You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/340768506

ANSYS CFD Simulations of Supersonic and Hypersonic Flows

Conference Paper · February 2020

CITATIONS READS

0 1,494

4 authors, including:

Krishna Zore John Stokes


ANSYS Inc. Pune India ANSYS
14 PUBLICATIONS   9 CITATIONS    11 PUBLICATIONS   8 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

AIAA High Lift Prediction Workshop-3 View project

AIAA Aeroelasticity Prediction Workshop-2 View project

All content following this page was uploaded by Krishna Zore on 19 April 2020.

The user has requested enhancement of the downloaded file.


6th National Symposium on Shock Waves - IITM (NSSW- 2020)

ANSYS CFD Simulations of Supersonic and Hypersonic Flows


Krishna Zore1, Isik Ozcer2, Luke Munholand3, John Stokes4
1
ANSYS Inc, Hinjewadi, Pune, 411057, India
2,4
ANSYS Inc, Montreal, Canada
3
ANSYS Inc, Lebanon, USA
Krishna Zore: krishna.zore@ansys.com

Abstract This paper presents supersonic and hypersonic flow simulations carried out using
ANSYS Fluent, including the use of mesh adaptation techniques to achieve fast, robust, and
accurate solutions. Two supersonic test cases are taken from the NASA AIAA 3rd Sonic Boom
Prediction Workshop1 (SBPW): the ‘Axie’ configuration at a Mach number (M) of 1.6,
designed to match the near field pressure signature of the inviscid NASA Concept 25D with
flow-through nacelle at 3 body lengths from the centerline; and the ‘Biconvex’ case, also at
M=1.6, and designed to create shocks that would intersect the engine exhaust jet plume3,4. For
these cases, the pressure signatures at multiple body lengths from the front and aft body show
excellent agreement with the experimental measurements. A third, hypersonic test case is taken
from Ref. 5, a 2-D incident shock wave at M=5, designed to study the interaction between the
shock wave and the boundary layer. For this case, wall pressure, skin-friction coefficient and
mean-flow boundary layer profiles, before/aftershock impingement shows good agreement
with the experimental data.

Keywords; Supersonic, Hypersonic, PUMA, CFD, Turbulence

1 Introduction

Civil aviation is thriving for subsonic flight and has become an essential mode of transport
nationally and globally. However, supersonic commercial flights are prohibited, with
breakthroughs required to develop environmentally acceptable and economically viable
aircraft. Noise due to sonic boom propagation is a further additional barrier in developing
commercial supersonic jets2. The National Aeronautics and Space Administration (NASA)
conducts various High-Speed Research (HSR) programs, in partnership with academia and
industries, to develop new methods and tools to mitigate the sonic boom phenomenon. Related
to these efforts, the American Institute of Aeronautics and Astronautics (AIAA) has organized
series of Sonic Boom Prediction Workshops1 (SBPW), providing test cases from the HSR
programs to encourage the worldwide Computational Fluid Dynamics (CFD) community to
develop and apply best practices and establish state-of-the-art techniques of predicting near-
field sonic boom signatures and atmospheric pressure propagations.
The ability of ANSYS Fluent to predict near-field sonic boom pressure signatures at
supersonic flow conditions are assessed using two AIAA 3rd SBPW test cases: the ‘Axie’
configuration, at M=1.6, designed to match near field pressure signature of the inviscid NASA
concept 25D with flow-through nacelle at 3 body length (BL) from the centerline8-12; and the
‘Biconvex’ tail shock generator, also at M=1.6, designed to create shocks that intersecting the
engine exhaust jet plume. For the latter, comparisons of near-field pressure signature at
Hnose=15 inches BL are made with NASA Ames 9- by 7-Foot Supersonic Wind Tunnel (9x7)
test data3,4. For the third test case, the hypersonic flow prediction capabilities are assessed on a
2-D incident shock wave at M=5, designed to study shock boundary layer interactions5 (SBLI)
6th National Symposium on Shock Waves - IITM (NSSW- 2020)
that are vital in the successful design of control surfaces, high-speed inlets, and thermal
protection systems7 of spacecraft.
Details of the test cases, geometries and meshes are illustrated in section 2. The ANSYS
Fluent solver numerics are described in section 3. Section 4 provides a discussion of the results
for each of the test cases. Summary conclusions and an outlook to future work are found in
section 5.

