Percent shear area determination in charpy impact testing
AFRL_RQHV_Sagerman_Aviation2016
1. Hypersonic Experimental Aero-thermal Capability
Study through Multilevel Fidelity Computational
Fluid Dynamics
Denton G. Sagerman ∗
University of Dayton, Dayton, Ohio, 45469, USA
James A. Tancred†
Wright-Patterson Air Force Base, Ohio, 45433, USA
Markus P. Rumpfkeil ‡
University of Dayton, Dayton, Ohio, 45469, USA
Barry Hellman §
Wright-Patterson Air Force Base, Ohio, 45433, USA
As true with all hypersonic flight, the ability to quickly and accurately predict the
aerothermodynamic response of an aircraft in the early design phase is important to not
only lower cost, but also to lower the computational and experimental time required to
test such parameters. The Mach 6 High Reynolds Number Facility at Wright-Patterson
Air Force Base in Dayton, Ohio has been non-operational for the past twenty years, but a
recent resurgence in the need for accurate hypersonic test facilities has led to the reacti-
vation of the wind tunnel. With its restoration, the facility is to include new capabilities
to assess hypersonic aerothermodynamic effects on bodies in Mach 6 flow. Therefore, the
objective of this research is to conduct tests in the Mach 6 wind tunnel in which the surface
temperature distribution on tunnel models is determined from temperature sensitive paint
(TSP). Surface pressure and temperature readings, from pressure taps and thermocouples
installed on the models, as well as TSP wall temperature distributions will be used for com-
parison with results from computational fluid dynamics (CFD) analysis codes of differing
fidelity levels. The comparisons can then be utilized to gain confidence in the accuracy of
the aero-thermal response captured at Mach 6 wind tunnel conditions. Two computational
codes will be used in order to validate the aero-thermal capabilities of the wind tunnel: the
Configuration Based Aerodynamics (CBAero) tool set, an inviscid panel code with viscous
approximation capabilities; and the Unstructured Langley Approximate Three-Dimensional
Convective Heating (UNLATCH) code, a boundary-layer approximation solver using flow
solutions from the Euler code Cart3D. The three tunnel model geometries that will be
used for this research are the Reference Flight System model G (RFSG), a Generic Hyper-
sonic Vehicle (GHV), and the Hypersonic International Flight Research Experimentation
Program-Flight 1 (HIFiRE-1) payload geometry. Due to current unresolved issues with
CBAero and UNLATCH, as well as wind tunnel scheduling delays, one-to-one comparisons
of temperature distributions between TSP and the computational codes have not been
obtained yet. However, TSP temperature distributions for the HIFiRE-1 geometry have
been successfully ascertained at Mach 5.85, and verification studies of the inviscid analyses
using CBAero and UNLATCH have been performed and will be presented in this paper.
∗Master’s Candidate, Dept. of Mech. and Aerospace Engineering, sagermand1@udayton.edu, Student Member AIAA
†Aerospace Engineer, Air Force Research Laboratory, James.Tancred@us.af.mil, Member AIAA
‡Assistant Professor, Dept. of Mech. and Aerospace Engineering, Markus.Rumpfkeil@udayton.edu, Senior Member AIAA
§Aerospace Engineer, Air Force Research Laboratory, Barry.Hellman@us.af.mil, Senior Member AIAA
Cleared for Public Release, Distribution Unlimited: 88ABW-2016-2580
1 of 41
American Institute of Aeronautics and Astronautics
2. Nomenclature
a Speed of sound
α or AOA Angle of attack
CD Drag coefficient
CL Lift coefficient
CM Pitching moment coefficient
Cp Pressure coefficient
J Flow functional
M Mach number
ˆn Local surface normal
p Pressure
q Dynamic pressure
R Universal gas constant
Re Reynolds number
T Temperature [Kelvin]
Vsurface Surface velocity vector
V Velocity vector
γ Specific heat ratio
ρ Density
Subscripts
0 Total or stagnation condition
∞ Freestream condition
I. Introduction
The ability to predict and test surface temperatures at hypersonic conditions in the early design stages
can have a significant impact on the success of a design. This research aims to gain confidence in the ability
to obtain aerothermodynamic data from the Mach 6 High Reynolds Number Facility at Wright-Patterson Air
Force Base (WPAFB). An initial goal set for the effort includes the analysis of three geometries: the RFSG,
a GHV, and the HIFiRE-1 payload geometry. For this effort, confidence is postured through the comparison
between TSP temperature distributions from wind tunnel data and those distributions determined through
viscous approximation via the computational codes CBAero and UNLATCH.
All experimental testing for this research will be conducted in the Mach 6 High Reynolds Number Facility,
WPAFB, OH, seen in Figure 1. The test section conditions for the Mach 6 tunnel are shown in Table 1.
Table 1. Mach 6 Wind Tunnel Testing Conditions1
Test Section 12 in. diameter open jet, 17 to 28 inches long
Mach Number 6.0 (5.85 measured during facility calibration)
Temperature Range 440 to 640 degrees F
Reynolds Number Range 10 to 30 million/ft.
Run Time Continuous 5 minutes at 2000 psi
Density Altitude Sim. Simulates Flight from 30,000 ft. to 60,000 ft.
Stagnation Temperature 900 to 1100 Rankine
p0 Range 700 to 2100 psia - exhausting to atmosphere
Although actual testing will be conducted with TSP, pressure sensitive paint (PSP), which has very
similar adhesion capabilities, has been used to test appropriate adhesion due to its availability at the first
time of testing. Two models were evaluated to test appropriate surface roughness. The test models which
can be seen in Figure 2 are similar to the HIFiRE-1 cone geometry, but have no instrumentation and have a
2 of 41
American Institute of Aeronautics and Astronautics
3. Figure 1. Mach 6 High Reynolds Number Wind Tunnel at Wright-Patterson Air Force Base1
fine glass bead finish with a roughness of 32 RA. One model has a blunt nose, while the other has a pointed
nose in order to test the adhesion during flow impact at various surface angles. These tests confirmed that
the paint will stick long enough to the surface at Mach 6 conditions to obtain meaningful measurements.
Figure 2. TSP Adhesion testing models
The outline of the remainder of the paper is as follows. Section II presents the three employed geometries
and Section III gives a quick overview of the computational analysis performed. Section IV details the results
for all three models obtained from CBAero. Section V presents inviscid results for the RFSG and GHV models
whereas Section VI shows results for a coupled Euler/Boundary Layer solver for the HIFiRE-1 geometry.
