is not present in this model.
5.3.2 Turbulence
At a certain value of the Reynolds number the ow will start to show turbulent regions. For a at
plate the Reynolds number at which a ow is considered to be fully turbulent is Re∞ > 0.5e6 [4].
So it is an appropriate assumption that the ow for this model will be fully turbulent.
There are a lot of models that can represent turbulence in a fully turbulent ow, therefore more
turbulent models are suited for this analysis. The Shear-Stress Transport (SST) k − ω model will
be used. This model is a commonly used turbulence model in aerospace engineering. The k − ω
models are suited for high Reynold numbers and include compressibility, shear ow spreading and
they include some low Reynold number features. The advantage of the SST model in comparison
with the standard k − ω model is that it is better suited for adverse pressure gradients and ow
separation. Since these two features are important for a drag analysis, this turbulence model will
be used. This model has an extra option for low Reynolds number ows, but in this analysis a high
Reynolds number is present and therefore this extra option is not used.
The SST k − ω model introduces two extra equations to be solved, which are the following:
∂ ∂ ∂ ∂k
(ρk) + (ρkvi ) = Γk + G̃k − Yk + Sk
∂t ∂xi ∂xj ∂xj
(5.6)
∂ ∂ ∂ ∂ω
(ρω) + (ρωvi ) = Γω + Gω − Yω + Dω + Sω
∂t ∂xi ∂xj ∂xj
Where G̃k is the generation of turbulence kinetic energy due to mean velocity gradients, Gω is the
generation of ω , Γk and Γω are the eective diusivity of k and ω respectively. Yk and Yω are the
dissipation of k and ω due to turbulence respectively. Dω is the cross-diusion term and Sk and Sω
are optional sources in the model.
It is beyond the scope of this report to explain these equations. But basically the k − ω models are
empirical models based on model transport equations for the turbulence kinetic energy (k) and the
specic dissipation rate (ω ). The rst equation of 5.6 refers to the turbulence kinetic energy and
the second equation refers to the specic dissipation rate.
10
analysis, the thrust that the engine will deliver is constant and so the pressure jump is also constant.
Dening the actuator disk in such a way that it models the engine properly is tricky. A problem
that is faced is that the velocity of the air inside the engine becomes very high in order to satisfy
the static pressure jump of the actuator disk at the end of the engine. The ow inside the engine
can even become supersonic, introducing a high pressure drag due to shock waves. This behaviour
of air is not present in a real engine, therefore the drag of the plane will be exaggerated due to the
high pressure drag induced by the supersonic ow in front of the actuator disk. But the actuator
disk will still give a good indication of how the ow looks with an engine included.
Although Froude's Momentum Theory is only applicable for 1D incompressible inviscid ow it will
still be used partially. The rst part of Froude's Momentum Theory holds for 3D compressible and
viscid ows as well. But it is not possible to calculate the necessary pressure jump to generate the
required thrust analytically with Froude's Momentum Theory for this type of ow.
The thrust of the engines is determined in the report of the predecessor [7] and is 12454 N per
engine.
The part of Froude's Momentum Theory that is used is how the thrust is coupled to the velocity:
2
Fengine = ṁ∆v = Aengine v1 ρ1 (v2 − v1 ) = πrengine v1 ρ1 (v2 − v1 ) (5.7)
Where v1 and p1 are the velocity and pressure in front of the actuator disk respectively. v2 is the
velocity after the actuator disk. The pressure jump that produces the force that equals the thrust
will be found by trial and error.
5.3.4 Solvers
ANSYS FLUENT has two types of discrete solvers, pressure-based solvers and density-based solvers.
Both make use of the control-volume techniques. Originally the pressure-based solver was developed
for incompressible ows, but it is now capable of dealing with compressible ows by making use of
the coupled pressure-based solver. The coupled pressure based-solver is more suitable for dealing
with compressible ows with large regions of low Reynolds number in it. Both solvers could be used,
but because of the fact that our analysis is not dominated by large low Reynold number regions the
density-based solver is used.
Equations 5.2, 5.3 and 5.5 are nonlinear, so the solution will be obtained by an iteration process.