2 Test Cases, Geometry and Mesh

The test cases used for validating ANSYS Fluent capabilities in predicting supersonic and
hypersonic flows are summarized in Table 1. Geometries for the Axie and Biconvex
configurations are taken from the SBPW31 in the generic ‘STEP’ CAD format and are shown
in Fig. 1a and 1b, respectively, and Fig. 1c shows a schematic of the 2-D incident shock from
Ref. 5. The water-tight computational domains for the Biconvex configuration is prepared
using ANSYS Discovery SpaceClaim Multipurpose 3D Design Modeler (SCDM 2019 R3),
whereas, the geometric sketch and water-tight meshes for Axie and 2-D incident shock are
prepared using ANSYS ICEM CFD 2019 R3.

Table 1. Test case summary

Configuration Flow Type Mach Number Mesh Topology Mesh Adaption


Axie Supersonic 1.6 Hybrid HexTet Yes
Biconvex Supersonic 1.6 AllPoly Yes
2-D Incident Shock Hypersonic 5 Hex Yes

(a) (b) (c)

Fig. 1 Geometries, (a) Axie, (b) Biconvex, and (c) 2-D incident shock

The computational meshes for all configurations are shown in Fig. 2. Fig 2a depicts the
hybrid inviscid unstructured ‘HexTet’ mesh of the Axie configuration. The fluid domain is
designed to capture the shocks efficiently and ensure flow is unequivocally either entering or
exiting the domain at all non-wall boundaries of the Axie configuration.
The ‘AllPoly’ unstructured viscous mesh for the Biconvex case, prepared using ANSYS
Fluent Meshing tool 2019 R3, is shown in Fig. 2b. The fluid core and viscous boundary layers
are filled with general polyhedra and poly-prisms, respectively. The poly-prisms are generated
by inflating surface polygons, with the first layer height specified to satisfy a wall Y+=1 and a
growth rate from the wall of 1.18.
The structured ‘Hex’ mesh prepared for the 2-D incident shock using ANSYS ICEM
CFD 2019 R3 is shown in Fig. 2c. The bottom flat plate is 500 mm long, with mesh in the
6th National Symposium on Shock Waves - IITM (NSSW- 2020)
turbulent boundary layer inflated from Y+=1 for the first layer using biasing with a growth rate
of 1.15. The shock generator, that is, the top plate, is at an angle of β=100, designed to produce
an incident oblique shock that impinges on the bottom plate boundary layer 350mm from the
leading edge.

(a) (b)

(c)

Fig. 2 Meshes, (a) Axie, (b) Biconvex, and (c) 2-D incident shock.

The solution-based mesh adaption techniques, resultant mesh sizes and their influence
on results are covered in the results and discussion section.

3 Flow Solver Numerics

The CFD simulations are performed using ANSYS Fluent 2019 R3, a cell-centered finite-
volume solver, using its fully implicit density-based solution method with the AUSM12 flux-
vector splitting scheme applied to convective fluxes. The simulations presented in this paper
use the Green-Gauss node-based and Least Squares cell-based spatial discretization methods
for computing secondary diffusion terms, velocity derivatives and scalars at cell faces.
Similarly, Third-Order MUSCL and second-order upwind spatial discretization schemes are
applied to the flow (pressure, momentum, and energy) and turbulence model equations
respectively. Furthermore, a Warped-Face Gradient Correction (WFGC) method is enabled to
account for gradient accuracy degradation due to very high aspect ratio, non-flat-faced cells in
the boundary layers and any highly deformed cells with cell centroid outside of the control
volume. Finally, options were set to enable special High-Speed Numerics (HSN) and
6th National Symposium on Shock Waves - IITM (NSSW- 2020)
Convergence Acceleration for Stretched Meshes (CASM) to stabilize and accelerate solution
convergence. The integrated expression language is used for defining mesh refinement and
coarsening criterion for dynamic mesh adaption.