Finally, Section VII shows preliminary experimental results for the HIFiRE-1 model and comparison to
computational results and Section VIII concludes this paper.
3 of 41
American Institute of Aeronautics and Astronautics
4. II. Geometries
The three geometries chosen for benchmarking are the RFSG reusable booster system, the generic hy-
personic vehicle (GHV), and the HIFiRE-1 all shown in Figure 3. Once again, the goal of the research is
to provide comparisons between wall temperature distributions from TSP to those provided by CBAero and
UNLATCH. The comparisons can give confidence in a new capability of the wind tunnel to measure wall
temperatures at Mach 6. Such measurements can give insight into the thermal response of bodies in Mach 6
flow, which in turn can inform subsequent structural analyses of future design iterations. If performed early
in the design cycle, insights from the measurements may be able to guide designers to avoid unacceptable
outer moldline (OML) decisions that can hinder success of the design later in the design process.
Reusable Booster System Generic Hypersonic Vehicle HIFiRE 1
Figure 3. Hypersonic Geometries to be tested in Mach 6 High Reynolds Number Wind Tunnel2
II.A. RFSG
The Reference Flight System model G (RFSG) was developed under the Air Force Research Laboratory
(AFRL) Future responsive Access to Space Technologies (FAST) program. This vehicle design was developed
based on lessons learned from the FAST program’s three ground technology development efforts. The vehicle
is a reusable first stage which would be paired with an expendable upper stage to provide responsive and
affordable space access. This model was chosen for this research because of the extensive aerodynamic
database compiled for this model as well as its unique geometric features such as the wing-body combination
and wing tip vertical fins. An instrumented drawing of the RFSG geometry as well as the wind tunnel model
itself can be seen in Figures 4 and 5, respectively.
Figure 4. RFSG wind tunnel model with instrumentation
4 of 41
American Institute of Aeronautics and Astronautics
5. Figure 5. RFSG wind tunnel model
II.B. Generic Hypersonic Vehicle
In order to foster collaboration between government and private hypersonic communities, the AFRL created
a family of Generic Hypersonic Vehicles (GHV) that are publicly released. This allows for data to be more
easily disseminated among those in the hypersonic community in order to advance the field. The GHV is
a slender, low aspect ratio wing-body configuration enclosing a supersonic combustion ramjet (SCRAMjet)
engine with an inward-turning inlet. The vehicle is conceptually designed to have a metallic structure with
a thermal protection system capable to withstand Mach 6 cruise conditions.3
The model was chosen for its
complexity as well as its ease of distribution within the community. The GHV model used for this research
has an added stingmount to allow for compatibility with the sting available in the Mach 6 tunnel. The
stingmount adaptation to the GHV was incorporated in the computational modeling for this research. The
top and side views of the model with instrumentation are shown in Figure 6.
Top view
Side view
Figure 6. GHV wind tunnel model with instrumentation
5 of 41
American Institute of Aeronautics and Astronautics
6. II.C. HIFiRE-1
The Hypersonic International Flight Research Experimentation (HIFiRE) Program is a bi-lateral hypersonic
flight test program between AFRL and the Australian Defense Science and Technology Group (DTSG),
formerly the Australian Defense Science and Technology Organization (DSTO). HIFiRE Flight 1 took place
on March 22, 2010 to measure boundary layer transition (BLT) and shock boundary layer interaction (SBLI)
at Mach 7 flight conditions.4
Figure 7 exhibits the dimensioned HIFiRE-1 wind tunnel model to be used
in this research, with the positions of three pressure taps and two thermocouples shown as well. The
thermocouples will be used to calibrate the TSP on the vehicle’s surface. The finished wind tunnel model
used in testing can be seen in Figure 8.
Figure 7. HIFiRE-1 wind tunnel model with instrumentation
Figure 8. HIFiRE-1 wind tunnel model
III. Computational Analysis
To gain confidence in the experimental results obtained from the Mach 6 wind tunnel, computational
analysis tools were employed. The two tools used were the Configuration Based Aerodynamics (CBAero)
tool set and the Unstructured Langley Approximate Three-Dimensional Convective Heating (UNLATCH)
code. Both analysis tools were developed by the National Aeronautics and Space Administration (NASA).
The use of UNLATCH requires flow solutions from the Euler solver Cart3D.5,6
Prior to reporting results from
CBAero and UNLATCH, grid convergence studies were performed on the computational meshes built for each
6 of 41
American Institute of Aeronautics and Astronautics
7. analysis. Final meshing strategies were chosen from the grid convergence studies that, when used for analysis,
incurred small discretization errors in the integrated quantities of interest derived from flow solutions. For
CBAero, grid convergence studies were performed that monitor inviscid CL and CD due to input surface
meshes for the geometries of interest. For UNLATCH, Cart3D meshing strategies were evaluated such that
final CL and CD integrated quantities no longer change with increasing Cartesian volume grid resolution.
Once meshing strategies were chosen for each tool, attempts were made to obtain surface wall temperatures
for each geometry. Details will be given in the next three sections.
IV. CBAero
The CBAero software package is an engineering-level inviscid flow solver that incorporates various panel
methods in order to evaluate the aerodynamic and aerothermodynamic response of arbitrary vehicle config-
urations at supersonic and hypersonic speeds. A beneficial attribute of the code is its use of unstructured
surface triangulations to define input geometry.7
The use of the term panel denotes that an integrated quan-
tity obtained from one surface cell of a geometry is independent of any quantity obtained from any other
cell. In other words, there is no flow interaction or communication from one cell to the next. To provide an
estimate of the surface velocity vectors at hypersonic speeds, Equation (1) can be used.
Vsurface = ˆn × V∞ × ˆn (1)
Various inviscid panel methods are available for analysis such as Modified Newtonian, Tangent Cone, and
Tangent Wedge.8
The methods determine approximated values of pressure coefficients acting normally on
each cell of the geometry. A different method may be assigned to different sets of cells or components
that, when subjected to a freestream velocity, more closely align with the assumptions of an available panel
method. Once the inviscid solution is obtained from a chosen panel method strategy, CBAero provides its
own viscous approximation to estimate surface skin friction. From the skin friction estimate, heating rates
and temperatures at the body wall may be determined.9
For this effort, the Modified Newtonian method is used for the inviscid approximation of surface pressures
on the RFSG, GHV, and HIFiRE-1 geometries. The conditions for CBAero analysis, corresponding to Mach
6 wind tunnel conditions, are shown in Table 2. The grid convergence studies and inviscid analysis results
are shown in Sections IV.A- IV.C for each geometry.
