Beruflich Dokumente
Kultur Dokumente
, S.M. Dash
, and Z. Liu
,
Combustion Research & Flow Technology, Inc. (CRAFT Tech)
Pipersville, PA 18947
Phone: 215-766-1520/Fax: 215-766-1524
A scalar fluctuation model (SFM) that solves partial differential equations for
energy/species variance and corresponding dissipation rates is presented, along with several
applications to high-speed fuel/air mixing problems. The model is implemented in a k-
epsilon turbulence model framework with unified compressibility and low Re extensions,
specialized for high speed aero-propulsive flow applications. It is used to predict the spatial
variation of turbulent Prandtl and Schmidt numbers using time-scale relations, providing
more accurate and reliable solutions than those based on user-specified average-values. Over
the past several years, the authors and coworkers have systematically upgraded the SFM to
treat flows of increasing complexity, using a building-block approach to ensure that
modifications made to improve the analysis of more complex cases will not degrade the
model performance in analyzing fundamental cases. A GUI-driven building-block data base
(BBDB) tool has been developed to facilitate the validation/calibration process, which
contains the various data sets we are working with (experimental and LES) along with grids
and solution files, and scripts to take CFD output and put it into the format required to
compare with the data. This paper will describe the latest version of the SFM, its
application to select fundamental cases in the BBDB, and a detailed description of its
analysis of the SCHOLAR fuel/air mixing/combustion data in which we have examined grid
resolution sensitivities and compared results using the SFM with those using different values
of constant turbulent Prandtl and Schmidt numbers.
I. INTRODUCTION
Accurate modeling of scalar transport, such as internal energy and species concentration, is needed to properly
predict the temperature and fuel/air distribution in high-speed aero-propulsive devices, and thus to predict
combustion efficiency and overall performance. The effect of turbulence plays a dominant role in scalar transport
and is accounted for in a RANS framework through the application of a turbulent Prandtl (Pr
t
) and turbulent
Schmidt (Sc
t
) number for thermal and species mixing, respectively. It is now well recognized that the application of
constant-average values of these numbers for complex flows is inadequate since: different values apply to different
very basic flows (e.g., 0.7 is used for round jets, 0.9 is used for boundary layers); these numbers vary widely in
complex flows; and, the values differ with varying levels of compressibility in the flow. Turbulence predictions for
high speed flows are complicated by compressibility effects which have a first-order influence on fundamental
*
Research Scientist, 3313 Memorial Parkway South, Suite 108, Member AIAA.
President & Chief Scientist, 6210 Kellers Church Rd., Associate Fellow AIAA.
t
t x x x x
j j j
2
+ = + +
(2)
2
,
( )
( )
1 2
3 4 5
j e
e e
j j e j e
e e
e
e
u
e
t
C C
t d d
t x x x k k
C P C C
d k d d T
k k k
e
j
x
+ = + + +
+ + +
(3)
where the dissipation rate is defined as,
e
k k
e e
x x
=
(4)
and the heat transfer time scale is,
,
e
m e
e
k k
= (5)
The compressibility correction used in the k- model
7
has the relation,
2 2
P P M P M
k k 1 T k 2 T
=
(6)
In this equation, P
k
is the production term in the transport equation for k and
T
M
(7)
In Eqn. (3),
T
is a near-wall damping function included to capture low-Reynolds number behavior. Details of
this term are provided in Ref. [1] for the temperature variance formulation. An analogous form is used for the energy
variance dissipation rate, with the scalar fluctuation time scale provided by the ratio k
e
/
e
. Constants used in the k
e
-
e
model, summarized in Table 1, are consistent with those proposed by Sommer et al.
11
with the exception of
, k
and
,e
, whose value of 1.0 has been found to provide better agreement with data. Wall thermal boundary
treatments using a wall-function approach
12
or a wall resolved approach
11
are both available.
