Professional Documents
Culture Documents
The first test case is that of an oblique shock on a 15 Constant Input Parameters
degree wedge at Mach 2.5 [1]. In this case the author
There are some solution parameters that remained
compares the results of a simulation using the NASA
constant throughout the project. This is a listing of all
fluid dynamics simulation code WIND with the exact
parameters set by the user in FLUENT that purposely
analytical solution in Anderson (Compressible) [2]. The
remained the same for all data sets. Gravity was
exact solution for supersonic flow over an infinite wedge
neglected. All solutions are for steady state; there were
is known for certain parameters including pressure,
no transient effects included. The density-based solver
density, and temperature ratios across the oblique shock.
was used, as was absolute velocity formulation. The
The author examines several different two-dimensional,
FLUENT material ‘air’ was changed from constant
structured grids varying the spacing.
density to ideal gas to allow for compressible flow
The second test case is a shock on a 10 degree cone at solutions as all cases included in the project were above
Mach 2.35 [3]. While the differing Mach numbers and Mach 0.3. All solutions used the implicit, Roe-FDS,
half angles make a direct comparison difficult, it can be least-squares cell based solver.
seen in the exact results that the shock on a wedge is
Also constant throughout the project was the inlet and
stronger than the shock on a cone at the same Mach and
outlet boundary condition specification. While the
half angle as the ratio of properties across the shock is
operating condition changed, the inlet and outlet of the
notably lower for the cone. This is due to the relief of
mesh were set to the FLUENT ‘Pressure Far-Field’
energy around the cone that is not present in the wedge
boundary type. This boundary is intended for use when
case. The change in geometry is reflected with a change
the boundary is assumed to be at infinity. Using the
in the boundary conditions. Where the wedge is modeled
Pressure Far-Field allows for the direct input of Mach
with a 2-D geometry with an assumed infinite depth, the
number as a solution variable, as opposed to needing to
cone is modeled in 2-D but with an axi-symmetric
input the velocity as calculated based on the desired Mach
boundary at the cone axis of symmetry.
number and free-stream atmospheric conditions. Also,
In the third test case, the 10 degree cone case is the intent of the project is to examine the flow
expanded to a range of cones with a half-angle of 5, 10, characteristics of a vehicle traveling through a quasi-
15, 20, 25, and 30 degrees run at Mach 10. The purpose infinite atmosphere, which is what the Pressure Far-Field
of this case is to compare the pressure coefficient for a is intended to simulate. Throughout this project there are
computational data set versus a slender body references to FLUENT results. Unless otherwise stated,
approximation from Anderson (Hypersonic) [4]. Also FLUENT data is the data produced by the user for this
included in [4] is an experimental data set which matches project.
well with the slender-body approximation up to a cone
half-angle of around 30 degrees.
Supersonic 15 degree Wedge
Literature Review = Δy). Table 1 gives the Δx and Δy values for each grid,
along with the shock wave width (linear measurements in
The literature for this case is a National Aeronautics inches). The maximum Δy is the Δy at the inlet and the
and Space Administration (NASA) National Program for minimum at the inlet and the minimum Δy is the Δy at the
Applications-Oriented Research in CFD (NPARC) outlet (the same number of nodes are along both the inlet
website designed to demonstrate the validation and and outlet). Each grid was run for both a 1st order and 2nd
verification of the WIND CFD code for a 15 degree half- order upwind scheme. The pressure ratio across the shock
angle wedge at Mach 2.5 [1]. As this website is intended was the initial parameter for comparison, but there was no
to provide an example of CFD code verification, it is a appreciable difference in the pressure ratio for any of the
good example of a thorough examination of a CFD code. grids used. Each grid was able to predict the pressure
There are no experimental results referenced at the ratio well, but the size of the shock wave varied greatly;
website, but the exact analytic solution for the case is as the shock width should be zero (for practical
simple enough to be used directly. The only feature in the measurement purposes), this can be used as a direct
flow is the presence of a shock wave; the remainder of the measurement of the error in the solution.
flow is steady and undisturbed. The flow parameters are
constant on either side of the shock and the relationship Table 1: Grid Size and Shock Width for Grid Convergence Study
between the flow characteristics on the two sides of the
max min shock Time
shock is given by a collection of ratios, where the grid size Δx
Δy Δy width req’d
subscript 1 indicates free-stream conditions ahead of the to
inches solve
shock and the subscript 2 indicates conditions behind the
1st <5
shock. order 36x24 0.500 0.500 0.3660 3.100 sec
360x240 0.050 0.050 0.0366 0.310 8 min
Mesh Characteristics and Solution Methods 400x200 0.045 0.060 0.0439 0.189 4 min
2nd <5
This problem used a 400x200 mesh (200 from the order 36x24 0.500 0.500 0.3660 1.240 sec
ramp to the far wall, 400 from the inlet to the outlet), 12
360x240 0.050 0.050 0.0366 0.124 min
which is in the range between the NASA author’s Grid C 8 min
400x200 0.045 0.060 0.0439 0.126
and Grid D (from Wedge Study #2). A structured 2-D
mesh was created by meshing each edge of the domain
with the given number of nodes and allowing Gambit to Figure 1 is a plot of the shock width versus the
mesh the face using the ‘map’ method. The 15 degree longitudinal grid spacing, Δx. Figure 2 is a plot of the
wedge allowed for a mesh with limited skew in the shock width versus the transverse grid spacing Δy. In
elements. The mesh was then brought into FLUENT both cases the error seems to be decreasing at nearly the
where the inviscid assumption was applied. Using a same rate for both the 1st order and 2nd order solutions.