Table 2. CBAero Flow Analysis Conditions
Case 1 Case 2 Case 3 Case 4
P0 [psi] 700 700 2000 2000
T0 [Rankine] 900 1000 900 1000
T∞ [Kelvin] 63.7 70.8 63.7 70.8
ρ∞ [ kg
m3 ] 0.1951 0.1756 0.5575 0.5017
IV.A. RFSG
Grid convergence studies were performed with four different input surface meshes for the RFSG geometry.
The family of surface meshes, along with each respective cell count, used for the RFSG model in the CBAero
analysis is shown in Figure 9. The grid convergence study for the RFSG geometry using CBAero is shown
in Figure 10, where both CL and CD are shown to converge with increasing number of surface cells. Using
the most refined mesh having 189, 044 cells, the inviscid Cp distribution for the RFSG is shown in Figure 11
for Case 1 of Table 2, at −10 and 0 degrees AOA.
7 of 41
American Institute of Aeronautics and Astronautics
8. Ncells = 36, 934 Ncells = 43, 390
Ncells = 78, 814 Ncells = 189, 044
Figure 9. CBAero surface meshes for the RFSG geometry2
Ncells
104
105
106
CL
-0.18
-0.175
-0.17
AOA = -10
qD
= 0.85 Bars
qD
= 2.44 Bars
Ncells
104
105
106
CL
-0.11
-0.105
-0.1
AOA = -5
Ncells
104
105
106
CL
-0.04
-0.038
-0.036
AOA = 0
Ncells
104
105
106
CL
0.03
0.035
0.04
AOA = 5
Ncells
104
105
106
CL
0.095
0.1
0.105
AOA = 10
RFSG CBAero Grid Convergence Study for CL
CL
Ncells
104
105
106
CD
0.114
0.116
0.118
AOA = -10
q
D
= 0.85 Bars
q
D
= 2.44 Bars
Ncells
104
105
106
CD
0.08
0.085
AOA = -5
Ncells
104
105
106
CD
0.064
0.066
0.068
AOA = 0
Ncells
104
105
106
CD
0.055
0.06
0.065
AOA = 5
Ncells
104
105
106
CD
0.065
0.07
0.075
AOA = 10
RFSG CBAero Grid Convergence Study for CD
CD
Figure 10. CBAero Grid Convergence Study for RFSG geometry for various surface meshes
8 of 41
American Institute of Aeronautics and Astronautics
9. α = 0 degrees α = −10 degrees
Figure 11. Cp distribution from CBAero for RFSG2
IV.B. GHV
A total of five surface meshes for the GHV geometry were used for a CBAero grid convergence study. These
GHV grid models can be seen in Figure 12. Due to the assumption of panel analysis that no cell interacts
with one another, a geometry having an inner moldline (IML), such as a pipe, tube, inlet, or engine, causes
the panel analysis to output erroneous results for the IML. This is because there is no mechanism, under
the former assumption, for the flow to be redirected within an enclosed space, such as reflection or turning
of flow within an engine. The only valid output from CBAero panel analysis for the GHV is from the OML,
which includes the fuselage, wings, and tails. However, the grid convergence study performed for the GHV
with CBAero reports CL and CD for all cells of the geometry, including OML and IML. This was done
to simplify the analysis and to prevent the complexity of isolating the OML from the IML. It is assumed
in this grid convergence study that if CL and CD of the OML and IML have small changes for increasing
surface mesh resolution, then the surface mesh of high resolution identified in the study may be used for
panel analysis of the model. A later comparison of Cp at a cross-section of the wing between CBAero and
Cart3D (which accounts for flow interaction) shows good agreement between the Cp distributions of the two
codes, giving confidence for this grid convergence study. The results of the CBAero grid convergence study
for the GHV are shown in Figure 13. From the grid convergence study, a surface mesh of 221, 454 cells was
determined to produce grid-converged values of CL and CD for the GHV. Lastly, this surface mesh was used
to run CBAero to produce inviscid results (again reporting OML and IML to reduce analysis complexity),
which are displayed in the form of Cp contours in Figure 14 again for Case 1 of Table 2, at −10 and 0 degrees
AOA.
9 of 41
American Institute of Aeronautics and Astronautics
10. Ncells = 135, 604 Ncells = 144, 082
Ncells = 174, 018 Ncells = 221, 454
Ncells = 478, 394
Figure 12. CBAero surface meshes for the GHV geometry2
10 of 41
American Institute of Aeronautics and Astronautics
11. Ncells #105
2 3 4
CL
-0.1
-0.09
-0.08
AOA = -10
q
D
= 0.85 Bars
q
D
= 2.44 Bars
Ncells #105
2 3 4
CL
-0.04
-0.02
0
AOA = -5
Ncells #105
2 3 4
CL
0
0.02
0.04
AOA = 0
Ncells #105
2 3 4
CL
0.06
0.07
0.08
AOA = 5
Ncells #105
2 3 4
CL
0.12
0.125
0.13
AOA = 10
GHV CBAero Grid Convergence Study for CL
CL
Ncells #105
2 3 4
CD
0.02
0.025
AOA = -10
q
D
= 0.85 Bars
q
D
= 2.44 Bars
Ncells #105
2 3 4
CD
0.005
0.01
0.015
AOA = -5
Ncells #105
2 3 4
CD
0.005
0.01
0.015
AOA = 0
Ncells #105
2 3 4
CD
0.014
0.016
0.018
AOA = 5
Ncells #105
2 3 4
CD
0.036
0.038
0.04
AOA = 10
GHV CBAero Grid Convergence Study for CD
CD
Figure 13. CBAero Grid Convergence Study for GHV geometry for various surface meshes
α = 0 degrees α = −10 degrees
Figure 14. Cp distribution from CBAero for GHV2
IV.C. HIFiRE-1
In order to determine a surface mesh resolution that produces grid-converged, integrated CL and CD values
for the HIFiRE-1 payload, four surface meshes were generated that are shown in Figure 15. Convergence
of lift and drag coefficients becomes apparent at resolutions near the 24, 064-cell mesh that was studied.