The turbulent Prandtl number with near-wall damping is given by
e
T
e
C f
k
Pr
C f k
= (8)
and the turbulent thermal diffusivity is calculated from,
,
t
t
T
v
C f k
Pr
m e
= = (9)
where C
=0.14, which differs slightly from the value presented by Sommer et al.
11
to better simulate the turbulent
Prandtl number for jet mixing problems. The coefficients C
and f
= ()
1/4
to define a wall distance y* = u
y/
,
which can be applied in both attached and detached
flows. The expression for the near-wall damping function f
+
= +
(10)
Use of the internal energy as a basis for this model removes the chemical source terms from the transport
equations that appear for the temperature-based formulation. This model has also been formulated so that it reduces
exactly to the nonreacting formulation of Brinckman et al.
1
when the fluid mixture specific heats are constant. The
model predicts both low and high speed flows using a single set of calibration coefficients. For reacting flow
simulations using a pdf-based model, the required temperature fluctuations may then be extracted from k
e
as
described by Calhoon and Kenzakowski
14
.
Table 1. Scalar Variable Model Constants
C
d1
C
d2
C
d3
C
d4
C
d5
C
1
A
+
k,e
,e
2.0 0.0 0.72 2.2 0.8 0.1 45 1.0 1.0
C. Mixture Fraction Variance Equations
Transport equations for the mixture-fraction variance k
f
=<ff> and its dissipation rate
f
are similar to the
energy variance model, except that the source term on the k
f
equation is based on the mixture fraction f, where f is
generally taken to be the effective dilution ratio of an inert species (e.g., f=1 in the unmixed fuel jet and f=0 in the
surrounding airstream). The mixture fraction variance equations are given as:
2
,
( )
( )
2
f
f f
t
f
j j k f j j
u k
k k
D f j
D D
t
t x x x x
+ = + +
(11)
2
,
( )
( )
1 2
3 4 5
f
f f
j j f j f
f f
f
f
u D
f j
t
D D C C
t d d
t x x x k k
C P C C
d k d d f
k k k
f
j
x
+ = + + +
+ + +
(12)
where f is tracked as a conserved scalar. In non-reacting problems, the mass fraction of an individual species
behaves as a conserved scalar, and a mixture fraction can be derived by normalizing one of the species mass
fractions. This approach provides computational efficiency by eliminating the need to solve an additional transport
equation for mixture fraction. In reacting problems, the same approach can be used with any inert species, as its
mass must be conserved. In the absence of an inert species, an additional transport equation for the mixture fraction
itself must be solved to assure that the variance k
f
is based on a conserved scalar (see Ref. [2]).
Here, the dissipation rate is defined as
f
k k
f f
D
x x
=
(13)
and the mass transfer time scale, based on the turbulent velocity and species mixing time scales, is
,
f
m f
f
k
k
= (14)
The turbulent mass diffusivity is calculated from:
,
t
t
t
D C f k
Sc
m f
= = (15)
American Institute of Aeronautics and Astronautics
4
and the turbulent Schmidt number with near-wall damping is:
f
t
f
C f
k
Sc
C f k
= (16)
where the model parameters C
, C
, f
and f
are as described earlier for the energy variance model. The model
constants for the mixture fraction variance are those used in the energy variance model as given in Table 1.
III. Building-Block Data Base (BBDB) Tool
In calibrating and validating the SFM, a building-block approach is now being used where we start with very
basic cases and work our way forward analyzing cases of increasing complexity. If we need to add corrections or
modify the SFM for more complex cases to get good agreement with data, we have to go back to the more basic
cases and rerun them to ensure that the changes made do not degrade the earlier comparisons. This can be a very
time-consuming and tedious process, which we have now expedited by constructing a GUI-driven BBDB tool. This
tool has the basic features summarized in Table 2 below, with some of the cases we have included shown in Table 3.
