Knight 2012
Knight 2012
Knight 2012
a r t i c l e i n f o a b s t r a c t
Available online 10 January 2012 The aerothermodynamic loadings associated with shock wave boundary layer interactions (shock
Keywords: interactions) must be carefully considered in the design of hypersonic air vehicles. The capability of
Hypersonics Computational Fluid Dynamics (CFD) software to accurately predict hypersonic shock wave laminar
Computational fluid dynamics boundary layer interactions is examined. A series of independent computations performed by
Shock waves researchers in the US and Europe are presented for two generic configurations (double cone and
cylinder) and compared with experimental data. The results illustrate the current capabilities and
limitations of modern CFD methods for these flows.
& 2011 Elsevier Ltd. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2. Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1. Double cone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2. Cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3. Details of computations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1. Governing equations and boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.1. Conservation of mass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.2. Conservation of momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.3. Conservation of energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.4. Conservation of vibrational energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.5. Equation of state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.6. Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2. Thermochemistry model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3. Thermodynamic and transport properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4. Numerical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.1. Drikakis et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.2. Gaitonde . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4.3. Lani . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.4.4. Nompelis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.4.5. Reimann . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
n
Corresponding author. Tel.: þ1 732 445 4464; fax: þ 1 732 445 3124.
E-mail address: [email protected] (D. Knight).
1
Currently at European Space Research & Technology Centre, Keplerlaan 1,
Postbus 299, 2200 AG Noordwijk, The Netherlands.
2
Currently at Department of Mechanical and Aerospace Engineering, Ohio
State University, Columbus, OH 43210, USA.
0376-0421/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.paerosci.2011.10.001
D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26 9
3.4.6. Walpot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4. Results for double cone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.1. Run 42 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.1.1. Assessment of accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.1.2. Flowfield structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2. Run 40 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2.1. Assessment of accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3. Run 80 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5. Results for cylinder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.1. HEG I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.1.1. Assessment of accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.1.2. Flowfield structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.2. HEG III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5.2.1. Assessment of accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1. Introduction
and base flows with and without plume interaction [2]. Subgroup
3 of WG10 examined CFD capability for shock wave–turbulent
Renewed interest in hypersonic air vehicles such as the Boeing
boundary layer interactions focusing on five generic configura-
X-51 (Fig. 1) has focused research on topics critical to hypersonic
tions: 2-D compression corner (Fig. 3a), 2-D expansion–compres-
flight. The design of hypersonic air vehicles involves numerous
sion corner (Fig. 3b), 2-D shock impingement (Fig. 3c), 3-D single
engineering disciplines including aerothermodynamic analysis. In
fin (Fig. 2a) and 3-D double fin (Fig. 2b). All cases were perfect gas
particular, the interaction of shock waves with the vehicle
flows. All 2-D shock interaction flow simulations were performed
boundary layers can lead to regions of enhanced aerothermody-
using either Direct Numerical Simulation (DNS) or Large Eddy
namic loading, and therefore accurate modeling of shock wave
Simulation (LES) models, and all 3-D shock interaction flow
boundary layer interactions (‘‘shock interactions’’) is essential.
simulations were performed using RANS methods. The report
During the past decade two NATO Research Technology Organiza-
concluded that DNS and LES results for 2-D shock wave turbulent
tion (RTO) Working Groups (WGs) have assessed the capabilities for
boundary layer interactions showed significant progress in pre-
prediction of aerothermodynamic loads in high speed flight. AGARD
dicting the flow; however, the Reynolds numbers of the simula-
Working Group 18 (WG18) examined the Computational Fluid
tions were relatively low due to computational resource
Dynamics (CFD) capability for prediction of 2-D and 3-D shock wave
requirements, and no comparison with experimental surface heat
laminar and turbulent boundary layer interactions for three generic
transfer measurements was performed. The report concluded that
configurations: single fin (Fig. 2a), double fin (Fig. 2b) and hollow
new RANS concepts for 3-D shock wave turbulent boundary layer
cylinder flare (Fig. 2c) [1]. The single and double fin configurations
interactions showed improvement in prediction of the flow;
involved shock wave turbulent boundary layer interactions, while the
however, heat transfer was not accurately predicted.
hollow cylinder configuration included both laminar and turbulent
This paper presents the results of the participants in RTO AVT
shock boundary layer interactions. All cases were perfect gas flows.