However, a mesh of 80, 384 cells, the highest mesh resolution for the HIFiRE-1 CBAero grid convergence
study, was used for further analysis at little extra computational cost.
Qualitative viscous results were obtained from CBAero with the HIFiRE-1 model. These results are
shown in Figure 17 in the form of heating rate distributions on the body at 0 and 10 degrees angle-of-attack
for Cases 1 and 4 of Table 2. Contours are reported in W
cm2 . Note that there is higher heating in stagnation
regions such as the nose tip and the ramp aft of the constant-area section of the payload. Also, Case 4
shows higher heating than Case 1, which is an expected result due to the higher temperature and pressure
condition of Case 4.
11 of 41
American Institute of Aeronautics and Astronautics
12. Ncells = 7, 184 Ncells = 24, 064
Ncells = 48, 224 Ncells = 80, 384
Figure 15. CBAero surface meshes for HIFiRE-1 geometry2
Ncells
103
104
105
CL
-0.4
-0.35
AOA = -10
qD
= 0.85 Bars
qD
= 2.44 Bars
Ncells
103
104
105
CL
-0.26
-0.24
-0.22
AOA = -5
Ncells
103
104
105
CL
#10-3
-5
0
5
AOA = 0
Ncells
103
104
105
CL
0.16
0.18
0.2
AOA = 5
Ncells
103
104
105
CL
0.35
0.4
AOA = 10
HIFiRE 1 CBAero Grid Convergence Study for CL
at Notch 90 deg
CL
Ncells
103
104
105
CD
0.34
0.35
0.36
AOA = -10
q
D
= 0.85 Bars
q
D
= 2.44 Bars
Ncells
103
104
105
CD
0.235
0.24
0.245
AOA = -5
Ncells
103
104
105
CD
0.265
0.27
0.275
AOA = 0
Ncells
103
104
105
CD
0.285
0.29
0.295
AOA = 5
Ncells
103
104
105
CD
0.35
0.36
0.37
AOA = 10
HIFiRE 1 CBAero Grid Convergence Study for CD
at Notch 90 deg
CD
Figure 16. CBAero Grid Convergence Study for HIFiRE-1 geometry with Notch at 90 deg for various surface meshes
12 of 41
American Institute of Aeronautics and Astronautics
13. Case 1, α = 0 degrees Case 1, α = 10 degrees
Case 4, α = 0 degrees Case 4, α = 10 degrees
Figure 17. Surface heating rates from CBAero for Notch at 90 deg2
IV.D. Limitations in CBAero Analysis
Some difficulties arose in the analysis of the RFSG, GHV, and HIFiRE-1 geometries when using CBAero.
Firstly, providing a surface mesh that was not closed, or water-tight, caused issues with the viscous solver
of CBAero. If the mesh retained open seams or holes, the solver could not easily integrate over the body to
produce friction and heating approximations. Secondly, due to the assumption of a panel analysis discussed
in Section IV.B, CBAero is rendered invalid as an analysis tool for a full configuration that includes an
engine, such as the GHV. If only the OML is considered, the panel analysis is valid. However, a mesh
that excludes its IML is then no longer water-tight. Finally, initial results of wall temperature distributions
for the RFSG and the HIFiRE-1 geometries showed signs of either over-predicting temperature by a large
margin or indicated asymmetry in the distributions even when no side-slip was involved. Causes for the
over-prediction and asymmetries in the temperature distributions were not determined during the period of
performance of this effort, and therefore temperature results are excluded from this work.
13 of 41
American Institute of Aeronautics and Astronautics
14. V. Cart3D Inviscid Flow Solver
Medium-fidelity computational flow simulations were performed using Cart3D, a software suite employing
automated and error-adapted5,6
Cartesian volume mesh generation to solve the Euler flow equations. While
monitoring a user-specified functional, such as lift, drag, or pitching moment, Cart3D uses ”adjoint-weighted”
error estimation5,10
to cluster nodes in regions of the flow-field where error in the flow solution degrades the
accuracy of the chosen functional. In this effort, the value of pitching moment coefficient, Cm, was monitored
to drive adjoint-weighted mesh refinement of the off-body flow-field for the HIFiRE-1 payload. A combined
functional J = 0.6CD + 0.4CL was used to drive the adaptations for the RFSG and GHV geometries.
Using the mesh adaptation capability of Cart3D, grid convergence studies were performed for the RFSG,
GHV, and HIFiRE-1 models. The integrated quantities of CL and CD were monitored for various adaptation
levels and initial volume grid resolutions. Grid-converged solutions were reached but a discussion of the
Cart3D grid convergence study for the HIFiRE-1 payload is left to Section VI. The eventual goal for
employing Cart3D is to provide grid-converged Euler flow solution data to UNLATCH.
V.A. RFSG
To conduct the Cart3D grid convergence study of the RFSG, four initial volume meshes were staged to
reach five different levels of mesh adaptation. A surface mesh of 461, 492 cells was first created to input
into Cart3D’s automatic Cartesian mesh generator, cubes. The surface mesh is shown in Figure 18. Cubes
then takes the surface mesh and creates an initial volume grid, denoted as Nstartcells, of Cartesian cells
such that the surface mesh provides a wall boundary that defines the geometry at hand. Adjoint-adapted
solutions were then obtained from Cart3D by staging the initial volume mesh from five to nine adaptations.
The volume meshes of three different angles-of-attack after nine adaptations for the RFSG are showcased in
Figures 19 and 20. Note how the adaptation of the mesh is capable of resolving flow features and highly
contoured regions, such as the leading bow shock and nose of the vehicle. For completeness, the residual
histories for the RFSG geometry are shown in Figure 21, showing that, by the ninth adaptation, the residual
drops by approximately three to four orders of magnitude. To further showcase the flow features about
the vehicle, Cp, Mach, and static density contours are given for the same angle-of-attack values in Figures
22-24. Stagnation can be seen readily at the nose and wing leading edges in the surface Cp contours while
strong flow interactions are apparent near the nose in the Mach and density fields. Finally, the results of
the grid convergence study of the RFSG are shown in Figures 25-27 for CL, CD, and CM , respectively. Grid
convergence is reached near about one million cells, corresponding to roughly seven to nine adaptations for
starting meshes greater than Nstartcells = 5, 500 cells.