4 Results and Discussions

In this section, computational results for each of the three cases are presented. Additionally,
the ability of solution-based mesh adaption to capture pressure signatures with reduced cell
counts is also highlighted.

4.1 Axie Configuration

The Axie configuration is the simplified version of the inviscid NASA Concept 25D 8-12
designed with Cart3D to match near field pressure signature with flow-through nacelle at 3
body lengths from the centerline. The motivation behind this test case is to reduce simulation
complexity for CFD code verification. The simulation operating conditions are illustrated in
Table 2.

Table 2. Axie operating conditions

Mach Number Angle of Attack (α) Ref. Body Length Altitude Temperature
1.6 00 32.92 m 15760 m 216.65 K

(a) (b)

Fig. 3 Axie Configuration, (a) Hybrid HexTet adapted mesh, and (b) Pressure signature
‘dp/Pinf’ contour

The mesh resulting from the applied solution-based mesh adaption is shown in Fig. 3. It
captures all the oblique shocks and expansion fans originating from the nose/tail and shoulders
of the Axie configuration. Fig 3a shows the symmetry plane of the finest adapted mesh, which
𝑑𝑝
has 198million elements, and Fig 3b shows the pressure signature contour (Eq. (1))
𝑝𝑖𝑛𝑓
overlaid with the measuring body length markers, BL-1 (Z=32.92m), BL-3 (Z=98.76m), and
BL-5 (Z=164.6m).
𝑑𝑝 (𝑝−𝑝𝑖𝑛𝑓 )
= (1)
𝑝𝑖𝑛𝑓 𝑝𝑖𝑛𝑓
6th National Symposium on Shock Waves - IITM (NSSW- 2020)

(a) (b)

(c)

Fig. 4 Axie Configuration, Sonic Boom Pressure Signatures, (a) BL-1, (b) BL-3, and (c) BL-5

There are no experimental measurements available for the Axie configuration. The
reference against which results are verified is the ensemble mean of multiple results from
multiple solvers on fine meshes from the 2nd AIAA SBPW. Fig 4. shows the ANSYS Fluent
2019 R3 predicted sonic boom pressure signatures compared with this ensemble mean at the
three measurement locations.
Fig 4a. illustrates the comparison at BL-1. Here, results from a 2-D axisymmetric
simulation (red curve) with the mesh adapted (refined) until the solution becomes mesh
independent, shows excellent agreement with the ensemble mean. Similarly, the 3-D conical
fluid domain with meshes adapted up to a fine (198 million (green curve)) and medium (101
million elements (orange curve)) resolution also show good match with the ensemble mean,
for both the positive ‘oblique shock’ and negative ‘expansion fan’ pressure signatures. The
coarse (49 million element mesh (blue curve)) is shown to be insufficient to resolve these
pressure signatures.
Comparisons in Fig 4b. between computed and ensemble mean pressure signatures at
BL-3 are very similar to those at BL-1.
The computed pressure signatures at BL-5 shown in Fig 4c., on the other hand, show a
slightly different resolution compare to the ensemble mean. These differences are minimal on
the oblique shock side but are significant on the expansion fan side. This is still seen with the
fine (198 million element mesh (green curve)), indicating the need for still further mesh
refinement. The axisymmetric adapted mesh (red curve), which is adapted until a mesh
6th National Symposium on Shock Waves - IITM (NSSW- 2020)
independent solution is achieved, shows that a good match with the ensemble mean can be
achieved with sufficient mesh refinement.

4.2 Biconvex Configuration

The Biconvex is one of the few generic horizontal tail shocks generators4 studied at NASA
Ames 9x7 Supersonic Wind Tunnel to understand shock interactions with nozzle jet plume and
their impacts on the sonic boom pressure signatures. Table 3 summarizes the computational
operating conditions, where it is noted that the Reynolds number is slightly higher than those
reported for the experiment4. However, this difference was verified by the workshop organizing
committee to have a negligible impact on the pressure signature predictions. While the
experiments were carried out at multiple Nozzle Pressure Ratios (NPRs), from 1 (no flow) to
14 (under-expanded), only an NPR 8 (design condition) is considered here.