136 Subtopic No. 2. The focus is an assessment of CFD for the
All turbulent flow simulations were performed using Reynolds-
specific issue of shock interactions and control surfaces in non-
Averaged Navier–Stokes (RANS) models. The report concluded that
equilibrium laminar flows. The paper presents a comparison of
while turbulent perfect gas shock interaction predictions were
computed and experimental results for two new configurations.
accurate for 3-D mean pressure and primary separation locations,
The first configuration is a double cone (Fig. 4a) from Calspan
nevertheless the skin friction and heat transfer were poorly predicted.
University of Buffalo Research Center (CUBRC). The second con-
RTO Working Group 10 (WG10) conducted a detailed exam-
figuration is a cylinder (Fig. 4b) from the German Aerospace
ination of CFD capability for six areas relevant to hypersonic
Center (DLR). Experimental data are available for two different
flight: boundary layer instability and transition, real gas flows,
enthalpy conditions under laminar flow conditions.
laminar hypersonic viscous–inviscid interactions, shock–shock
interactions, shock wave–turbulent boundary layer interactions
2. Experiments
than the predicted levels. The test conditions are shown in Table 1 3. Details of computations
where subscripts o and 1 represent reservoir and test section
conditions, respectively. The values represent best estimates The participants in the study, listed in Table 3, were solicited on
based on measurements made in the reservoir and freestream the basis of their expertise in CFD simulation of shock interactions in
in flow calibration studies. hypersonic flows. All details of the computations were determined
individually and independently by each participant including selec-
tion of the thermochemistry and transport models, inclusion or
omission of a separate vibrational energy model, choice of numerical
2.2. Cylinder algorithms, determination of the numerical grid and grid refinement
strategy, etc. In this section we present the governing equations and
A test campaign [4] to investigate the flow past a cylindrical boundary conditions, and the thermochemistry, thermodynamic and
model was performed at the high enthalpy shock tunnel (HEG) of transport models used by each participant. In addition, a brief
the German Aerospace Center (DLR). The cylindrical model, with a description of the numerical algorithms is provided.
radius of 45 mm and a span of 380 mm, was mounted on the
nozzle centerline with its axis transverse to the flow. It was
3.1. Governing equations and boundary conditions
equipped with 17 pressure transducers and the same number of
thermocouples to measure surface pressure and heat flux dis-
For convenience of comparison between the participants, the
tributions. The transducers were distributed along six rows
governing equations for unsteady, laminar, compressible flow of a
located close to the plane of symmetry at the mid-span location
mixture of reacting gases are written in Cartesian tensor notation
(10, 20 and 30 mm to the left and right of the plane of symmetry,
(using the modified Einstein summation convention3 where
covering a circumferential angle of 7601 with respect to the
convenient for readability). A different coordinate system may
inflow direction).
have been used for the governing equations by the participants.4
The large shock standoff distance of this configuration permitted
All computations assumed laminar flow.
the application of optical measurement techniques for the determi-
nation of the gas properties in the shock layer. Holographic inter-
ferometry and time resolved Schlieren were applied to measure 3.1.1. Conservation of mass
density distributions in the shock layer and the temporal evolution The conservation of mass for each specie is
of the bow shock shape. Freestream static and pitot pressures and
@ra @ra uj @
stagnation heat transfer on a sphere were recorded at each run for þ ¼o
_ aþ ðra uaj Þ ð1Þ
@t @xj @xj
calibration, normalization and statistical purposes.
The measurements on the cylinder were carried out at different for a ¼ 1, . . . ,n where ra is the density of species a, uj is jth
total enthalpies (HEG conditions I and III, 22.4 MJ/kg and 13.5 MJ/kg, Cartesian component of the mass-averaged velocity of the mix-
respectively) and with air as a test gas. The HEG reservoir and ture, and o _ a is the net rate of production of species a per unit
freestream data for the measurements are listed in Table 2.
