Professional Documents
Culture Documents
code
applied for most of the
flowfield, but with the
UNS numerics of the
CRUNCH CFD
code
required where fine-
scale features must be
resolved, such as in the
vicinity of transition
trips on the forebody
and in the fuel injector
regions of the
combustor. Several levels of grid
adaptation are often required for which the
CRISP CFD
code
2
is utilized. Figure 2
shows product formation contours at
several stations in the elliptical combustor
of a generic inward-turning scramjet,
where fuel is injected from the walls using
flush, angled injectors.
Table 1. Codes Utilized for Scramjet Applications
The combustion efficiency, obtained
with the original and adapted UNS grids
indicates that unless the grid is adapted, we
overestimate this efficiency (due to
numerical diffusion effects). Since there is
only half-plane symmetry, the fuel/air
mixing details for over 10 injectors must be
resolved. Using conventional, cell-splitting
adaptation can lead to a large increase in
the number of nodes (see Table in Figure 2
we go from 7M to 13.5M cells in Pass 1
which did provide a grid resolved solution
since Pass 2 results were essentially
Figure 2. Grid adaption effects on combustion efficiency and grid size.
3
American Institute of Aeronautics and Astronautics
AIAA-2006-8037
identical). We have been working on hybrid cell splitting/cell stretching adaptation concepts which remedy this
issue but still require further development to resolve tangling issues
11
.
While conventional load balancing (same number of cells per domain) using domain decomposition with MPI is
effective for aerodynamic problems, it is problematic for scramjet combustor problems where varied zones in the
flow have differing work loads. Using iterative matrix-split chemical kinetic techniques, we may be performing 100
iterations per CFD time-step in fuel injection regions (where ignition reactions have small time-scales), and minimal
iterations in other regions. Using dynamic load balancing, based on work per node,
3
has led to effective load
balancing in such situations and is being supplemented by use of tabulated procedures (ISAT, ANN, etc.). The fuel
injection region of combustors can be very CPU intensive, particularly for problems where ignition kinetics is
required, requiring 10-15M cells and solving 20 or more coupled pdes (5 gas dynamic, 9 or more chemical species
and 8 turbulent/ transition model equations see next section). Analysis of this region takes about of the overall
CPU time (including grid adaptation) and full end-to-end runs on 256-512 processors are routinely performed in
several days.
III. Models Utilized
The modeling utilized in performing scramjet simulations is summarized in Table 2. Basic models are used for
preliminary design and to
expedite early stages of
design optimization.
Advanced models are used
for interpreting experimental
data and for refining designs.
Transitional models, which
solve pdes for onset location
and for intermittency (3D
blending from laminar to
turbulent flow), have been
calibrated for hypersonic
flows and compared with
varied shock tunnel data sets
4
. These models account for
the fluctuation noise levels in
test facilities (prescribed as
inflow condition) and are
extremely useful in
suggesting trip location modifications in going from full-scale testing to a quieter flight environment. Analyzing the
transitional flow downstream of trips using the intermittency model is still developmental and has required some
local modifications to the source terms used in the low Re turbulence models. Referring to Figure 3, it has proven
important to both resolve the grid surrounding individual trip elements (Figure 3a), and to modify the turbulent
production levels in the immediate vicinity of the trip to produce results consistent with the measured heat transfer
levels behind the trip (Figure 3b). As per the discussion in the introduction, doing this properly has strong
implications on mixing/combustion and overall performance.
Table 2. Basic and advanced models used for scramjet applications
0
10
20
30
40
50
60
70
80
90
100
110
120
0 10 20 30 40 50
Test Data
CFD, Trips
CFD, TripswBoost
(a) Multi-Element UNS grid (b) Localized Pressure gradient correction to low Re
Turbulence Model
Figure 3. Modeling Trip Effects on Transitional Behavior.
4
American Institute of Aeronautics and Astronautics
AIAA-2006-8037
Our transitional modeling work is focused on being able to use CFD to bridge the gap in designing a flight
vehicle, using full-scale ground test data as the primary means of transitional model calibration and validation. The
transitional model has already demonstrated the ability to deal with differences in noise levels and wall
temperatures, and we are making very good progress in treating pressure gradient effects associated with nose
bluntness and mild compressions.