Figures 3 and 4 show case descriptions, which summarize some of the cases included in the BBDB. For each case,
we have the data (experimental or LES) in line plot format and scripts for the CFD solution files that provide direct
graphical comparisons with the data with
no user intervention. Several cases can
be run sequentially, so that after one
makes some modifications to the SFM, a
sequence of runs can be performed
overnight with the comparisons available
in the morning, plotted as per the scripts
set up for each case. The BBDB tool has
also proven useful in comparing results
of different codes and we have been
developing code-specific interface
routines to facilitate making such
comparisons. Once we have completed
installation of the BBDB on a dedicated
web-site, outside users will be able to
use it as well, as shown schematically in
Figure 2.
Table 2. Building Block Data Base (BBDB) Tool
Tool kit / data base provides:
user-friendly / web-based environment for accessing and using validation case files
detailed description of validation case
files with available experimental or LES data
solution files ( using various turbulence models, etc)
grid file
restart file and instructions on how to restart (code specific)
current solution file: case specific processing (scripts)
- utilities to format data files for direct comparison
Data organized in a Building-Block manner
automated run sequencing is available so we can repeat a series of most relevant cases to see
how a modeling change impacts comparisons with data and earlier solutions
Figure 2. Schematic of Web-based BBDB System used for SFM
Validation.
American Institute of Aeronautics and Astronautics
5
Table 3. Select SFM Validation Data Sets and GUI Driven Data Retrieval
Figure 3. Beach Hydrogen/Air Coaxial Reacting Jet
Data Set.
Figure 4. SCHOLAR Angled Fuel Injection
Combustion Data Set.
American Institute of Aeronautics and Astronautics
6
IV. Fundamental Case Studies
A. Compressibility-Effects Air/Air Studies
The effects of compressibility on jet mixing problems are quite significant, and can be illustrated by the
comparisons shown below for a round, Mach 2 jet into still air (Seiner
15
), and for a transverse jet exhausting into a
supersonic approach flow with boundary layer (Spaid and Zukoski
16
). Including the unified compressibility
correction of Reference [7] into the k- turbulence model reduces the rate of jet decay, and increases the extent of the
transverse jet induced separation bubble, in agreement with the data, as shown in Figures 5 and 6 below.
Figure 5. Analysis of Seiner Mach 2 Jet into Still
Air with and w/o Compressibility-Correction (CC).
Figure 6. Analysis of Spaid and Zukoski Transverse Jet
Data with and w/o CC
B. Low Speed vs. High Speed Hot, Round Jet Behavior
Our studies using SFM have indicated fundamental differences in the behavior of subsonic (Figure 7a) and
supersonic (Figure 7b) hot jets. For hot subsonic jets, the velocity and temperature fluctuations behave in similar
manner as their intensities asymptote toward the same level after the mixing reaches the axis (as confirmed by
comparisons with data sets such as those of Lockwood and Moneib
17
). In contrast for supersonic jets, the velocity
fluctuations on the centerline asymptote, but the temperature fluctuations peak and then decay. The SFM predicted
trends are confirmed by comparisons with an LES calculation for the supersonic jet in Figure 7b.
(a). Hot Subsonic J et (b). Hot Supersonic J et
Figure 7. Low Speed vs. High Speed Hot, Round Jet Behavior, Fluctuation Intensity
American Institute of Aeronautics and Astronautics
7
C. Scalar Fluctuation Differences in Variable Density Subsonic and Supersonic Round Jets
As shown in the comparisons below (Figure 8)
performed using SFM, for subsonic jets with the same
density ratio (produced by either temperature or
molecular weight), the temperature and species
fluctuations are identical. This should be contrasted
with the behavior in supersonic jets where it is seen
that temperature fluctuation levels are significantly
higher than species fluctuation levels, and they peak
and decay rather than flattening after the mixing
reaches the axis. This difference in behavior
exemplifies how compressibility affects scalar
fluctuations, and modeling this behavior properly is a
matter of on-going research, as we look to analyze
more challenging supersonic flows, with shocks and
other complexities.
Figure 8. Subsonic and Supersonic Variable Density
Ratio Jets.
V. Basic Fuel / Air Mixing and Combustion Studies
A. Beach/Evans Coaxial H
2
/Air Jet Study
In the experiment of Evans et al.