At the experimental conditions the flow in the shock layer is
3
subject to non-equilibrium chemical relaxation processes that sig- The modified Einstein summation convention implies summation over all
values of a Roman index (e.g., j) when it appears twice in a given term. Summation
nificantly affect the density distribution and hence the shock standoff over a Greek index (e.g., a) is identified by a summation sign S.
distance. Therefore, this test case represents a useful basis for the 4
For example, the governing equations for the double cone configuration may
validation of the physico-chemical models used in CFD codes. be written in cylindrical polar coordinates.
D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26 11
Fig. 3. Configurations for WG10. (a) 2-D compression. (b) 2-D compression–
expansion. (c) Shock impingement.
Table 1
Test conditions for double cone.
Table 2
Test conditions for cylinder.
Quantity I III
Table 3
Participants.
Name Organization
where the translational and rotational states of the gas are The heat flux is given by (14) and nd is the number of diatomic
assumed in equilibrium at temperature T and the vibrational species. The source term is
state is characterized by the vibrational temperature Tv. nd nd
X X
o_ v ¼ qatv þ o_ a eva ð17Þ
a¼1 a¼1
3.1.4. Conservation of vibrational energy
The conservation of vibrational energy is where qatv is the translational–vibrational energy transfer defined
! by the Landau–Teller model [5]
@ev @ev uj @qvj @ Xnd
v a
þ ¼ r e u þ o_ v ð15Þ
@t @xj @xj @xj a ¼ 1 a a j eav ðTÞeav ðT v Þ
qatv ¼ ra ð18Þ
ta
where ev is the vibrational energy per unit volume
nd
X and ta is the relaxation time [6,7].
ev ¼ ra eva ð16Þ An equation for vibrational energy was used in most, but not
a¼1 all, of the computations. Details are presented in Table 4.
D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26 13
Table 4 Table 6
Included vibrational energy equation. Mixture species and reactions.
Xn
ra Participant Reference
p ¼ RT ð19Þ
M
a¼1 a Drikakis Gupta et al. [28]
Gaitonde Grossman and Cinnella [34], Chase et al. [35]
where R is the universal gas constant and M a is the molecular Lani Panesi [36]
weight of species a. Nompelis Blottner et al. [37]
Reimann Bottin [38]
Walpot Blottner et al. [37]
Participant Reference
3.2. Thermochemistry model Drikakis Gnoffo et al. [27], Gupta et al. [28]
Gaitonde Blottner et al. [37], White [39]
The same thermochemistry model was used by all participants Lani Barbante [40]
for each configuration. The double cone was modeled as a two- Nompelis Blottner et al. [37], Nompelis [41]
Reimann Gupta et al. [28]
component mixture (N2 and N). The cylinder was modeled as a
Walpot Blottner et al. [37]
five-component mixture (N2, O2, NO, N and O). The reactions used
by the participants are listed in Table 6. References for the
reaction rates are given in Table 7. Table 10
Mixture thermal conductivity and viscosity models.
7.0E+04 7.0E+04
4.0E+04 4.0E+04
P (Nt/m )
P (Nt/m )
3.0E+04 3.0E+04
2.0E+04 2.0E+04
1.0E+04 1.0E+04
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
Fig. 5. pw for Run 42. (a) Gaitonde. (b) Lani. (c) Drikakis et al. (d) Nompelis. (e) Reimann. (f) Walpot.
D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26 15
systems. CNS3D has been previously used in various hypersonic third-order upwind-biased MUSCL approach modulated by the
flows, with and without ablation effects [11–13]. Additional minmod limiter. Additional details are provided in Gaitonde [17].
details are provided in Tissera et al. [12].
