Professional Documents
Culture Documents
, J. Papp
, R. Wilmoth
, and N. Sinha
**
Combustion Research and Flow Technology, Inc., Pipersville, PA, 18947
A hybrid continuum/non-continuum methodology is being developed to model flows in
which rarefaction effects are important while the continuum approach is accurate for parts
of the flowfield. Depending on altitude, rocket and missile exhaust plume expansion can be
modeled using a combination of continuum, direct simulation Monte Carlo (DSMC), or free-
molecular (FM) techniques. The ability to automatically couple these techniques allows for
efficient and accurate modeling of length scales on the orders of kms from the nozzle exit
plane, and parametric studies can easily be performed. The methodology is demonstrated
for an axisymmetric (0 angle of attack) and a 3D (>0 angle of attack) missile configuration.
For the latter calculations, the DSMC solution is coupled to an outer DSMC simulation and
a FM code. The Automatic Efficient Generalized Interface Surface Toolkit (AEGIS Toolkit)
which enables the coupling between the continuum and DSMC codes is described.
Extensions to the methodology for particulates and unsteady flow phenomena are among the
capabilities.
I. Introduction
omputational modeling of rocket and missile exhaust plumes requires accurate physical models and
sophisticated numerical methods. The nearfield plume can be modeled using conventional continuum-based
CFD codes. As this flow expands, it becomes more rarefied. At high altitudes, the interaction of this plume with a
rarefied freestream requires modeling the intermolecular interactions more directly. In this region, the direct
simulation Monte Carlo (DSMC) method is appropriate. When the flow is more rarefied, a free-molecular
approximation is sufficient.
C
Hybrid CFD-DSMC simulations of the plume flowfield typically consist of a continuum nearfield simulated
using continuum-based CFD methods while the rarefied farfield is treated using DSMC methodology. Rarefied
flow modifications are implemented into the continuum flow solver to extend its range of applicability to higher
altitudes and more rarefied conditions, in order to further reduce the computational effort on the non-continuum
solver. An interface surface is generated using some type of continuum/non-continuum breakdown criteria.
Continuum flow values are then interpolated onto this surface and utilized as an inflow boundary for the DSMC
simulation.
Particulates from solid motors are also of interest for high-altitude plume flow characteristics and plume
radiance. Past efforts to simulate particles within the rarefied regime have decoupled the particulate simulation from
the rarefied gas simulation
1
. Now the particulates are included in both the CRAFT CFD
2-3
solution and output this surface with appropriate properties in the format needed by the Direct Simulation Monte
Carlo (DSMC) Analysis Code (DAC) developed at NASA Johnson
4
. However, the methodology and the
procedures used are general. This methodology has been shown previously for zero angle of attack at high altitudes
where axisymmetric calculations are sufficient
5
. The present study includes a generic missile flying at high altitude
*
Research Scientist, 6210 Kellers Church Rd, Pipersville, PA, AIAA Member.
Research Scientist, 6210 Kellers Church Rd, Pipersville, PA, AIAA Member.
Senior Research Scientist, 6210 Kellers Church Rd, Pipersville, PA, AIAA Member.
Principal Research Scientist, 124 Burnham Place, Newport News, VA, AIAA Associate Fellow.
**
Vice President and Technical Director, 6210 Kellers Church Rd, Pipersville, PA, AIAA Associate Fellow.
American Institute of Aeronautics and Astronautics
1
AIAA 2007-5704
at a moderate angle of attack. Since the non-zero angle of attack breaks the symmetry, full 3D simulations are
required. Thus, 3D CRAFT CFD
and DAC calculations are performed. Application of the AEGIS toolkit to this
3D problem is described. The coupling of two DAC flow regions is also described. This coupling allows for a
larger farfield domain without the inefficiencies encountered in a single large DSMC calculation. Results from
coupling DAC results to a free-molecular code are also presented.