Figure 18. RFSG Surface Mesh with Ncells = 461, 492
14 of 41
American Institute of Aeronautics and Astronautics
15. α = 0, Ncells = 616, 619 α = 5, Ncells = 960, 120 α = 10, Ncells = 910, 778
Figure 19. RFSG Volume Meshes through y − cutplane2
α = 0, Ncells = 616, 619 α = 5, Ncells = 960, 120 α = 10, Ncells = 910, 778
Figure 20. RFSG Volume Meshes through x and y − cutplane2
Solution Iteration
log(L1Residual)
0 1000 2000 3000
-10
-5
0
5
10
M = 6, α = 0 degrees
Solution Iteration
log(L1Residual)
0 1000 2000 3000
-10
-5
0
5
10
M = 6, α = 5 degrees
Solution Iteration
log(L1Residual)
0 1000 2000 3000
-10
-5
0
5
10
M = 6, α = 10 degrees
Figure 21. RFSG convergence histories after 9 adaptations2
M = 6, α = 0 degrees M = 6, α = 5 degrees M = 6, α = 10 degrees
Figure 22. RFSG Cp contours2
15 of 41
American Institute of Aeronautics and Astronautics
16. M = 6, α = 0 degrees M = 6, α = 5 degrees M = 6, α = 10 degrees
Figure 23. RFSG Mach contours2
M = 6, α = 0 degrees M = 6, α = 5 degrees M = 6, α = 10 degrees
Figure 24. CART3D ρ contours2
16 of 41
American Institute of Aeronautics and Astronautics
20. A final comparison for the grid convergence study is given between Cart3D and CBAero results about a
centerline cutplane. Such a comparison gives confidence that both analysis codes are producing acceptable
results that reach proper magnitudes in the integrated quantities of interest. For comparison a centerline cut
of the RFSG was taken to extract the pressure coefficient value found along the cut for flow conditions at
Mach 5.85 and angle-of-attack values of 0, 5, and 10 degrees. Very acceptable agreement between upper and
lower surface Cp distributions along the centerline was determined from Cart3D and CBAero, which is shown
by the close data trends in Figure 28. The cutplane itself is shown in the same figure. This gives confidence
that both Cart3D and CBAero produced grid-converged inviscid quantities for the RFSG geometry.
α = 0
Position along x-axis [in]
0 2 4 6 8 10
Cp
-0.2
0
0.2
0.4
0.6
0.8
1
RFSG Cp
comparison along y-axis cutplane at AOA=0
CART3D
CBAero
Upper Surface
Lower Surface
α = 0
Position along x-axis [in]
0 2 4 6 8 10
Cp
-0.2
0
0.2
0.4
0.6
0.8
1
RFSG Cp
comparison along y-axis cutplane at AOA=5
CART3D
CBAero
Upper Surface
Lower Surface
α = 5
Position along x-axis [in]
0 2 4 6 8 10
Cp
-0.2
0
0.2
0.4
0.6
0.8
1
RFSG Cp
comparison along y-axis cutplane at AOA=10
CART3D
CBAero
Lower Surface
Upper Surface
α = 10
Figure 28. CART3D vs. CBAero for RFSG along y-cutplane also shown
20 of 41
American Institute of Aeronautics and Astronautics
21. V.B. GHV
In a similar fashion to the RFSG, the GHV was analyzed using Cart3D during a grid convergence study.
Four initial volume grids were staged through five different adjoint-guided adaptations. Adaptation levels
ranged from five to nine adaptations. An input surface mesh of 209, 196 cells was created for input into
cubes. This input mesh is provided in Figure 29. The residual dropped approximately three to four orders of
magnitude. Examples of the mesh adaptation and the residual reduction are again shown for flow conditions
at Mach 5.85 and angle-of-attack values of 0, 5, and 10 degrees in Figures 30-31 and Figure 32, respectively.
Flow features are also illustrated, respectively, with Cp, Mach, and density contours in Figures 33-35. Grid
convergence results are shown in Figures 36-38 for CL, CD, and CM . The final determination from the GHV
grid convergence study with Cart3D was that grid-converged integrated quantities are reached, at least,
when seven adaptations are used on an initial Cartesian volume grid with cell number greater than roughly
Nstartcells = 5400.
Again, a comparison is shown between Cart3D and CBAero for a Cp distribution along a cutplane of the
geometry. A cut was made along the port (left) wing of the GHV, and the Cp distribution was extracted.
The cutplane can be viewed in Figure 39, and the Cp distributions on upper and lower surfaces for 5 and
10 degrees AOA are shown in Figure 40. There is discrepancy between the distributions from both analysis
codes. However, trends are quite similar. At these angles-of-attack, the flow can be highly separated,
which is not modeled by either code. Cart3D can capture three-dimensional effects, but cannot accurately
model separation since separation is a highly viscous phenomenon. CBAero has no ability to model three-
dimensional flow effects or even communication of flow from one point to the next. Therefore, even though
there is discrepancy in the two distributions, there is still confidence that the proper trends are being resolved
by both codes.