3.4.3. Lani
The double cone Run 40 configuration was computed using a
3.4.2. Gaitonde cell-centered finite volume algorithm based on AUSM þ with
All results were obtained with the GASP code [14]. For the Least Squares reconstruction and the Venkatakrishnan limiter.
double-cone case, the Roe scheme was used (see, e.g., [15] for The double cone Run 42 was computed using a novel Contour
formulation with chemistry) with the van Albada limiter. For the Residual Distribution method with the BXC scheme. The cylinder
cylinder cases, the AUSMþ scheme [16] was employed with the configuration was simulated using a cell-centered finite volume
7.0E+06
7.0E+06 Run 42 - park I
Experiment 6.0E+06
6.0E+06 514x256 grid
1024x512 grid
5.0E+06
5.0E+06
Q (W/m2)
4.0E+06
Q (W/m2)
4.0E+06
3.0E+06
3.0E+06
2.0E+06
2.0E+06
1.0E+06 1.0E+06
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
7.0E+06
7.0E+06 Max value 8.1 10
7
Experiment 6.0E+06
6.0E+06 Computed (48x128, VL)
Computed (48x128, WENO5th)
Computed (96x256, VL) Experiment
Computed (96x256, WENO5th) 5.0E+06 Computed
5.0E+06
Q (W/m2)
4.0E+06
Q (W/m2)
4.0E+06
3.0E+06 3.0E+06
2.0E+06 2.0E+06
1.0E+06 1.0E+06
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
7.0E+06 7.0E+06
Q (W/m2)
4.0E+06 4.0E+06
3.0E+06 3.0E+06
2.0E+06 2.0E+06
1.0E+06 1.0E+06
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
Fig. 6. qw for Run 42. (a) Gaitonde. (b) Lani. (c) Drikakis et al. (d) Nompelis. (e) Reimann. (f) Walpot.
16 D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26
algorithm based on AUSM þ with Least Squares reconstruction LU-SGS scheme have been used. Additional details are provided in
and the Venkatakrishnan limiter. Reimann and Hannemann [24].
Fig. 7. Flowfield structure for Run 42. (a) Overall flowfield. (b) Enlargement of flowfield. (c) Enlargement of flowfield.
D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26 17
and 4% for heat transfer relative to the individual measured indicate sensitivity of the upstream propagation (Fig. 5a) to the
values [25]. grid refinement; in particular, the predicted upstream propaga-
tion location for the 1024 512 grid are in excellent agreement
4.1. Run 42 with experiment, while the predicted upstream propagation for
the 512 256 grid underestimates the upstream propagation by
4.1.1. Assessment of accuracy approximately 4 mm.
The computed and experimental surface pressure are pre- The assessment of the accuracy of the predictions of ‘‘peak’’
sented in Fig. 5. All six simulations accurately predict the initial pressure is problematic for two reasons. First, it is not at all
plateau pressure (0:05 m ox o 0:09 m), although some of the evident that the experimental transducer at x ¼0.1133 m (the
simulations underestimate the upstream propagation of the first location of the highest measured p) corresponds to the peak
shock boundary layer interaction at x 0:05 m. Gaitonde’s results pressure in the experiment. Consequently, a computed peak
Experiment
6.0E+04 0.00300
6.0E+04
0.00325 Experiment
0.00350 Computed
5.0E+04 0.00370 5.0E+04
0.00400
P (Nt/m2)
P (Nt/m2)
4.0E+04 4.0E+04
3.0E+04 3.0E+04
2.0E+04 2.0E+04
1.0E+04 1.0E+04
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
7.0E+04 7.0E+04 4
4 Max value 8.5 10
Max value 7.2 10
1.0E+04 1.0E+04
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
7.0E+04 7.0E+04
P (Nt/m2)
4.0E+04 4.0E+04
3.0E+04 3.0E+04
2.0E+04 2.0E+04
1.0E+04 1.0E+04
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
Fig. 8. pw for Run 40 (y Time counter is notional and does not represent absolute time). (a) Gaitonde y . (b) Lani. (c) Drikakis et al. (d) Nompelis. (e) Reimann. (f) Walpot.
18 D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26
4.0E+06
Q (W/m2)
4.E+06
3.0E+06 3.E+06
2.0E+06 2.E+06
1.0E+06 1.E+06
0.0E+00 0.E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
7.0E+06
7.0E+06
6.0E+06
Experiment Experiment
6.0E+06 Computed (48x128, VL) 0.0031
Computed (48x128, WENO5th) 0.0198
Computed (96x256, VL) 5.0E+06 0.0364
5.0E+06 Computed (96x256, WENO5th) 0.0530
0.0697
Q (W/m2)
4.0E+06 0.0863
Q (W/m2)
4.0E+06
0.1029
0.1195
3.0E+06 0.1362
3.0E+06
0.1528
0.1695
2.0E+06 2.0E+06 0.1861
1.0E+06 1.0E+06
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
7.0E+06 7.0E+06
Q (W/m2)
4.0E+06 4.0E+06
3.0E+06 3.0E+06
2.0E+06 2.0E+06
1.0E+06 1.0E+06
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X (m) X (m)
Fig. 9. qw for Run 40 (y Time counter is notional and does not represent absolute time). (a) Gaitonde y . (b) Lani. (c) Drikakis et al. (d) Nompelis. (e) Reimann. (f) Walpot.