Table 3. Biconvex operating conditions

Angle of Temperature Reynolds Engine plenum Engine plenum


Mach Attack (K) Number Pre. ratio Temp. ratio
Number (α) per meter (Ptot/Pinf) (Ttot/Tinf)
1.6 00 207.778 14.84x10^6 8.0 1.768

Fig. 5 Biconvex Configuration, Polyhedral Unstructured Mesh Adaptation (PUMA)

Fig. 5 shows ‘AllPoly’ the original unstructured mesh without refinement (9.2 million
elements), a mesh with a Body of Influence (BOI)-based refinement (44.6 million), and two
meshes using Polyhedral Unstructured Mesh Adaption (PUMA-4_14.9million & PUMA-
12_27.7million). The solution-based PUMA is performed using ANSYS Fluent 2019R3 using
a newly added ability to specify adaption controls employing a flexible and intuitive expression
language. A custom field function (CFF) for absolute ‘dp/Pinf’ (“Eq. (1)”) “iso-min=0.003 &
iso-max=0.3” is used to create a field variable cell register (field_value_ref) for refinement,
6th National Symposium on Shock Waves - IITM (NSSW- 2020)
and region-based cell register (region_noref) for coarsening criterion, later this criterions are
applied in a logical expression (‘AND(field_value_ref,NOT(region_noref))’) in a adaption
control. Additionally, the maximum refinement levels, minimum cell volume, and dynamic
adaption frequency (iterations between adaptions) are set to 5, 5e-09, and 250 respectively.

(a) (b)

Fig. 6 Biconvex configuration, pressure signature ‘dp/Pinf’ contour overlaid with PUMA, (a)
Overall view including measurement location, and (b) Shock plume interaction view

Fig. 7 Biconvex configuration, sonic boom pressure signature ‘dp/Pinf’ cut at BL (Hnose) =
15inches

The pressure signature ‘dp/Pinf’ (“Eq. (1)”) contour overlaid by the PUMA-12 adapted
mesh is shown in Fig. 6. Fig. 6a depicts the overall view of the pressure signature captured by
the dynamic mesh adaption. Approximately 12 refinement levels are required to capture the
pressure signature at the measuring location. Fig. 6b shows a close-up view of the pressure
signature highlighting the shock plume interaction overlaid by the adapted mesh. It is seen that
the oblique shock and expansion fans originated from the leading and trailing edge of the
Biconvex shock generator, respectively, intersect the nozzle jet plume.
Fig. 7 shows the sonic boom pressure signature measured at BL (Hnose)=15 inches from
the nose cone. The baseline, unrefined mesh (‘no-refinement_9.2million element’ (blue curve))
6th National Symposium on Shock Waves - IITM (NSSW- 2020)
fails to resolve a near-field pressure signature. However, the ‘BOI-refined_44.6million’ mesh
(red curve) predicts near-field pressure signature accurately and shows excellent agreement
with the experimental measurement. The oblique shocks and expansion fans originated from
the body parts such as nose, support blade strut, nozzle, and Biconvex shock generator
interacting nozzle jet plume are well resolved. However, the minimum element size required
was approximately 3 mm, which significantly increases the total element count and with that
the computational cost. To reduce the computational cost, solution-based PUMA was used to
optimize the mesh. The PUMA-4_14.9 million element mesh (purple curve), with up to four
levels of refinement compared to the unrefined baseline mesh, predicts forebody shocks
accurately. However, the blade strut oblique shock and the Biconvex shocks passing through
nozzle jet plume are not captured well. With 12 levels of refinement, as seen in the ‘PUMA-
12_27.7million’ (green curve) results, all the shocks are predicted accurately and show
excellent agreement with the experimental measurement. This indicates that the use of a
PUMA-optimized mesh gives similar accuracy as the BOI-refined mesh but with ~38% fewer
elements and ~30% less computational cost (time).