The linearization in this iteration process can be done explicit or implicit. The implicit approach is
unconditionally stable but takes more computation time, while the explicit approach is faster but
has the so called CFL stability criterium restricting the ratio between the time step and the spatial
step. So the implicit method gives us more freedom in terms of time step and spatial step, which is
an advantage in the process of building the model.
For the discretization of the ow equations (5.2, 5.3 and 5.5) the second order upwind scheme will
be used. The Turbulent Kinetic Energy equation (5.6) and the Specic Dissipation Rate equation
(5.6) which are introduced by the turbulence model will be discretized by the rst order upwind
scheme. The upwind discretization scheme is very stable and can cope with coarse meshes. This is
very useful because we start with a coarse model and then rene it. There is however a drawback
of the upwind scheme and that is that it introduces articial dissipation.
Upwind schemes are used in the solvers because of their stable behaviour, allowing coarses meshes
and more workable numbers of elements. But the stable behaviour of these upwind scheme come
with a cost, they introduce so called artical dissipation. Which is mathematical dissipation that
11
stabilizes the scheme, however this dissipation is physically not present. This articial dissipation
increases with higher gradients in the ow, therefore structured meshes are preferable with upwind
schemes. Structured meshes allow the user to direct and align elements according to those gradients,
resulting in less articial dissipation. For this analysis unstructured meshes are used (see section
5.4.1) and so arranging elements according to their gradients is not possible. The only way to reduce
this articial dissipation is by grid renement.
Each boundary must have a boundary condition in order to solve the system of equations. A lot
of boundary conditions are already present in ANSYS FLUENT, and the following boundaries are
assigned with the following boundary conditions:
Inlet
Velocity inlet: The inlet velocity is 243 m/s and is normal to the inlet plane. The turbulence
intensity is set to 3 and 5.
Far eld
Pressure far eld: M∞ is set to 0.82 with direction (1,0,0) and the turbulence intensity to 3 and
5. This boundary condition computes the parameters of the ow with help of characteristics,
so it follows from the interior elements.
Symmetry plane
Symmetry: This boundary condition is applicable for symmetry planes.
Outlet
Pressure outow: At the outlet the turbulence intensity is set to 3 and 5.
Airplane
Wall: The roughness height and roughness constant are not known, therefore this is set to
default, which is 0 and 0.5 respectively.
Actuator disk
Fan: The fan boundary condition is an interior boundary condition. It corresponds to the
explained theory of 5.3.3. The constant pressure jump is set to 10000 Pa.
5.4 Mesh
Generating a good quality mesh is a key step for obtaining a correct solution. The mesh should also
include a boundary layer to solve the turbulence equations. For the nal analysis several models
had to be build. The main model that will be used for the analysis, a model without engines and
a model with a coarse mesh to perform a grid renement study. The reason why a model with a
coarse mesh is used for the grid renement study is explained in section 5.6.1. The main model and
12
the model without engines will have the same mesh parameters, the coarse mesh will have dierent
mesh parameters.
But rst the CAT-part should be simplied, small details should be deleted.
There are however two drawbacks of unstructured meshes in this case. Because of the fact that
the mesh is unstructured, the elements can not be orientated in directions with high gradients.
Therefore more elements are needed for regions with high gradients. The other drawback is that
this method induces more articial dissipation in comparison with structured meshes. By rening
the grid this phenomena will be minimized.
The patch conforming method requires a higher quality geometry model than the patch independent
method, but this is not a problem.
The drawback of the tetra elements is that more elements are needed in comparison with other
types of elements.
The trailing edges of the wing and the empennage have a nite thickness in the CAT-part. This
causes badly shaped elements around the trailing edges, which are of great interest for the analysis.