pressure greater than the experimentally measured peak pressure distribution would be needed. Consequently, a strict comparison
is not necessarily in error.5 Second, the transducer size is finite between the computed and experimental peak pressure is inap-
and therefore integrates the pressure over a small surface area. propriate in regions of extremely rapid changes. Nevertheless, a
Depending on the grid resolution, the lateral scale of the trans- visual comparison of the computed and experimental pressure
ducer may be equivalent to several times the grid spacing along profile in the vicinity of the peak pressure does indicate reason-
the surface, and therefore filtering of the computed pressure ably close agreement. Finally, the secondary plateau (x Z0:13 m)
is accurately predicted by all participants. All six simulations
resulted in steady flows in agreement with experiment.
5
The converse, of course, is not true. A computed peak pressure which is less
The computed and experimental surface heat transfer are
than the experimentally measured peak pressure (and outside the experimental presented in Fig. 6. The computed separation point at
error bars) is in error. The same holds true for heat transfer. x¼0.055 m is in good agreement with experiment for nearly all
D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26 19
simulations. It is interesting to note that while the refined grid peak heat transfer (Fig. 6) equals or exceeds the measured peak
(1024 512) results of Gaitonde show closer agreement with the heat transfer for all computations except Nompelis (Fig. 6d) and
surface pressure in the vicinity of separation, the coarser grid Riemann (Fig. 6e). The predicted heat transfer in the second
(512 256) results agree better with the experimental heat plateau region shows reasonable agreement with experiment
transfer in this region. This raises a question regarding the for all simulations. The convergence of all simulations to a steady
accuracy of the Park model for this type of flow. The computed state is in agreement with the experiment.
Fig. 10. qw for Run 40 vs t (experiment). (a) qw vs t for Run 40 (experiment). (b) qw vs t for Run 40 (experiment).
3.0E+04 3.0E+06
2.5E+04 2.5E+06
Experiment Experiment
Computed Computed
2.0E+04 2.0E+06
Q(W/m2)
P(N/m )
2
1.5E+04 1.5E+06
1.0E+04 1.0E+06
5.0E+03 5.0E+05
0.0E+00 0.0E+00
0 0.025 0.05 0.075 0.1 0.125 0.15 0 0.025 0.05 0.075 0.1 0.125 0.15
X(m) X(m)
Fig. 11. pw and qw for Run 80 (experiment and computation). (a) qw vs t for Run 80 (experiment). (b) qw for Run 80 (experiment). (c) pw for Run 80 (Nompelis). (d) qw for
Run 80 (Nompelis).
20 D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26
4.1.2. Flowfield structure structure illustrated in Fig. 7b and a rapid pressure rise on the
The structure of the double cone flowfield is illustrated in surface resulting in the formation of a separated region delimited
Fig. 7 for Run 42 based upon the computations by Gaitonde. by the separation and reattachment points at x 0:05 m and
Fig. 7a displays the entire flowfield. The first cone shock gener- x 0:11 m, respectively (Fig. 7a). The separation point is slightly
ated by the 251 cone is attached to the cone apex and is located downstream of the upstream influence point where the surface
very close to the first cone surface. A second cone shock is pressure begins to rise (Fig. 5a). Within the separation region the
generated by the 551 cone. The interaction of these two shock pressure is approximately constant (Fig. 5a) and the heat transfer
waves (shock–shock interaction) creates a complex wave is substantially reduced (Fig. 6a). The separation of the boundary
80000
70000
60000
50000
p (N/m )
40000
30000
20000 Experiment
Computed
10000
0
-80 -60 -40 -20 0 20 40 60 80
θ (deg)
80000
70000 Experiment
Computed (Park)
Computed (Dunn and Kang)
60000
50000
p (N/m )
40000
30000
20000
10000
0
-80 -60 -40 -20 0 20 40 60 80
θ (deg)
Fig. 12. pw for HEG I. (a) Gaitonde. (b) Lani. (c) Nompelis. (d) Reimann. (e) Walpot.