18
, a Mach 2 hydrogen jet is exhausted into a co-axial stream of vitiated air at
Mach 1.9. This is a basic coaxial H
2
/air high-speed jet mixing and combustion case with no geometric complexities.
The jet issues from a nozzle with inner diameter d =0.6525 cm and outer diameter d
o
=0.9525 cm. A schematic of
the experimental apparatus and flow conditions for the inner and outer jets is provided in Figure 3. The case was
analyzed using a basic H/O reaction mechanism (7 species/8 reactions) with both the Prandtl and Schmidt numbers
prescribed (Pr
t
=Sc
t
=0.7), and using the SFM model. Figure 9(a) and (b) show the predicted variation of H
2
along
the jet centerline, and a profile of H
2
O at a location about 22 diameters downstream of the nozzle exit plane. The
predicted contours of Pr
t
and Sc
t
for this case are shown in Figure 10 and do not vary significantly, with values
ranging from about 0.55 - 0.75 in the developed jet. Using averaged values of 0.7 appears to do an adequate job for
this basic case.
(a). H
2
variation along jet Centerline (b). H
2
O profile at x/d ~22
Figure 9. Beach/Evans Coaxial H
2
/Air Jet, Comparison of Predictions to Data.
American Institute of Aeronautics and Astronautics
8
(a). Turbulent Prandtl Number (b). Turbulent Schmidt Number
Figure 10. Prandtl and Schmidt number variations for Beach/Evans Jet.
Figure 11. Schematic of experimental set-up
for the Burrows and Kurkov
19
combustor.
B. Burrows and Kurkov H
2
Slot Jet Ducted Mixing and Combustion Study
This experiment consisted of sonic hydrogen injection
through a backward facing step into a Mach 2.44 vitiated air
(or with N
2
in the non-combusting case) stream. Figure 11
presents a schematic of the experimental set-up. The flow
conditions for this case are summarized in Table 4. Contours
of Prandtl, Schmidt and Lewis number are shown in Figure
12. For the mixing case, the variations are seen to be modest,
and the analysis of this case with constant values has been
shown to be adequate. For the combusting case, note the
pronounced change in these transport parameters at the
ignition location which has led to markedly improved
comparisons with the data (see Refs. [2]-[4]). The response of
the SFM to the changes in turbulence produced by heat release due to combustion is an extremely promising feature
of this model and has led to improvements in comparisons with data for several case, including the SCHOLAR data
to be discussed below.
Mixing Experiment Combustion Experiment
Figure 12. Burrows & Kurkov Study Contours of Pr
t
, Sc
t
and Le
t
.
American Institute of Aeronautics and Astronautics
9
Table 4. Vitiated air and fuel exit conditions for Burrows and Kurkov
19
experiment.
Vitiated Air Hydrogen
Static Pressure 101.325 kPa 101.325 kPa
Static Temperature 1270 K 242 K
Mach Number 2.44 1.0
H
2
mass fraction 0.0 1.0
O
2
mass fraction 0.258 0.0
N
2
mass fraction 0.486 0.0
H
2
O mass fraction 0.256 0.0
VI. Analysis of SCHOLAR Data
A. Overview
The SCHOLAR data set (see Figure 4) is intended to validate CFD models in an environment which typifies that
in a scramjet combustor, with flush/angled injectors. The 3D SCHOLAR angled fuel injector mixing and
combustion experiments were conducted at NASA Langley Research Center Ref. [20-21], providing data for CFD
model validation, including mean temperature, wall pressure, and species measurements. These data sets are useful
for analyzing the capabilities of the SFM to capture turbulent mixing effects in more complex flows involving
hydrogen injection into three-dimensional cross-flows. Data was obtained at 4 stream-wise planes downstream of
the injector, with the X-location of each stream-wise plane provided in Table 5.
Table 5. X-location of planes.