Scalar fluctuation modeling is playing a very dominant role in obtaining accurate values for fuel/air mixing. The
most recent form of the turbulent pdes implemented
, - 13 15
solve for energy and species variance and related
dissipation rates, and are used to obtain local values of turbulent Prandtl (Pr
t
) and Schmidt (Sc
t
) numbers, governing
thermal and species turbulent diffusion. Most CFD codes require specification of constant values for these
parameters but their values in fuel injection regions can vary substantially as shown by the Sc
t
contours in Figure 4
Calibration and validation of these models suffers from lack of scalar fluctuation data in high speed flows and LES
solutions of unit fuel injection problems
9
are providing supportive data. Figure 5 shows one such LES solution with
contours of time-averaged mean flow (u, T, Y
H2
) and corresponding RMS fluctuations exhibited. This case
represented H2 angled fuel injection into a high speed airstream emulating the environment in a Mach 10
combustor.
Figure 4. Schmidt number contours fuel injection in high speed stream comparisons.
Figure 5. LES simulation of fuel injection flowfield.
5
American Institute of Aeronautics and Astronautics
AIAA-2006-8037
While RANS comparisons with the LES results for mean and fluctuating velocity and temperature were
quite promising, species comparisons were not as favorable, with RANS species fluctuation levels significantly
underpredicting those of the LES calculation (see Ref. 9). An inspection of these results indicated that species
fluctuations in the RANS solution were too low and not showing any dependence on pressure gradients, as was
evident in the velocity/temperature fluctuations in the RANS and LES solutions, and in the species fluctuations in
the LES solution. A pressure gradient correction term has been added to the species fluctuation equation that greatly
improves the comparisons as shown in Figures 10-12. Figure 6 compares species fluctuation contours at several
axial stations downstream of the injector with the original formulation (as used in Ref. 9), and, with the new
formulation with the pressure correction.
x =1
Orig. p
x =3
Orig. p
x =5
Orig. p
x =7
Orig. p
Figure 6. Comparison of RANS species fluctuation levels with original and new pressure-gradient corrected
formulation.
It is seen that fluctuation levels are significantly higher using this correction, and that the contours show
more spread. In Figure 7, we compare fluctuation levels using the pressure correction with the LES results. Both
fluctuation levels and overall spread are now in reasonable accord with the LES results. Finally, we show in Figure
8, Schmidt number contours obtained using the original and new pressure-corrected formulation, with the new
formulation yielding contours that appear more realistic and are in reasonable accord with the Prandtl number
contours (see Ref.9). We are currently analyzing several fuel/air mixing data sets with this improved model and will
be presenting these comparisons in an upcoming paper
16
.
6
American Institute of Aeronautics and Astronautics
AIAA-2006-8037
x =1
LES p
x =3
LES p
x =5
LES p
x =7
LES p
Figure 7. Comparison of RANS pressure-gradient corrected species fluctuations with LES results.
x =1
Orig. p
x =3
Orig. p
x =5
Orig. p
x =7
Orig. p
Figure 8. Schmidt number comparisons original vs. pressure-gradient corrected formulation.
7
American Institute of Aeronautics and Astronautics
AIAA-2006-8037
IV. Studies Supporting System Design
Earlier studies had been performed for rectangular scramjet designs, with interdigitated flush and wedge
injectors (see J ANNAF Combustion Science Meeting papers presented over the past several years), to assess the
ability of the CFD codes of Table 1 to reproduce data obtained in varied LENS shock tunnel tests performed at full-
scale and for duplicated flight conditions. These studies exhibited the need to develop advanced transitional and
scalar fluctuation models, and to use grid adaptation in transitional trip and fuel injector regions. The conclusion
from these varied comparative studies was that CFD could provide very reasonable comparisons with data under
conditions where strong burning occurred without ignition aids. Substantive DoD computer (AMRDEC and HPCC)
resources were used to support a substantive number of end-to-end comparative studies and were key to the progress
made in generating accurate solutions in a timely manner.
Of most recent interest in our Army scramjet work are inward-turning concepts, which utilize a sugar-scoop
inlet which compresses the flow in a shock-free manner into an elliptical combustor with flush, angled injectors.
These concepts have been tested in the Army/CUBRC LENS facility with an inward-turning conceptual flowpath
shown in Figure 9,
17
. Such propulsive flowpaths are now being integrated into hypersonic missile designs and
numerous CFD calculations have been performed in support of systems studies. This flowpath is not straightforward
to calculate and UNS numerics are beingused for the entire end-to-end calculation. While on-design inlet
performance is generally quite good, achieving good combustion efficiency in a reasonable length is a design
concern. A basic issue is that of fuel injector design and spacing with studies for fixed fuel/air stoichiometry and
injector size shown in Figure 10. Using a smaller number of injectors provides good penetration but air passes
between injectors, while using a larger number of injectors results in poor penetration and unburnt air in the central
core. In this preliminary study, ten injectors worked best but our solution is far from optimal.
Figure 9. Inward turning model and flowpath.
Figure 10. Circular combustor injector design study.