Therefore the wings are slightly extended and the slender surface is replaced by an edge (see gure
5.2). The wings and the empennage are created with discrete surfaces. They contain therefore a
lot of unnecessary edges which cause badly shaped elements. Therefore the surfaces that form the
wing and the empennage are merged into bigger surfaces with help of the merge tool in ANSYS
(see gure 5.2)
13
(a) Discrete surfaces and trailing edges (b) Merged surfaces and adapted trailing edges
For the main model and the model without engines the amount of layers in the boundary layer is
set to 40. The rst layer height is set to 1e−4 m and the grow rate to 1.4. For the coarse mesh the
boundary contains 20 layers and the rst layer is 5e−4 m and the grow rate is 1.4
14
Table 5.3:
Part Element size in [m] Part Element size in [m]
Cockpit 0.075 Back and front side pylon 0.03
Fuselage 0.1 Side pylon 0.075
Leading edge wing 0.075 Inlet engine 0.05
Wing 0.1 Inside engine 0.075
Trailing edge wing 0.075 Outlet engine 0.05
Winglet 0.075 Outside engine 0.1
Leading edge empennage 0.075 Actuator disk 0.05
Empennage 0.1 Nose 0.075
Trailing edge empennage 0.075
Table 5.4:
Part Element size in [m] Part Element size in [m]
Cockpit 0.1 Back and front side pylon 0.03
Fuselage 0.15 Side pylon 0.01
Leading edge wing 0.1 Inlet engine 0.05
Wing 0.15 Inside engine 0.075
Trailing edge wing 0.1 Outlet engine 0.05
Winglet 0.1 Outside engine 0.1
Leading edge empennage 0.075 Actuator disk 0.05
Empennage 0.1 Nose 0.075
Trailing edge empennage 0.075
Figure 5.3 shows some important views of the mesh. It can be seen that the boundary layer is
created nicely, even for the complex geometry like the engine. The coarse mesh looks a lot the same
but only more coarse.
15
(a) Total mesh
16
(d) Boundary layer
5.5 Results
First an analysis was performed with the engines. It turned out that the resulting drag of this
analysis was very high. In order to determine the reason of this high drag an analysis without the
engine was performed. An analysis with a coarser mesh was also performed, but this analysis will
be touched at the grid renement study (section 5.6.1). The other two analyses will be touched
here.
Some important views of the pressure, velocity and temperature elds can be found at appendix I
Also a y + contour plot and a y ∗ contour plot can be found at gure 5.5.
17
(a) y + values
(b) y ∗ values
From the y + contour plot it is clear that the values of y + lay in the specied range of y + < 300.
y ∗ is another parameter that looks a lot the same as y + and this should also be lower than 300,
which is also the case. And at the velocity eld plots a boundary layer is clearly visible. At the wall
(airplane) the velocity of the ow is zero as is it supposed to be. This all means that the turbulent
model is working properly.
18
At table 5.5 the generated lift and drag can be found of the main model with the ne mesh and
the coarse mesh. The DL value is very low, for commercial airplanes this should be in the range of
8 − 15.
As said earlier two possible reasons are examined. As explained in 5.3.3, the actuator disk can cause
a supersonic shock in the engine resulting in a high pressure drag.
The other possible reason that is examined is the ow behavior around the pylon of the engine. At
appendix I at gure I.1b it can be seen that there is a stagnation regime present at the front side
of the pylon and from gure I.2b it is clear that there is a wake regime at the back of the pylon,
which can cause high drag. Therefore an analysis without the engine is performed, if this analysis re-
sults in much better lift over drag ratio it is clear that the problem lays in the modeling of the engine.
Figure 5.6: Flow around the end of the airplane without engine
And the generated lift and drag can be found at table 5.5.
19
It is clear from the table that by removing the engine and thereby also the actuator disk the drag
does decrease, but not very much. In fact the engine of the original model contributes for 15% of
the total drag of the airplane. Which is a high value, but not uncommon according to M. Sadraey
[2].
Table 5.5:
Cases Lift Drag L
D
Pressure Viscous Total Pressure Viscous Total
Orignal model 165909 94 166003 56594 10262 66856 2,48
Model without engine 238843 81 238924 46732 10165 56897 4,20
Coarse grid 179909 198 180108 59095 15156 74251 2,43
5.6 Validation
Computational models can come up with solutions that are not real at all. Therefore it is always
necessary to examine your results.
The validation will consist of two parts. The rst part will be a grid renement study and the
second part will be a comparison between the obtained results and expected data.