Plane X-location
3 274.2 mm
5 426.6 mm
6 777.0 mm
7 1234.2 mm
Computational analyses by Rodriguez and Cutler
22
showed sensitivity to boundary conditions including
temperature, species composition, and inflow turbulence parameters. They found their solutions to be most sensitive
to the values used for Pr
t
and Sc
t
, and concluded that it is unlikely that constant values for these parameters can be
calibrated to be valid for a wide range of flows. They suggested that the use of models providing for variable Pr
t
and Sc
t
values are necessary to properly simulate the SCHOLAR combustion experiment. Keistler et al.
23
proposed
a model for variable Pr
t
and Sc
t
based on enthalpy and concentration variances, which they used to analyze the
SCHOLAR combustor experiment but comparisons with the data have not shown overall consistency.
Our analysis predicts variable Pr
t
and Sc
t
using the SFM formulation, which includes compressibility
corrections. The facility nozzle from the upstream heater was modeled to provide inflow boundary conditions for
mean and turbulence quantities, to the test section. Near-wall modeling was used for no-slip walls, along with a
constant wall temperature. Early analyses considered the effect of wall temperature and it was found to have a
limited impact on results. This correlates well with the results of Ref. [22], which showed a minor effect from the
temperature of the wall.
B. Grid Sensitivity
A systematic approach was used to model the
SCHOLAR experiment. Three levels of grid refinement
were performed, denoted as Grid A, Grid B, and Grid C
for subsequent levels of refinement. A representative
view of Grid C in the vicinity near the injector is
provided in Figure 13. This domain was decomposed
into an isolator section upstream of the injector, the
injector section containing the angled fuel injector, and a
downstream combustor section. The injector nozzle was
included in a block that extended into the duct. Grid
dimensions are shown in Table 6. The large increase in
injector nozzle grid points was primarily directed at
American Institute of Aeronautics and Astronautics
10
Figure 13. Schematic of SCHOLAR grid near
injector ( every other grid point shown), Plane 5 (X =
426.6mm) shown for reference.
better resolving the nozzle wall boundary layer and turbulence production, with the intent of improving early jet
core mixing predictions.
Table 6. Scholar Grid Dimensions.
Grid A Grid B Grid C
Duct Block 245x95x33 305x113x41
565x145x65
Isolator(black) 209x145x65
Injector(green) 173x145x65
Combustor(purple) 201x145x65
Injector Nozzle Block 134,976 cells 449,196 cells 1,258,353 cells
Total Cells 868,928 1,811,116 6,456,177
Figure 14. SCHOLAR mixing experiment, Wall
pressure grid sensitivity.
Mixing and combustion cases were considered for all
three grids using SFM. Comparisons of lower wall pressure
for the three grids (Figure 14) for the mixing case primarily
show a sharpening of the peaks, particularly downstream of
injection. The general peak locations, however, are basically
the same. Comparisons of results to contours developed from
experimental data are shown in Figure 15 and indicate that
grid refinement helps to better capture the radial extent of the
mixing, but the jet core mixing is being somewhat under
predicted. A three-fold increase in mesh points from Grid B
to Grid C did not significantly improve the core mixing
prediction. This type of trend has been noted in comparisons
to other data sets involving transverse fuel injectors
24
and
work is ongoing to understand the model performance in
these mixing configurations.
(a) mean temperature
(b) mole fraction N
2
Figure 15. SCHOLAR mixing experiment, Comparison of CRAFT predictions to measured data at Plane 5,
grid sensitivity.
American Institute of Aeronautics and Astronautics
11
Wall pressure comparisons to data for all three grids for the combustion case are shown in Figure 16.