V. Fuel Injector Optimization
To optimize fuel injector patterns/conditions to yield the highest combustion efficiency for a fixed combustor
length, multi-variate genetic-based optimization
7
is being used, with a schematic of the GUI-driven framework
8
American Institute of Aeronautics and Astronautics
AIAA-2006-8037
implemented shown in Figure 11. The genetic optimization procedure has been coupled to the CRUNCH CFD
code and the grid generation package, GRIDGEN. The methodology was chosen because the search procedure is
inherently parallel and has worked well in a number of earlier multi-variant design applications. Figure 12 shows the
fuel injector optimization performed for a rectangular combustor with upper and lower interdigitated, flush fuel
injectors. Parameter space included spacing, cant angle and off-set with the fuel/air ratio (=1.2) and injector diameter
kept constant. Each case was performed on 64 processors, with 5 cases running simultaneously and 6 design levels
performed to date yielding a close-to-optimal solution.
Figure 11. GUI Driven automated optimization of CFD problems with several variables, constraints and
objectives with complex design landscapes.
Figure 12. Fuel injector optimization study.
9
American Institute of Aeronautics and Astronautics
AIAA-2006-8037
VI. Conclusion
This paper has discussed computational modeling advances which we have found to be needed to reproduce
hypersonic scramjet propulsive flowpath data sets obtained with full-scale models at duplicated flight conditions in
the LENS shock-tunnel facility. Transitional modeling and trip design has been an area of major concern, and the
well-calibrated hypersonic engineering model being utilized is showing great promise for use in transitioning from
ground test to flight, since it includes effects of noise levels and wall temperatures, which are the principal
differences in these two environments. In view of the marked sensitivity of mixing/combustion to inflow (shock and
transitional) characteristics (as discussed in the introduction), the role and utility of subscale (ground or flight)
and/or direct/semi-direct ground tests in supporting practical flowpath design needs to be more carefully examined.
We have found that the user-specified selection of constant turbulent Prandtl and Schmidt numbers poses a
major uncertainty in the prediction of mixing/combustion efficiency, with predictions most often requiring the
tuning of these values (and possibly using zonal strategies) to match varied data sets. Inclusion of scalar
fluctuation models which predict these transport parameters removes this uncertainty and is providing more
consistent agreement with varied data sets in a fully-predictive manner.
Design optimization has been focused on optimizing fuel injector patterns and operational characteristics. This
multi-variate process is CPU intensive since each design case requires a bare-minimum of 64 CPUs and we
would like to run 5-10 cases simultaneously, for 5-10 design levels(25-100 3D calculations). Efficiency
enhancements to the CFD such as dynamic load balancing and use of chemical-kinetic tabulated procedures are
helping, with other enhancements now in a developmental state.
Substantive design-oriented developmental work is now in progress that includes:
1. trip design/optimization (including fluidic concepts that can pre-condition the fuel);
2. transpiration modeling (supported by LENS shock-tunnel tests);
3. examination of alternate fuel concepts to reduce H
2
volumetric requirements; and,
4. integration of inward-turning propulsive flowpath designs with vehicle concepts for varied mission
objectives.
This developmental work will be discussed in upcoming papers.
Acknowledgments
The work described in this paper has been primarily supported by several U.S. Army Research, Development,
and Engineering Center (USARDECOM) programs monitored by Dr. Billy J . Walker, AMSRD-AMR-SS. The
authors acknowledge contributions made by: Billy J . Walker (USARDECOM) and Dr. Michael Holden (CUBRC) in
interpreting varied data sets; by Dr. Chandrasekhar Kannepalli (CRAFT Tech) in performing the LES calculations
used to calibrate the scalar-fluctuation models; and by Drs. Kevin W. Brinkman and William H. Calhoon, J r.
(CRAFT Tech), in their continued support for applying the RANS scalar-fluctuation models.
References
1
Holden, M.S., Studies of Scramjet Performance in the LENS Facilities, AIAA Paper 2000-3604, 36th
AIAA/ASME/SAE/ASEE J oint Propulsion Conference, Huntsville, AL, J uly 17-19, 2000.
2
Cavallo, P.A., and Grismer, M.J ., A Parallel Adaptation Package for Three-Dimensional Mixed-Element Unstructured
Meshes, Journal of Aerospace Computing, Information, and Communication, Vol. 2, No. 11, pp. 433-451, November, 2005.
3
Dash, S.M., Perspective On Flowfield Modeling Advances Needed To Support Hypersonic Scramjet Design and
Evaluation, 40th CS, 28th APS, 22nd PSHS, 4th MSS, Charleston, SC, J une 13-17, 2005.