Figure 29. GHV Surface Mesh with Ncells = 209, 196
21 of 41
American Institute of Aeronautics and Astronautics
22. α = 0, Ncells = 588, 407 α = 5, Ncells = 920, 597 α = 10, Ncells = 788, 754
Figure 30. GHV Volume Meshes through y − cutplane2
α = 0, Ncells = 588, 407 α = 5, Ncells = 920, 597 α = 10, Ncells = 788, 754
Figure 31. GHV Volume Meshes through x and y − cutplane2
Solution Iteration
log(L1Residual)
0 1000 2000
-10
-5
0
5
10
M = 6, α = 0 degrees
Solution Iteration
log(L1Residual)
0 1000 2000
-10
-5
0
5
10
M = 6, α = 5 degrees
Solution Iteration
log(L1Residual)
0 1000 2000
-10
-5
0
5
10
M = 6, α = 10 degrees
Figure 32. GHV convergence histories after 8 adaptations2
M = 6, α = 0 degrees M = 6, α = 5 degrees M = 6, α = 10 degrees
Figure 33. GHV Cp contours2
22 of 41
American Institute of Aeronautics and Astronautics
23. M = 6, α = 0 degrees M = 6, α = 5 degrees M = 6, α = 10 degrees
Figure 34. GHV Mach contours2
M = 6, α = 0 degrees M = 6, α = 5 degrees M = 6, α = 10 degrees
Figure 35. GHV ρ contours2
23 of 41
American Institute of Aeronautics and Astronautics
27. Figure 39. GHV Slice along port side wing cutplane
Position along x-axis [in]
6.5 7 7.5 8 8.5 9 9.5 10
Cp
-0.05
0
0.05
0.1
0.15
0.2
GHV Cp
comparison along y-cutplane of wing at AOA=5
CART3D
CBAero
Lower Surface
Upper Surface
α = 5
Position along x-axis [in]
6.5 7 7.5 8 8.5 9 9.5 10
Cp
-0.05
0
0.05
0.1
0.15
0.2
GHV Cp
comparison along y-cutplane of wing at AOA=10
CART3D
CBAero
Lower Surface
Upper Surface
α = 10
Figure 40. CART3D vs. CBAero for GHV along y-cutplane of wing
27 of 41
American Institute of Aeronautics and Astronautics
28. VI. UNLATCH Boundary Layer Solver
The ability to predict convective heating over a three-dimensional body at hypersonic speeds is very
important in the early stages of hypersonic vehicle design. The UNLATCH (Unstructured Langley Approx-
imate Three-Dimensional Convective Heating) engineering code can be used to calculate the heating rates
on general three-dimensional bodies using inviscid flow field results from an unstructured flow solver, such
as Cart3D. This is done by reducing the three-dimensional boundary-layer equations to the same form as
the axisymmetric boundary-layer equations.11
Given the inviscid flow solution from Cart3D for the HIFiRE-1 payload geometry, the following cases in
Table 3 were run using UNLATCH in order to approximate the viscous effects of the boundary layer on the
body. These are the same conditions at which the wind tunnel experiments were ran. All cases were ran
with a constant wall temperature of 300K.
As a note, there was considerable difficulty in reaching flow solutions from Cart3D that could be suc-
cessfully input into UNLATCH. Stagnation regions and areas of high flow interaction must be resolved to
a significant degree in order for UNLATCH to begin its viscous approximation. As a result, a very large
number of trades in Cart3D adaptation strategy were made in order to find a combination of adaptation and
resolution that could be acceptable for UNLATCH. Due to this difficulty, only solutions for the HIFiRE-1
geometry were successfully input into UNLATCH at this point.
Table 3. UNLATCH Run Conditions
Case 1 Case 2 Case 3 Case 4
V∞ [m
s ] 936 986 936 986
p∞ [m
s ] 3570 3570 10200 10200
T∞ [K] 63.7 70.8 63.7 70.8
VI.A. Grid Generation
In order to sufficiently resolve the flow solution about the HIFiRE-1 payload geometry, a different Cart3D
adaptation strategy had to be employed. The stagnation region near the nose of the geometry had to be
discretized with enough resolution such that the resulting solution on the surface was relatively smooth. To
do this, the nose itself was identified as a separate component from the body. When the initial volume mesh
was generated, refinement of the grid was denoted to take place at the nose, allowing the adjoint refinement
of the mesh to identify the nose specifically as a location to cluster cells. The surface mesh used for building
the volume grid is shown in Figure 41. The volume grid with higher resolution near the nose is shown in
Figures 42 and 43. Note that an added benefit to this meshing approach is the high resolution of the shock
structure surrounding the body. A downside, however, is that the cell count increases rapidly with each
adaptation, since the starting mesh tends to be much larger.
Figure 41. HIFiRE-1 Surface Mesh used for CART3D
28 of 41
American Institute of Aeronautics and Astronautics
29. α = 0, Ncells = 6, 408, 869 α = 5, Ncells = 6, 688, 258 α = 10, Ncells = 6, 456, 957
Figure 42. CART3D Volume Meshes through y − cutplane2
α = 0, Ncells = 6, 408, 869 α = 5, Ncells = 6, 688, 258 α = 10, Ncells = 6, 456, 957
Figure 43. CART3D Volume Meshes through x and y − cutplane2
VI.B. HIFiRE-1 Grid Convergence Study and UNLATCH Results
Using this new meshing strategy, a grid convergence study was conducted as performed in Section V. In this
case, due to the extra computational cost of the larger volume grids, three initial volume grids were staged
through seven different adaptation levels. Flow feature are again showcased in Figures 44-46. Integrated
quantities of CL, CD, and CM from the study are shown in Figures 47-49. Residual reduction was only
reduced to approximately two orders of magnitude in this study. Time for this effort ran out before a root
cause for the low residual reduction could be determined. Even though the residuals are higher than preferred
in CFD studies (desiring reductions to at least four orders of magnitude for hypersonic flow and preferably
eight orders of magnitude in most cases), the integrated quantities do not appear to suffer from this. In fact,
the coefficients monitored reach values that experience negligible changes at eight and nine adaptations for
initial volume grids with cell counts greater than roughly 40, 000 Cartesian cells.
M = 5.85, α = 0 degrees M = 5.85, α = 5 degrees M = 5.85, α = 10 degrees
Figure 44. CART3D Cp contours for HIFiRE-1 geometry2
29 of 41
American Institute of Aeronautics and Astronautics
30. M = 5.85, α = 0 degrees M = 5.85, α = 5 degrees M = 5.85, α = 10 degrees
Figure 45. CART3D Mach contours for HIFiRE-1 geometry2
M = 5.85, α = 0 degrees M = 5.85, α = 5 degrees M = 5.85, α = 10 degrees
Figure 46. CART3D ρ contours for HIFiRE-1 geometry2
30 of 41
American Institute of Aeronautics and Astronautics
34. Using this insight, a Cart3D solution reaching eight adaptations was input into UNLATCH. The heating
rate contours, in W
cm2 , predicted by UNLATCH for the HIFiRE-1 geometry are shown in Figure 50. When
compared to the heating rates from the CBAero analysis, shown in Figure 17, the overall magnitude is
similar. The shadowed, or leeward, facing regions at positive angle-of-attack were over-predicted by the
panel code with viscous approximations. Unfortunately, due to a bug in the UNLATCH source code, output
of surface wall temperature from UNLATCH is unavailable.