Substantial improvement is noted in going from Grid A to Grid C downstream of x =400mm. Grid A was very
coarse stream-wise in this section due to the long combustor, which caused a much more diffusive flow-field than
that of the more refined grids. Contours of mean temperature shown in Figure 17 indicate combustion in Grid A
starting to take place well upstream of the other two grids. However, the amount of combustion taking place is
much less, which coincides with the wall pressure result. The combustion onset location along the top wall for Grid
B and Grid C are in similar locations, providing some consistency in the results. Figure 18 shows mean temperature
and N2 mole fraction contours compared to measured data for all three grids. As in the mixing case, the core jet
mixing is under-predicted, even with the substantial increase in the number of grid points. More combustion
appears to be taking place on Planes 6 and 7 of Grid C compared to the experiment, but the overall contours provide
a better match to data than the coarser grids. These results clearly reveal an improved match to data with grid
refinement going from Grid A to Grid C. However, grid convergence
25
has yet to be realized. Work is progressing
on a formal method of grid convergence, which will be employed in further analyses of this experiment
26
.
(a) lower wall (b) upper wall
Figure 16. SCHOLAR combustion experiment, Wall pressure grid sensitivity.
Figure 17. SCHOLAR combustion experiment, Temperature grid sensitivity.
American Institute of Aeronautics and Astronautics
12
(a) Mean temperature
(b) mole fraction N
2
Figure 18. SCHOLAR combustion experiment, grid-sensitivity comparison.
American Institute of Aeronautics and Astronautics
13
Figure 19. SCHOLAR mixing experiment, wall
pressure predictions vs. data.
C. SCHOLAR Supersonic Mixing Experiment
Having established sensitivities to grid resolution, we
now focus on SFM model comparisons versus those with
constant Pr
t
and Sc
t
, with all calculations performed on
Grid C. The CRAFT CFD
using
variable and constant values of Pr
t
and Sc
t
equal to 0.9 is
compared to measured data from the SCHOLAR mixing
experiment in Figure 19. There is good overall agreement
with the data except for the early peak pressures, which appear to be due to shocks upstream of the injector. These
missed shocks also seem to cause the curves to be shifted downstream of the data. Geometry refinement was
performed to include the upstream facility nozzle to try to capture these shocks, but the two peaks upstream of x =
200mm could not be reproduced in work to date. The same inconsistency to experimental data was observed in
simulations documented by Danehy et al.
20
, suggesting that there may be some physical feature in the experimental
facility that is not being included in relatively simple isolator models. Sensitivity analyses and model refinements
are being considered which may reveal the source of this discrepancy with the data, and its impact on jet mixing.
Results from predictions using both variable and constant values of Pr
t
and Sc
t
equal to 0.9 are compared for
mean temperature and nitrogen mole fraction at a streamwise plane corresponding to x 427mm. As with the wall
pressure, the contours of mean temperature and nitrogen mole fraction between variable and constant Pr
t
and Sc
t
in
Figure 20 are very similar. Contours of local Pr
t
and Sc
t
are shown in Figure 21, exhibiting a range of Pr
t
and Sc
t
from 0.8 to 1.0 in the jet region, indicating that a constant value of 0.9 is an appropriate choice in the downstream
mixing region.
Figure 22 shows CRAFT predictions using variable Pr
t
and Sc
t
compared to the measured data. The general
shape of the jet region is similar between the two, but the extent of jet core mixing is larger in the data. As
mentioned earlier, this discrepancy with data does not seem to be related to grid resolution. While use of constant
values of Pr
t
and Sc
t
for the mixing case has a minimal effect compared to use of the SFM, it has a significant effect
on comparisons to data for the combusting case, as will be discussed.
(a) mean temperature (b) mole fraction N
2
Figure 20. SCHOLAR mixing experiment, Comparison of CRAFT predictions for variable and constant Pr
t
and Sc
t
.
American Institute of Aeronautics and Astronautics
14
(a) turbulent Prandtl number (b) turbulent Schmidt number
Figure 21. SCHOLAR mixing experiment, Predicted Pr
t
and Sc
t
.
(a) mean temperature (b) mole fraction N
2
Figure 22. SCHOLAR mixing experiment, Comparison of CRAFT predictions with measured data.
D. SCHOLAR Supersonic Combustion Experiment
Comparisons of results predicted with CRAFT CFD