Table 5.6:
Case Number of elements Factor of renement Lift Change Drag Change
Coarse main model 2,213,509 180108 74251
Fine main model 4,611,052 2.08 166003 7.83% 66856 9.96%
It is clear that the coarse mesh was still too coarse. Because a change in lift of 7.83% and a change
in drag of 9.96% by a mesh renement of factor 2 is slightly too big. The grid is considered to be
ne enough when the change in lift and drag is less than 5% when the mesh is rened by a factor of
2. Clearly the coarse mesh was not ne enough, but the main model can not be evaluated because
an analysis on a ner mesh is not present. But because the values of the grid renement study for
the coarse mesh are close to the acceptable values it is assumed that the grid is ne enough.
20
The basic geometry is a known geometry and therefore a big part of the behaviour of the ow is
known one forehand. Some key features should be present in the obtained data. The ow must
contain the following features:
Stagnation points on the nose of the plane, the leading edges of the wing and the empennage
and the front side of the engine and pylon. (see gure I.1)
If there are to be any supersonic ow regions they should be on the the front part of the wing
and empennage, the cockpit or in the engine. (see gure I.2)
In order to generate lift there should be a pressure dierence between the topside of the wing
and the downside of the wing. The topside should have a lower pressure regime and the
downside a higher pressure regime in comparison with each other. (see gure I.1c)
At appendix I it can be seen that these features are present in the ow.
5.7 Conclusions
The meshes that are used for these analyses all contain more than 512,000 elements. This means
that in order to perform the analyses a full license of ANSYS FLUENT is required. At rst this
license was not present and therefore a lot of time was put into optimization of a very coarse grid.
But 10 days for the ocial deadline I received the full license and from that moment it was possible
to build ner meshes. However 10 days is a short time span to produce several good quality meshes
and perform the analyses. Therefore it was not possible to perform all the analyses that were
required. An analysis on a ner mesh was required in order to complete the grid renement study
and an analysis without the actuator disk would have given more insight in the representation of
the engine by the actuator disk.
From section 5.6 can be concluded that the model is working ne. There is however still room for
some improvement, but it is to be expected that the obtained data will not change a lot.
The most outstanding outcome of this analysis is that the drag is very high of this airplane cong-
uration. A few improvements can be made in order to reduce this drag greatly, which can be found
in section 5.8.
The lift that is generated by this airplane is too big, this means that the wing area is too big for
this angle of incidence. The angle of incidence can be reduced or the wing area could reduced, or a
combination of both of them (see 5.8) .
The innovative part of this airplane is the empennage in combination with the twin engine on the
back of the fuselage. With an eye on the recommendations in 5.8 it is possible to state that an
airplane in this conguration can work properly. From table 5.5 it is clear that the engines do not
induce a ridiculous high drag, and also the ow around the empennage does not seem to cause a lot
of problems, which was an important aspect of this analysis. There were worries of supersonic ow
regimes in between the engines, but from the velocity eld plots it is clear that this is not the case.
And from the velocity eld plots it is also clear that the empennage does not introduce ow regimes
that cause high drag. There is no reversion of the ow present on the empennage and neither does
it introduce a big wake area.
The main reasons for the high drag are other parts of the airplane which can be improved rather
easily and will be discussed in the recommendations.
21
5.8 Recommendations
The recommendations are split up into two kinds of recommendations. There are some recommen-
dations regarding the model (mesh, solver, etc.) and there are some recommendations regarding
the design of the airplane. The biggest gain in drag reduction can be found in new shapes for the
pylon and the connection of the wing to the fuselage.
5.8.1 Model
The model is working ne, but there is still room for improvement. Because the right license for
the software was obtained 10 days before the nal deadline, there was not much time to produce
a ecient mesh. Therefore the mesh could be optimized in order to reduce the computation time.
If the computation time must be decreased further it is also possible to switch to an explicit solver
when the model is complete.
The boundary layer could be made even ner to capture the viscous behaviour of the boundary
layer better. A y + value lower than 1 is preferred, nevertheless this will not improve the drag of the
airplane. Another improvement for the boundary layer is the transition from the boundary layer
to the outer ow. At gure 5.3d it can be seen that this transition is very rude and this should
be made more smooth. Nevertheless this transition did not cause convergence problems for this
analyses.