D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26 21
layer causes a significant deflection of the flow resulting in a shock interaction (Fig. 7b) with another contact surface and
separation shock (Fig. 7a). The separation shock and second cone reflected shocks. The downstream of the second cone shock is
shock interact (shock–shock interaction) (Fig. 7b) and form a subsonic while the flow beneath the contact surface emanating
triple point with a consequent reflected shock and a contact from the crossing shock interaction is supersonic. One of the
surface [26] (contact discontinuity). The reattachment of the reflected shocks from the crossing shock interaction further
boundary layer generates a reattachment shock (Fig. 7b) with a reflects from the surface and subsequently reflects from the sonic
consequent rapid increase in surface pressure and heat transfer at line as an expansion (Fig. 7c) which interacts with the cone
x 0:12 m (Figs. 5a and 6a). The interaction of the reflected shock surface thereby reducing the surface pressure and heat transfer at
from the triple point and reattachment shock results in a crossing x 0:12 m (Fig. 5a).
1E+07
9E+06
8E+06
7E+06
6E+06
Q (W/m )
5E+06
4E+06
3E+06 Experiment
Computed
2E+06
1E+06
0
-80 -60 -40 -20 0 20 40 60 80
θ (deg)
1E+07
9E+06
8E+06
7E+06
Q (W/m )
6E+06
5E+06
4E+06
3E+06
2E+06 Experiment
Computed (Park)
Computed (Dunn and Kang)
1E+06
0
-80 -60 -40 -20 0 20 40 60 80
θ (deg)
Fig. 13. qw for HEG I. (a) Gaitonde. (b) Lani. (c) Nompelis. (d) Reimann. (e) Walpot.
22 D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26
4.2. Run 40 heat transfer at several different points in time, while Fig. 9b
displays only one selected instant in time. The former simulations
4.2.1. Assessment of accuracy show significant unsteadiness in the flowfield as expected from
The computed and experimental surface pressure are pre- the surface pressure predictions. A similar conclusion was
sented in Fig. 8. Fig. 8(a), (c)–(f) show instantaneous surface obtained by Lani.
pressure at several different points in time, while Fig. 8b displays Nevertheless, the experimental heat transfer was observed to
only one selected instant in time. It is evident from the former five be steady as indicated in Fig. 10. The behavior of the experimental
simulations that the computed flowfield is unsteady and does not heat transfer at a location upstream of the separation region is
achieve a stationary (steady) configuration. A similar conclusion displayed in Fig. 10a and indicates that a steady heat transfer has
was obtained by Lani. been achieved. Similarly, the experimental heat transfer distribu-
The computed and experimental surface heat transfer are tion on the surface is displayed in Fig. 10b and indicates that a
presented in Fig. 9. Fig. 9(a), (c)–(f) show instantaneous surface steady distribution is achieved for the entire cone surface. By
8.0 5500.0
7.0 5000.0
6.0 4500.0
5.0 4000.0
4.0 3500.0
3.0 3000.0
2.0 2500.0
1.0 2000.0
0.0 1500.0
1000.0
500.0
13000
12000
12000
11000
11000
10000
10000
9000
9000
8000
8000
7000
7000
6000
6000
5000
5000
4000
4000
3000
3000
2000
2000
1000
1000
0
2000 0.30
1000 0.25
0 0.20
-1000 0.15
-2000 0.10
-3000 0.05
-4000 0.00
-5000
-6000
-7000
-8000
Fig. 14. Flow structure for HEG I (Walpot, Dunn and Kang Model). (a) Mach number. (b) Velocity (m/s). (c) T (K). (d) Tv (K). (e) T v T (K). (f) N.
D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26 23
analogy, a steady experimental surface pressure distribution is certainly not attributable to a particular grid since different grid
also achieved. convergence studies were performed by the participants.