4.3 2D Incident shock wave

Incident shock wave interactions with turbulent boundary layers are of fundamental importance
in modern aerothermodynamics and high-speed fluid dynamics7, leading to an interest in
evaluating the ability to predict these interactions computationally. The operating conditions
for the given test case are given in Table 4.

Table 4. Incident shock wave operating conditions

Mach Total Pressure Total Temperature Reynolds number Wall Temperature


Number (bar) (K) per meter (K)
5 21.2 410 40x10^6 300

Fig 8. depicts the experimental shadowgraphs and qualitative flow field diagram for
β=100 shock deflection angle. The numbered arrows at the bottom of the flow field diagram
show the positions of measuring pitot and static pressure probes for determining boundary-
layer properties. The flow separation S and reattachment R positions are marked with black
arrows.

Fig. 8 Incident shock wave, shadowgraph and qualitative flow field diagram5 for β=100
6th National Symposium on Shock Waves - IITM (NSSW- 2020)

Fig. 9 Incident shock wave, turbulent boundary-layer represented by turbulent viscosity ratio
contour overlaid by solution-driven mesh adaption for β=100

The predictions are made using the ANSYS Fluent 2019 R3 density-based solution
method with the Spalart Allmaras (SA) turbulence model. Fig. 9 shows the turbulent viscosity
ratio contour overlaid by the solution-based mesh adaptation, where refinement criterion is set
based on the field variable cell register, marking cells with static pressure gradient more than
60, maximum refinement levels 5, minimum cell volume 1e-12, and dynamic mesh adaption
frequency 100. The computational boundary-layer thickness on the flat plate represented by
the turbulent viscosity ratio contour looks very similar to the experimental shadowgraph
boundary-layer thickness (see Fig. 8). The inviscid shock impingement position of 350mm
from the leading edge of the bottom plate that is observed in the experiment is also similar to
what is predicted by the CFD simulation. Fig. 9 also shows the sections at which boundary
layer measurements were made. The top and bottom rows of Fig. 10 show x-velocity and static
temperature profiles, respectively, compared with the experimental data in sections 3, 7 and
10.
In section 3, before the shock boundary-layer intersection, all three meshes slightly
underpredict the velocity and temperature boundary layer profiles compared to experimental
data but show a good match on the edge of the boundary layer.
In section 7, just after the intersection, the velocity profile predictions show good
agreement with the experimental data, with minor underprediction near the wall. This may be
due to discrepancies in the recovery of the flow from the separation. The temperature profile
shows more significant underpredictions, the reason for which may be as discussed in Ref. 7,
that the complex nonequilibrium behavior of these turbulent interacting flow not being
captured in relatively simple eddy viscosity turbulence models.
In section 10, well behind the intersection location, the velocity profiles again show
better agreement with the experiment data, however, the profile curve slopes are slightly
different. A similar observation can be made for the temperature profile, which shows
underprediction in the near-wall region but good agreement towards the outer edge of the
boundary-layer.
6th National Symposium on Shock Waves - IITM (NSSW- 2020)

Fig. 10 Incident shock wave, x-velocity (top) and static temperature (bottom) profiles in the
boundary-layer at section 3, 7 & 10

Fig. 11 The incident shock wave, skin friction coefficient (left) and wall static pressure
(right) on the flat plate.

Fig. 11 shows the comparison of skin friction coefficient (left) and wall static pressure
(right) on the bottom plate between computation and experiment. The skin friction coefficient
predicted by all the meshes indicates an accurate prediction of flow separation and reattachment
locations, S and R, respectively, as marked by black arrows. However, in the separated flow
region and downstream of the shock boundary-layer interaction it shows slight underprediction.
The wall static pressure shows good agreement with the experimental data, including the
sudden rise in wall pressure after the shock boundary-layer interaction. A slight
underprediction of wall static pressure remains downstream of the intersection location.
These comparisons do not show any significant mesh dependence, indicating that the
discrepancies between computed and experimental results are not due to insufficient spatial
resolution but rather other factors, such as the turbulence model.
6th National Symposium on Shock Waves - IITM (NSSW- 2020)
5 Conclusion

This paper outlines the results of three public test cases relevant to supersonic and hypersonic
flows. In general, good agreement with available experiments is observed, with some areas for
future work and improvement identified. For the supersonic cases looking at the ability to
capture near field pressure signals accurately, the refinement of the mesh using Fluent’s PUMA
technology proved invaluable. Further efficiency gains might be possible using anisotropic
mesh adaptation, and this will be investigated in the future. For the hypersonic case focused on
shock boundary-layer interaction, mesh adaptation was also applied successfully. Further
improvements may be possible here with higher fidelity turbulence models, e.g. using scale-
resolving hybrid RANS-LES methods. Such further investigation will also be the subject of
future work.