Courant number of 1.0, the solution converged at 1190 This is because the 1st or 2nd order solver was 1st or 2nd
iterations for Mach 2.5 (default FLUENT convergence order in time, both being the same order in space. Using
criteria were used). the 2nd order in time solver does, however, give more
accurate results. In order to estimate the order of
Using almost all the same solver parameters, an error accuracy in time the Courant number should have been
study on the chosen mesh was performed to ensure that varied. This case is steady in time, but changing the
the mesh was created in such a manner as to provide Courant number while holding constant Δx and flow
accurate results. Two grids, aside from the one used for velocity would have the effect of changing the Δt. For the
this case, were created to illustrate error convergence with steady case the Δt is the time between iterations while
decreasing Δx and Δy (x being the axial or longitudinal solving.
coordinate and y being the transverse coordinate). The
first grid was a 36x24 grid, which is approximately an The effect of cell skew can be seen when comparing
order of magnitude coarser than the grid used for the data Figure 1 and 2; the longitudinal grid spacing decreases
collection (400x200), and the second grid was a 360x240 from the 360x240 to the 400x200 grid, but the transverse
grid, which is the same order of magnitude as the grid spacing increases, so the overall accuracy is less
400x200 grid but allows for a square grid in the entry (Δx when using the 2nd order solution. The grid should be
refined in both directions (in 2-D) to gain real benefit analytic solution is labeled as ‘% error’ whereas the
from increasing the number of nodes. This study gives difference between the FLUENT and the WIND code is
some credibility to the use of the 400x200 grid with the labeled as ‘% difference’ as it is a comparison between
2nd order upwind solver for this case. The ~2% gain in two computational solutions that may or may not be
accuracy is not significant for the increase in CPU time accurate. The results show good agreement with both
required when compared to the 360x240 grid. data sets (<5% difference). However, the comparison
between WIND and the analytic solution shows that the
WIND results are <1% different from the analytic
solution. This demonstrates that the FLUENT solution,
while good, could be done better. Given the behavior of
the residuals at the time of convergence, it is quite likely
that a further reduction of the convergence threshold
would yield a more accurate solution. This would lead to
a higher computational cost that may or may not be
worthwhile depending on the accuracy needed. For this
current activity, 5% error is considered sufficiently
accurate.
Mach
Figure 1: Shock Width vs Longitudinal Grid Spacing for Mach 2.5 number
15 deg Wedge Pressure Density Temperature
behind
ratio ratio ratio
the
(p2/p1) (ρ2/ ρ1) (T2/T1)
oblique
shock
NASA WIND
1.8684 2.4696 1.8561 1.3305
data [1]
Further Results
Figure 2: Shock Width vs Transverse Grid Spacing for Mach 2.5 15
deg Wedge As the intent of this project is to examine hypersonic
flow, it was decided to re-run this case at higher Mach
Comparison to Analytic or Experimental Results numbers to see the results. This specific case had low
computational cost, even for higher Mach numbers. The
At convergence, the values for Mach behind the shock
same mesh and initial conditions (with the exception of
as well as the ratio of pressure, density, and temperature
Mach number) were solved for Mach numbers of 3, 4, 5,
across the shock were calculated from the FLUENT
6, and 7. The solution at Mach 3 took 1300 iterations to
results. Table 2 summarizes the results from the solution
solve, which is expected as this is a higher velocity case
at Mach 2.5 with a comparison to both the analytic
and therefore higher property gradients across the shock.