This is a paradox. Six separate and independent simulations
have observed a significant unsteadiness in the flowfield, while 4.3. Run 80
the experiment shows steady flow. The discrepancy is certainly
not attributable to a particular flux algorithm, since a variety of A separate experiment (Run 80) was performed to assess the
different algorithms were used by the participants. Also, it is effect of the Reynolds number (Table 1). The Reynolds number is
80000
70000
60000
50000
P (N/m2)
40000
30000
Experiment
Computed
20000
10000
0
-80 -60 -40 -20 0 20 40 60 80
θ (deg)
80000
70000
60000
50000
P (N/m2)
40000
30000
20000
P (N/m )
Computed (Park)
10000 Computed (Dunn and Kang)
0
-80 -60 -40 -20 0 20 40 60 80
θ (deg)
Fig. 15. pw for HEG III. (a) Gaitonde. (b) Lani. (c) Nompelis. (d) Reimann. (e) Walpot.
24 D. Knight et al. / Progress in Aerospace Sciences 48–49 (2012) 8–26
1E+07
9E+06
8E+06
7E+06
6E+06
Q (W/m2)
5E+06
4E+06
3E+06
2E+06
1E+06
0
-80 -60 -40 -20 0 20 40 60 80
θ (deg)
1E+07
9E+06
8E+06
7E+06
6E+06
Q (W/m2)
5E+06
4E+06
3E+06
2E+06
1E+06
0
-80 -60 -40 -20 0 20 40 60 80
θ (deg)
Fig. 16. qw for HEG III. (a) Gaitonde. (b) Lani. (c) Nompelis. (d) Reimann. (e) Walpot.
3:6 105 m1 which is approximately one-half of the value for kg) and HEG III6 is a low enthalpy (ho ¼13.5 MJ/kg). Note that
Run 40. The development of the surface heat transfer distribution both cases have significantly higher enthalpy than the double
in time is shown in Fig. 11a, and the heat transfer distribution cone. The experimental data include surface pressure, heat
during the latter 5 ms of the test is shown in Fig. 11b. The transfer, holographic interferometry and Schlieren images. The
experimental heat transfer distribution reaches a steady state. estimated experimental uncertainty is shown in one of the five
The computed surface pressure and heat transfer for this config- sets of computed results for simplicity.
uration by Nompelis is shown in Fig. 11c and d, respectively, and
shows close agreement with experiment. The computed flowfield
reaches a steady state. 5.1. HEG I
experiment. This is a particularly significant result since HEG I is the Reynolds number was reduced by an approximate factor of two.
highest enthalpy case considered in the study. Virtually identical The computed flowfield for this case was observed to be steady in
results are obtained by Nompelis (Fig. 12c) for surface accommoda- agreement with experiment, and the computed surface pressure
tion factor g ¼ 0 (non-catalytic) and g ¼ 1 (fully catalytic for radi- and heat transfer displayed close agreement with experiment.
cals). All computed flowfields converge to steady state. The predicted heat transfer and surface pressure for the low
The computed and experimental surface heat transfer are enthalpy cylinder are in good agreement with experiment. For the
presented in Fig. 13. There is a consistent underprediction of heat high enthalpy cylinder case, the predicted pressure profiles show
transfer for all simulations with the exception of the computation close agreement with experiment while the predicted heat
by Nompelis with g ¼ 1 (where g refers to the catalytic efficiency transfer underpredicts the experiment, thus indicating the need
at the wall). These results indicate that accurate modeling of for accurate modeling of the effects of surface catalycity.
surface catalysis is critical to the prediction of the peak heat
transfer for this case.
[20] Druguet M-C, Candler G, Nompelis I. Effect of numerics on Navier–Stokes [33] Dunn M, Kang S-W. Theoretical and experimental studies of reentry plasmas.
computations of hypersonic double-cone flows. AIAA Journal 2005;43: Contractor Report 2232. NASA; 1973.
616–23. [34] Grossman B, Cinnella P. Flux-split algorithms for flows with non-equilibrium
[21] Wright M, Bose D, Candler G. A data-parallel line relaxation method for the chemistry and vibrational relaxation. Journal of Computational Physics
Navier–Stokes equations. AIAA Journal 1998;36:1603–9. 1990;88:131–68.