Case 1, α = 0 degrees Case 4, α = 0 degrees
Case 1, α = 10 degrees Case 4, α = 10 degrees
Figure 50. UNLATCH solution for surface heating rates for Notch at 90 degrees2
34 of 41
American Institute of Aeronautics and Astronautics
35. VII. Preliminary Experimental Results for HIFiRE 1 Geometry
With the Mach 6 wind tunnel remaining unused for the past 20 years, a resurgence in the need for
accurate high speed testing facilities has resulted in the wind tunnel becoming operational once again. A
point of interest in hypersonic testing as previous noted, is the surface temperature distribution which can
be obtained from using Temperature Sensitive Paint (TSP), in which images taken from a CCD array will
capture the instantaneous change in color of the paint on the surface.
The following test matrix, shown in Table 4, was used in testing the HIFiRE-1 geometry in the Mach
6 wind tunnel. This matrix was first completed without TSP to have clean readings of the pressure taps
and thermocouples without the interference of the paint in the instrumentation holes. Future tests will be
completed, in accordance with the same test matrix, with the use of TSP while continuing to take readings
from the instrumentation.
Table 4. Mach 6 Wind Tunnel Testing Matrix
Total Pressure (P0) Total Temperature (T0) Orientation
700 900 Notch Starboard
700 900 Notch 0
700 1000 Notch Starboard
700 1000 Notch 0
900 900 Notch Starboard
900 900 Notch 0
900 1000 Notch Starboard
900 1000 Notch 0
VII.A. Surface Pressure Readings
A sample of the results from initial testing can be seen for two orientations of the HIFiRE-1 geometry, with
the notch at 0 degrees from the z-axis and with the model rotated 90 degrees so the notch is facing starboard
along the y-axis. Results are shown for the pressure coefficient based on the test section dynamic pressure,
which was calculated using isentropic conditions, and temperature readings from the thermocouples, shown
in Figure 51. As expected the highest pressure readings come from the high inclination region at the flare
(Pw 3) and the temperature increases throughout the run at all positions. The TSP will be calibrated by
taking the temperature reading the the exact point at which the image is taken in reference to the initial
image which accounts for changes in the paint due to lighting in the tunnel with no air.
35 of 41
American Institute of Aeronautics and Astronautics
36. 6 8 10 12 14 16
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
I
Cp
= 0.1593
I
Time [sec]
C
p
Run: 10 P
0
= 696 psi T
0
= 913 R AOA = −10
6 8 10 12 14 16
−12
−10
−8
−6
−4
−2
0
2
4
6
8
10
12
Inject/Alpha[deg]
P
w
− 1
Alpha
Inject
Pressure Tap 1
6 8 10 12 14 16
−0.5
−0.43
−0.36
−0.29
−0.22
−0.15
−0.08
−0.01
0.06
0.13
0.2
I
C
p
= 0.045828
I
Time [sec]
C
p
Run: 10 P
0
= 696 psi T
0
= 913 R AOA = −10
6 8 10 12 14 16
−12
−10
−8
−6
−4
−2
0
2
4
6
8
10
12
Inject/Alpha[deg]
P
w
− 2
Alpha
Inject
Pressure Tap 2
6 8 10 12 14 16
0.5
0.63
0.76
0.89
1.02
1.15
1.28
1.41
1.54
1.67
1.8
I
C
p
= 1.5056
I
Time [sec]
C
p
Run: 10 P
0
= 696 psi T
0
= 913 R AOA = −10
6 8 10 12 14 16
−12
−10
−8
−6
−4
−2
0
2
4
6
8
10
12
Inject/Alpha[deg]
P
w
− 3
Alpha
Inject
Pressure Tap 3
6 8 10 12 14 16
520
540
560
580
600
Time [sec]
TC1[R]
Run: 10 P
0
= 696 psi T
0
= 913 R AOA = −10
6 8 10 12 14 16
−12
−10
−8
−6
−4
−2
0
2
4
6
8
10
12
Inject/Alpha[deg]
T
c
− 1
Alpha
Inject
Thermocouple 1
6 8 10 12 14 16
530
540
550
560
Time [sec]
TC2[R]
Run: 10 P
0
= 696 psi T
0
= 913 R AOA = −10
6 8 10 12 14 16
−12
−10
−8
−6
−4
−2
0
2
4
6
8
10
12
Inject/Alpha[deg]
T
c
− 2
Alpha
Inject
Thermocouple 2
Figure 51. Mach 6 Wind Tunnel Results for P0 = 700 psi T0 = 900 R at AOA = -10 with Notch starboard
36 of 41
American Institute of Aeronautics and Astronautics
37. VII.B. Temperature Sensitive Paint
The temperature sensitive paint (TSP) is calibrated by Innovative Scientific Solutions Incorporated (ISSI)
for certain temperature ranges, where a second-order accurate curve fit is applied to the calibration data
based on luminosity. The temperature is a function of the ratio of wind-off to wind-on luminosity. The paint
is applied as a thin layer with the assumption that the TSP is sensing the local surface temperature. TSP
results for the HIFiRE-1 geometry are shown in Figure 52- 56. With the camera mounted on the top of the
wind tunnel, the windward cases correspond to negative AOA and the leeward cases correspond to positive
AOA. The temperature contour from the TSP is shown with the predicted temperature along the centerline
shown as the green line. Two images at different times were taken for each AOA. The exact temperature
from the thermocouples is also shown as a green star, which shows how accurate the TSP readings are when
compared to the thermocouple readings.
Picture #2, CL+10 sec
α=−10
Centerline 2 sec.
Picture #2, CL+10 sec
α=−10
Centerline 8 sec.
Figure 52. Mach 6 Wind Tunnel TSP Results for P0 = 700 psi T0 = 900 R for HIFiRE-1 geometry at α = −10
37 of 41
American Institute of Aeronautics and Astronautics
38. Picture #10, CL+3 sec Picture #20, CL+8 sec
α=−5
Centerline 2 sec.
Picture #10, CL+3 sec Picture #20, CL+8 sec
α=−5
Centerline 8 sec.
Figure 53. Mach 6 Wind Tunnel TSP Results for P0 = 700 psi T0 = 900 R for HIFiRE-1 geometry at α = −5
Picture #10, CL+3 sec Picture #20, CL+8 sec
α=0
Centerline 2 sec.
Picture #10, CL+3 sec Picture #20, CL+8 sec
α=0
Centerline 8 sec.