4
Papp, J .L., Dash, S.M., Rapid Engineering Approach to Modeling Hypersonic Laminar-to-Turbulent Transitional Flows,
Journal of Spacecraft and Rockets, Vol. 42, No. 3, pp. 476-475, May-J une 2005.
5
Brinckman, K.W., Kenzakowski, D.C., and Dash, S.M., Progress in Practical Scalar Fluctuation Modeling for High-Speed
Aeropropulsive Flows, AIAA Paper No. AIAA-2005-0508, 43rd Aerospace Sciences Meeting and Exhibit, Reno, NV, J an. 10-
13, 2005.
6
Calhoon, W.H., J r., Brinckman, K., Tomes, J ., Mattick, S. and Dash, S.M.., Scalar Fluctuation and Transport Modeling
for Application to High Speed Reacting Flows AIAA Paper No 2005-.1452, 44th Aerospace Sciences Meeting and Exhibit,
Reno, NV, J an. 9-12, 2006.
7
Ahuja, V., and Hosangadi, A., Design Optimization of Complex Flowfields Using Evolutionary Algorithms and Hybrid
Unstructured CFD, Paper No. AIAA-2005-4984, 17th Computational Fluids Dynamic Conference, Toronto, Ontario, CA, J un.
6-9, 2005.
8
Calhoon, W.H., J r., Brinckman, K., Kenzakowski, D.C., Sinha, N., and Dash, S.M., Progress In Turbulent Combustion
Modeling For Rocket Plumes, 28th EPTS J ANNAF Meeting, San Diego, CA, Nov. 1-5, 2004.
10
American Institute of Aeronautics and Astronautics
AIAA-2006-8037
9
Ott, J .D., Kannepalli, C., Brinckman, K.W., and Dash, S.M., Scramjet Propulsive Flowpath Prediction Improvements
Using Recent Modeling Upgrades, AIAA Paper No. AIAA-2005-0432, 43rd Aerospace Sciences Meeting and Exhibit, Reno,
NV, J an. 10-13, 2005.
10
Ungewitter, R.J ., Ott, J .D., Ahuja, V., and Dash, S.M., CFD Capabilities for Hypersonic Scramjet Propulsive Flowpath
Design, AIAA Paper No. AIAA-2004-4131, 40th AIAA/ASME/SAE/ASEE J oint Propulsion Conference and Exhibit, Fort
Lauderdale, FL, J uly 11-14, 2004.
11
Brinckman, K.W., Cavallo, P.A., Calhoon, W.H., J r., Ott, J .C., and Dash, S.M. Improved Modeling Tools for High Speed
Reacting Flows, CRAFT Tech Final Report No.: CRAFTR-12/2005.013 (C236), 28 December 2005.
12
Papp, J .L. and Dash, S.M., Turbulence Model Unification and Assessment for High-Speed Aeropropulsive
Flows, AIAA Paper No. 2001-0880, 39
th
AIAA Aerospace Sciences Meeting and Exhibit, Reno, NV, J anuary 8-11,
2001.
13
Brinckman, K., Calhoon, W.H., J r., Mattick, S.J ., Tomes, J ., and Dash, S.M., Scalar Variance Model Validation for High-
Speed Variable Composition Flows AIAA Paper No. AIAA-2006-0715, 44
th
Aerospace Sciences Meeting and Exhibit, Reno,
NV, J an. 9-12, 2006.
14
Calhoon, W.H., J r., Brinckman, K., Tomes, J ., Mattick, S. and Dash, S.M., Scalar Fluctuation and Transport Modeling for
Application to High Speed Reacting Flows AIAA Paper No. AIAA-2006-1452, 44
th
Aerospace Sciences Meeting and Exhibit,
Reno, NV, J an. 9-12, 2006.
15
Brinckman, K. W., Calhoon, W.H., J r., and Dash, S.M., Scalar Fluctuation Modeling for High-Speed
Aeropropulsive Flows, to be published, AIAA Journal, 2007.
16
Mattick, S.J., Calhoon, W.H., J r., Brinckman, K.W., Ott, J .D., and Dash, S.M., Improvements in Analyzing Scramjet Fuel
Injection Problems Using Scalar Fluctuation Modeling Abstract submitted to 45
th
Aerospace Sciences Meeting and Exhibit,
Reno, NV, J an. 8-11, 2007.
17
Walker, S. H., Rodgers, F.C. and Esposita, A.L., Hypersonic Collaborative Australia/United States Experiment
(HYCAUSE) AIAA Paper No. AIAA 2005-3254, AIAA/CIRA 13
th
International Space Planes and Hypersonic Systems and
Technologies, Capua, Italy, May 16-20, 2005.
11
American Institute of Aeronautics and Astronautics