6 Acknowledgment

The authors would like to thank the AIAA Sonic Boom Prediction Workshop organizing
committee for providing data on supersonic test cases. Many thanks to Patrick Sharkey, Alok
Khaware and Deepak Yadav for valuable inputs during this study.

References

[1] https://lbpw.larc.nasa.gov/sbpw3/ “3rd AIAA Sonic Boom Prediction Workshop”.

[2]https://www.nap.edu/catalog/10283/commercial-supersonic-technology-the-way-ahead
“Commercial Supersonic Technology: The Way Ahead (2001)”.

[3] Susan E. Cliff, Marie F. Denison B., Shayan Moini-Yekta, Donald E. Morr, Donald A.
Durston, “Wind Tunnel Model Design for Sonic Boom Studies of Nozzle Jet Flows with Shock
Interactions”, AIAA 2016-2035.

[4] James C. Jensen, Marie Denison, Don Durston, and Susan E. Cliff., “Computational and
Experimental Study of Plume and Shock Interaction Effects on Sonic Boom in the NASA
Ames 9x7 Supersonic Wind Tunnel”, AIAA 2018-331.

[5] Schulein, E., Krogmann, P., and Stanewsky, E., “Documentation of Two-Dimensional
Impinging Shock/Turbulent Boundary Layer Interaction Flow,” DLR, German Aerospace
Center, Rept. IB 223-96 A 49, G¨ottingen, Germany, Oct. 1996.

[6] Schülein E. (2004) "Optical Skin Friction Measurements in Short-duration Facilities


(Invited)", 24th AIAA Aerodynamic Measurement Technology and Ground Testing
Conference, 28 June - 1 July 2004, Portland, Oregon, AIAA-2004-2115, 18p.

[7] Schülein E. (2006) “Skin Friction and Heat Flux Measurements in Shock Boundary Layer
Interaction Flows”, AIAA-Journal, Vol.44, No.8, Aug. 2006, pp.1732-1741.

[8] Irian Ordaz, Mathias Wintzer, Sriram K. Rallabhandiz, “Full-Carpet Design of a Low-
Boom Demonstrator Concept”, 33rd AIAA Applied Aerodynamics Conference, 22-26 June
2015, Dallas, TX, AIAA 2015-2261.
6th National Symposium on Shock Waves - IITM (NSSW- 2020)
[9] Mathias Wintzer, Irian Ordaz, James W. Fenbert, “Under-Track CFD-Based Shape
Optimization for a Low-Boom Demonstrator Concept”, 33rd AIAA Applied Aerodynamics
Conference, 22-26 June 2015, Dallas, TX, AIAA 2015-2260.

[10] Farhat, C., Maute, K., Argrow, B., and Nikbay, M., “Shape Optimization Methodology
for Reducing the Sonic Boom Initial Pressure Rise,” AIAA Journal, Vol. 45, No. 5, May 2007,
pp. 1007–1018.

[11] George, A. R. and Seebass, R., “Sonic Boom Minimization Including Both Front and Rear
Shocks,” AIAA Journal, Vol. 9, No. 10, October 1971, pp. 2091–2093.

[12] Ordaz, I., Geiselhart, K. A., and Fenbert, J. W., “Conceptual Design of Low-Boom
Aircraft with Flight Trim Requirement", AIAA Paper 2014-2141, June 2014.

[13] M. S. Liou. "A sequel to AUSM: AUSM+". Journal of Computational Physics. 129. 364–
382. 1996.

View publication stats

You might also like