[22] Nompelis I, Candler G. Numerical investigation of double-cone flow experi- [35] Chase M, Davies C, Downey J, Frurip D, McDonald R, Syverud A. JANNAF
ments with high-enthalpy effects. AIAA Paper 2010-1283. American Institute thermochemical tables. 3rd ed. American Chemical Society and American
of Aeronautics and Astronautics; 2010. Institute of Physics for the National Bureau of Standards; 1986.
[23] Schwamborn D, Gerhold T, Heinrich R. The DLR TAU-Code: recent applica- [36] Panesi M. Physical models for nonequilibrium plasma flow simulations at
tions in research and industry. In: Wesseling P, Onate E, Périaux J, editors. high speed re-entry conditions. PhD thesis. Italy: Pisa University; 2009.
ECCOMAS CFD 2006. The Netherlands: TU Delft; 2006. [37] Blottner F, Johnson M, Ellis M. Chemically reacting viscous flow program for
[24] Reimann B, Hannemann K. Numerical investigation of double-cone and multi-component gas mixtures. Technical Report SC-RR-70-754. Albuquer-
cylinder experiments in high enthalpy flows using the DLR TAU code. AIAA que, New Mexico: Sandia National Laboratories; 1971.
Paper 2010-1282. American Institute of Aeronautics and Astronautics; 2010. [38] Bottin B. Aerothermodynamic model of an inductively-coupled plasma wind
[25] Holden M. Private communication; 2009. tunnel. PhD thesis. Université de Liége; 1999.
[26] Courant R, Friedrichs KO. Supersonic flow and shock waves. 3rd ed. Springer- [39] White F. Viscous fluid flow. New York: McGraw Hill; 1974.
Verlag; 1976. [40] Barbante P. Accurate and efficient modeling of high temperature none-
[27] Gnoffo P, Gupta R, Shinn J. Conservation equations and physical models for quilibrium air flows. PhD thesis. Von Karman Institute; 2001.
hypersonic air flows in thermal and chemical nonequilibrium. Technical [41] Nompelis I. Computational study of hypersonic double-cone experiments for
Paper 2867. NASA; 1989. code validation. PhD thesis. University of Minnesota; 2004.
[28] Gupta R, Yos J, Thompson R, Lee K. A review of reaction rates and [42] Wilke C. A viscosity equation for gas mixtures. Journal of Chemical Physics
thermodynamic and transport properties for an 11-species air model for 1950;18:517–9.
chemical and thermal nonequilibrium calculations to 30 000 K. Reference [43] Magin T, Degrez G. Transport properties for partially ionized unmagnetized
Report 1232. NASA; 1990. plasmas. Physical Review E 2004;70.
[29] Park C. On convergence of computation of chemically reacting flows. AIAA [44] Magin T, Degrez G. Transport algorithms for partially ionized unmagnetized
Paper 1985-0247. American Institute of Aeronautics and Astronautics; 1985. plasmas. Journal of Computational Physics 2004;198:424–49.
[30] Park C. Assessment of two-temperature kinetic model for dissociating and [45] Hirschfelder J, Curtiss C, Bird R. Molecular theory of gases and liquids. 2nd ed.
weakly ionizing nitrogen. AIAA Paper 1986-1347. American Institute of New York: John Wiley & Sons; 1963.
Aeronautics and Astronautics; 1985. [46] Herning F, Zipperer L. Beitrag zur Berechnung der Zähigkeit technischer
[31] Park C. Two temperature interpretation of dissociation rate date for N2 and Gasgemische aus den Zähigkeitswerten der Einzelbestandteile. Gas- und
O2. AIAA Paper 1988-0458. American Institute of Aeronautics and Astro- Wasserfach 1936;79.
nautics; 1988. [47] Barbante P, Degrez G, Sarma G. Computation of non-equilibrium high-
[32] Park C. Assessment of two-temperature kinetic model for ionizing air. Journal temperature axisymmetric boundary layer flows. Journal of Thermophysics
of Thermophysics and Heat Transfer 1989;3:233–44. and Heat Transfer 2002;16:490–7.