A ner grid should be constructed to perform a correct grid renement analysis for the mesh of the
main model. In this article it was only possible to perform a grid renement study for a coarser
grid, which appeared to be almost ne enough but not yet. From this grid renement study it was
assumed that the used mesh for the main model is ne enough, but hard evidence is still needed.
A grid renement study for this analysis is especially important because of the articial dissipation.
This articial dissipation decreases with decreasing mesh size (see section 5.3.4). With help of a
correct grid renement study it is possible to state when this articial dissipation lays in a allowable
range.
Froude's Momentum Theory in combination with the actuator disk is used to simulate the eects
of the engine. But this is a simplied way to simulate it. The fan boundary condition in ANSYS
FLUENT only accepts a uniform pressure jump across the cross section of the actuator disk, so it can
not depend of the radius of the actuator disk. While in the reality the velocity prole at the exhaust
of the engine is more parabolic instead of constant. A way to include this in ANSYS FLUENT is by
adding a momentum source term into the equation of conservation of momentum, but this requires
a lot of experience and knowledge of working with ANSYS. If the engine is simulated like this the
drag caused by the engine would be decreased.
5.8.2 Airplane
It is clear that the drag for this airplane is way too high to be commercially attractive. And also
the lift is a little bit too big.
The wing area is designed in the nal report of W. Muhammad Ilyas Yusof et al. [7]. The design
is based on the optimal ratio of DL during ight, when the wing area is changed this value will also
change. So in that case the whole wing should be reconsidered. The angle of incidence could also
be changed, which will also reduce the lift. But in that case the wing is not operating on its optimal
point. Therefore it is recommended to reduce the wing area, this will also the reduce frontal area.
22
Another thing that stands out if we look at the velocity contour plot of the engine is the wake area
after the connection of the wing to the airplane, see gure 5.7. This wake area should not be present
there and is caused by the rapid change of the shape of the fuselage. This part of the fuselage should
be much smoother, which is widely used in commercial airplanes.
The other wake area that stands out from gure 5.7 is the wake area after the pylon. This wake
area can also be avoided. If the pylon was more shaped like gure 5.8 and 5.9 it would probably
cause less drag. Figure 5.8 and gure 5.9 are based upon the Fairchild Republic A-10 Thunderbolt
II. This pylon will unite the ow in a much smoother way, resulting in a lower pressure dierences
between the ow that is going to the right of the pylon and the ow that is going to the left of the
pylon than the original pylon. This will reduce the drag.
23
Also the front side of the pylon can be improved, there is a big stagnation point. This stagnation
point is caused by the blunt frontal area of the pylon. By reshaping the pylon so that the ow is
more sliced instead of pushed in front or at the side by the pylon the drag will reduce, see gure
5.9.
24
Bibliography
[1] Froude's Momentum Theory. <http://www.public.iastate.edu/ aero442/unit2.pdf>.
[2] Sadraey M. Aircraft Performances Analysis. VDM Verlag Dr. Müller, 2009.
[3] ANSYS Inc. ANSYS FLUENT 14.0 Theory Guide. 2009.
[4] Glenn Elert. The Physics Hypertextbook. <http://physics.info>, 1998-2015.
[5] Kalitzin G., Medic G., Iaccarino, G., Durbin P. Near-wall behahvior of RANS turbulence models
and implications for wall functions. Stanford University, 2004.
[6] LEAP CFD Team. Estimating the First Cell Height for correct Y+.
<http://www.computationaluiddynamics.com.au>, 2013.
[7] Muhammad Ilyas Yusof W., Asyraf Radzali S., Prakobboon Abdul Majid Nazli N.,
Jarrod Woolley P., Yu M., Izzuwan Adnan A., Al-Saad I. Low Noise, Short Range Aircraft.
RMIT.
25
I Appendix
26
(c) Pressure eld along the wings
27
(a) Total velocity eld
28
(c) Velocity eld along the wings
29
(a) Total density eld
30
(c) Density eld along the wings
31
(a) Total temperature eld
32
(c) Temperature eld along the wings
33