Figure 54. Mach 6 Wind Tunnel TSP Results for P0 = 700 psi T0 = 900 R for HIFiRE-1 geometry at α = 0
38 of 41
American Institute of Aeronautics and Astronautics
39. Picture #10, CL+3 sec Picture #20, CL+8 sec
α=+5
Centerline 2 sec.
Picture #10, CL+3 sec Picture #20, CL+8 sec
α=+5
Centerline 8 sec.
Figure 55. Mach 6 Wind Tunnel TSP Results for P0 = 700 psi T0 = 900 R for HIFiRE-1 geometry at α = 5
Picture #10, CL+3 sec Picture #20, CL+8 sec
α=+10
Centerline 2 sec.
Picture #10, CL+3 sec Picture #20, CL+8 sec
α=+10
Centerline 8 sec.
Figure 56. Mach 6 Wind Tunnel TSP Results for P0 = 700 psi T0 = 900 R for HIFiRE-1 geometry at α = 10
39 of 41
American Institute of Aeronautics and Astronautics
40. VII.C. Comparison between Experimental and Computational Results
The comparison for the pressure coefficient values from the wind tunnel experiments and the computational
results are shown in Figure 57. When the notch is at 90 degrees starboard, the instrumentation is directly
in the flow direction, resulting in higher pressure readings. The trends for pressure tap 1 and 2 match
well between the experimental and computational values. At pressure tap 3, shock induced boundary layer
separation causes a drop in surface pressure at the leeward 5 degree case, which, as expected, is not captured
in the inviscid CART3D solution. The panel method, CBAero, with viscous approximations accurately
predicts the drop in pressure at 5 degrees but does not predict the pressure recovery at 10 degrees when
compared to the the experimental results. The results for P0 = 2000 psi and T0 = 1000 R (HT/HP) as well
as P0 = 700 psi and T0 = 900 R (LT/LP) are shown for comparisons since they represent the extreme ends
of the test matrix.
AOA [deg]
-10 -8 -6 -4 -2 0 2 4 6 8 10
Cp
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
Experimental and Computational comparison for HIFiRE 1 at Pw = 1 at Notch 90 deg
Pw 1 HT/HP
Pw 1 LT/LP
Pw 1 CART3D
Pw 1 CBAero
Pw 1
AOA [deg]
-10 -8 -6 -4 -2 0 2 4 6 8 10
Cp
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Experimental and Computational comparison for HIFiRE 1 at Pw = 2 at Notch 90 deg
Pw 2 HT/HP
Pw 2 LT/LP
Pw 2 CART3D
Pw 2 CBAero
Pw 2
AOA [deg]
-10 -8 -6 -4 -2 0 2 4 6 8 10
Cp
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
Experimental and Computational comparison for HIFiRE 1 at Pw = 3 at Notch 90 deg
Pw 3 HT/HP
Pw 3 LT/LP
Pw 3 CART3D
Pw 3 CBAero
Pw 3
Figure 57. Comparison between Experimental and Computational results for HIFiRE-1 geometry
VIII. Conclusion
Although one-to-one comparisons of experimental and computational surface temperature distributions
at Mach 6 were not obtained in this work, TSP data was obtained from the Mach 6 High Reynolds Number
Facility at Wright-Patterson Air Force Base for the HIFiRE-1 payload geometry at Mach 5.85. TSP data was
not ascertained for all geometries presented in this effort due to scheduling delays in the experimental testing.
However, the mere demonstration of TSP capability at Mach 6 is a significant step forward for hypersonic
vehicle design. Such a capability can lead to augmented design studies in the future such that insight into
the thermal response of a geometry can be realized much earlier in the design process for hypersonic vehicles.
Furthermore, though output temperature distributions could not be obtained from the UNLATCH boundary
layer solver, the solver itself accepted flow solutions from Cart3D using a unique mesh adaptation strategy
40 of 41
American Institute of Aeronautics and Astronautics
41. developed specifically for this effort. If success of future work can demonstrate valid wall temperature results
from UNLATCH, there are powerful implications that can dramatically influence hypersonic vehicle design.
The most notable implication is that the cost of reaching a wall temperature solution is effectively reduced
to the cost of a single Euler flow solution. Breaching the barrier of the wall temperature problem of complex
geometries in hypersonic flow using a boundary layer approximation to Euler solutions, through the use of
Cart3D and UNLATCH, provides strong motivation to continue such work and to identify the problems
slowing the progress in the application and implementation of UNLATCH.
References
1Hayes, J. and Altman, S., “AFRL/RQ Mach 3 and Mach 6 High Re Facility,” AIAA Dayton-Cincinnati Section Lunch
and Learn.
2“Tecplot 360 EX 2015 R1,” Tecplot, Inc., 2015.
31Lt Brent Ruttle, J. S. and Liston, G., “Generic Hyersonic Vehicles for Conceptual Design Analyses,” AFRL/RQHT,
2130 Eighth Street, WPAFB, OH 45433-7542 USA, 2012.
4Kimmel, R. L. and Adamczak, D., “HIFiRE-1 Background and Lessons Learned,” AIAA, Vol. Accepted. 2012-1088, 2012.
5Nemec, M. and Aftosmis, M. J., “Adjoint-Based Adaptive Mesh Refinement for Complex Geometries,” AIAA Paper,
2008-0725, 2008.
6Aftosmis, M. J., Berger, M. J., and Melton, J. E., “Robust and Efficient Cartesian Mesh Generation for Component-Based
Geometry,” AIAA Paper, 97-0196, 1997.
7David J. Kinney, P., “CBAERO 4.1+ User’s Manual Version 1.2,” NASA Ames Research Center: CMB-2008-9, September
2012.
8Anderson, J. D., Hypersonic and High-Temperature Gas Dynamics, AIAA, 2006.
9Desktop Aeronautics, I., “Quick Start Guide to Cart3D,” DA v1.2 based on NASA v1.3.6, 2 May 2012.
10Nemec, M. and Aftosmis, M. J., “Adjoint Error Estimation and Adaptive Refinement for Embedded-Boundary Cartesian
Meshes,” AIAA Paper, 2007-4187, 2007.
11H. H. Hamilton, K. J. W. and DeJarnette, F. R., “Improved Approximate Method for Computing Convective Heating
on Hypersonic Vehicles Using Unstructured Grids,” AIAA, Vol. Accepted. 2006-3394, 2006.
41 of 41
American Institute of Aeronautics and Astronautics