solution from Anderson [2] and the NASA WIND code
It is interesting to note that the solutions at Mach 4, 5, and
results [1]. The difference between the FLUENT and the
6 required fewer iterations to converge to the same
accuracy. Only at Mach 7 were more iterations required, The literature for this case is a National Aeronautics
and at Mach 7 the solution would not converge to the and Space Administration (NASA) National Program for
same value of residuals as the other cases. Table 3 shows Applications-Oriented Research in CFD (NPARC)
the Mach behind the shock and ratio of temperatures website designed to demonstrate the validation and
across the shock for the Mach 2.5 as well as Mach 3-7 verification of the WIND and NPARC CFD codes for a
solutions with the percent error as compared to the 10 degree half-angle cone at Mach 2.35 [3]. This website
Anderson analytic solution [2]. is not nearly as thorough as the 15 degree wedge website
[1], so for a code validation study reference [1] should be
Table 3: Comparison of FLUENT Results to Analytic Solution at
Higher Mach Numbers
utilized. There are no experimental results referenced at
the website, but the exact analytic solution for the case is
simple enough to be used directly. As compared to the
M2 M2 % error
M1 wedge, there is more variability in the flow behind the
(Anderson) (FLUENT) M2
shock. For this reason the flow parameters at the cone
2.5 1.874 1.837 1.98%
surface differ from those just across the shock. For this
3 2.255 2.222 1.48%
case the subscript 1 indicates free-stream conditions,
4 2.930 2.899 1.05% subscript 2 indicates conditions behind the shock, and
5 3.505 3.468 1.05% subscript 3 indicates conditions at the cone surface.
6 3.993 3.953 1.01%
7 4.402 4.348 1.21%
Mesh Characteristics and Solution Methods
condition to use for line ‘B’ in Figure 4 is ‘Wall’, which FLUENT FLUENT
is the same boundary condition used on the ramp for the FLUENT WIND NPARC vs vs
15 degree wedge. Using a Courant number of 1.0, the WIND NPARC
There were a handful of separate meshes used for this Figure 12: Cone-Flare Final Mesh, Close-up
case. The initial mesh included a region behind the cone-
flare, extending approximately 50 mm aft of the body.
This resulted in a large number of iterations (~3000) A structured mesh was created close to the body,
where the low-end absolute pressure limit set in FLUENT while an unstructured mesh was used to fill the remainder
was reached. Even using the FLUENT solution steering of the solution space. The second mesh created (first one
set for hypersonic flow (Courant number variable from ending at the aft end of the cone-flare) used only one
0.5-4.0) the solution took a considerable time to converge. layer of structured mesh near the body. After running the
It was decided that accurately computing the wake in this solution, it was decided that the transition between the
case would require a much longer mesh. The drag finer, structured mesh at the body and the sparser,
increase due to the extremely low pressure region at the unstructured mesh further away from the body was too
aft of the hypersonic geometry is of great interest in the abrupt. A second layer of structured mesh near the body
overall design of hypersonic vehicles, but this specific was added to create a more smooth transition between the
cone-flare geometry is intended to model the forward end structured and unstructured mesh sections. This third
of a vehicle. It was mentioned in the literature review for mesh was the mesh used for the final results for this case.
this case that actual hypersonic vehicle flight conditions Using this mesh, the shock at the leading edge was
would almost certainly be in the turbulent regime. The completely within the structured mesh near the body.
wake properties would depend greatly on a turbulent flow
solution, and the cases chosen for comparison in this Initially the assumed boundary condition at the far-
project are for laminar flow. Taking into account all these wall was set as ‘Wall’ in this analysis. The far-wall was
factors, it was decided to place the mesh outlet at the aft created at a distance away from the cone-flare assumed
end of the cone-flare. Figures 11 and 12 show the mesh sufficient to prevent interaction between the flow at the
used for this case, which used an axi-symmetric 2-D body and any wall effects at the far-wall. After viewing
solver. The minimum node spacing in the mesh at the the contours from the initial solution, however, interaction
body was approximately 0.015-0.02 inches. between the flow around the body and the flow at the far-
wall could be seen. It was decided that a boundary
condition of ‘Symmetry’ at the far-wall would be more
appropriate. A Symmetry boundary condition allows the
flow properties to remain at free-stream at the far-wall as
the properties are assumed symmetric about the far-wall
boundary. Alternatively a slip condition at the far-wall
could have been used to reduce or eliminate the boundary
layer at the far-wall. The desire for this project is to more
directly simulate flight conditions as opposed to exactly
replicating experimental data, so the symmetry condition
is more appropriate for this case.
Figure 11: Cone-Flare Final Mesh, Full View
Three separate simulations were run for the cone-flare
case – inviscid, laminar (to match the experiment and
literature), and k-epsilon (all using the same grid). The
inviscid condition was used for the supersonic wedge,
supersonic cone, and hypersonic cone cases so it was
desired to examine the use of the inviscid case for the
more complex flow. Up to this point there has been no
mention of turbulence models, so a k-epsilon case was run
to see how the results would compare to the laminar case.
Figure 13 shows the residual value versus iteration for the
inviscid solution.
Stagnation Stagnation
Figure 16: Velocity Vectors Showing Recirculation for Laminar Temperature Pressure
Case (K) (kPa)
FLUENT 547 30.90
Analytic [2] 550 31.54
% error 0.55% 2.03%
FLUENT 547 30.90
Ref [5] 549 31.48
% difference 0.36% 1.84%
Further Results
Figure 25: Mach Contours for Mach 15 Case Table 9: Stagnation Temperature Comparison for Mach 6-20
Stagnation
Temperature
(K)
ideal gas
Mach FLUENT % difference
[2]
6 550 547 0.55%