II. Methodology
A. CRAFT CFD Navier-Stokes Solver
The Navier Stokes calculations described in this work are performed using the CRAFT CFD
code
2
. CRAFT
CFD
uses is a finite-volume, second or third order upwind Roe TVD formulation. A number of options are
available for thermophysical modeling of high-speed flows, turbulence modeling, and multiphase modeling
(Eulerian or Lagrangian). Some CRAFT CFD
Code
NUMERICS/PARALLEL
PROCESSING
1D/2D/AXI/3D Finite-Volume Discretization
Implicit, ADI and L/U, Higher-Order Upwind (Roe/TVD) Formulation
Fully Implicit Source Terms/Boundary Conditions
PNS Spatial Marching Capability
Domain-Decomposition Parallel Architecture with MPI
Shared Memory Parallelism
Preconditioning Extensions
GRID FEATURES
Grid Dynamics to Account for Moving Boundaries
Grid Patching/Blanking for Complex Geometries
Solution-Adaptive Gridding and Grid Embedding
Noncontiguous Grid Interfacing with Flux Preservation Across Domains
THERMO-CHEMISTRY
Multi-Component Real Gas Mixtures
Finite-Rate Chemistry/Arbitrary Number of Species and Reactions
Fully Implicit Source Term Linearization
MULTIPHASE FLOW Non-equilibrium Particle/Droplet Solvers (Eulerian and Lagrangian Formulations)
TURBULENCE
k /EASM Formulations with Compressibility/Vortical Upgrades
LES Subgrid Scale Models Algebraic and One-equation
RAREFIED FLOW
Slip Boundary Conditions
Vibrational non-local thermal equilibrium energy modeling
B. DAC/PDAC DSMC Solver
The Direct Simulation Monte Carlo (DSMC) Analysis Code or DAC is utilized to perform DSMC calculations
in the present work. The DAC package was primarily developed by LeBeau et al.
4
and is a collection of programs
that are used in the analysis of rarefied flows for either three-dimensional or axisymmetric geometries. DAC
follows the standard DSMC methodology
7
where simulated molecules, each representing many real molecules, are
statistically tracked as they travel through the computational volume and undergo collisions with each other and
with solid surfaces. A grid is used in the computational volume in order to facilitate the selection of nearest
neighbor collision pairs and in order to obtain statistical samples for macroscopic flowfield quantities.
The computational grid is divided into a Cartesian network of Level 1 cells, with equally spaced cells in each
coordinate direction
8
. The cell size is typically defined by the user as the maximum size needed to resolve the flow,
approximately the mean free path of the freestream. For regions of higher gas density, each Level 1 cell can be
independently refined an additional level. These Level 2 cells are themselves a Cartesian grid imbedded in each
Level 1 grid cell with equal spacing in each coordinate direction, so directional bias can be maintained. The Level 2
American Institute of Aeronautics and Astronautics
2
AIAA 2007-5704
grid refinement is required to provide accuracy by maintaining collision modeling cells on the order of a mean free
path. A new feature in DAC allows the refinement to be relaxed by using the nearest neighbor collision pairing
9
.
Automatic grid adaptation varies the number of Level 2 cells in each parent Level 1 cell based on the previous
solutions density. Grid adaptation allows features such as the plume induced separation and plume edge to be
realized as opposed to the original coarse grid solution. The main DAC features are presented in Table 2.
Extensions to DAC, called PDAC, include particulate and unsteady flow capabilities.
Table 2. Features Of DAC
Implements the Direct Simulation Monte Carlo Technique for 3D and Axisymmetric
Simulations
Tracks Molecular Motion Deterministically and Molecular Collisions Probabilistically
General Boundary Conditions for Wide Range of Problems
Domain-Decomposition Parallel Architecture with MPI
Distributed Memory Parallelism
Pre-Processor and DSMC Codes Can Be Run In Scalar or Parallel Modes
NUMERICS/
PARALLEL
PROCESSING
Automatic, Dynamic Load Balancing
Automatic Flowfield Grid Generation Using Two-Level Embedded Grid Technology
Automatic Flowfield Grid Adaptation Based on Previous Solution
GRID FEATURES
Unstructured Grid Definitions of Complex 3D Geometries Embedded Within Cartesian
Flowfield Grid Framework
Multi-Component Real Gas Mixtures
Translational, Rotational, and Vibrational Non-Equilibrium Energy Exchange
Finite-Rate Non-Equilibrium Chemistry/Arbitrary Number of Species and Reactions
Chemical Reactions Modeled on Molecular Level Via Probabilistic Kinetic Theory
Limited Reactions Sets Provided But Additional Sets Can Be Added by User
THERMO-
CHEMISTRY
Extended Reaction Sets for Missile Plume and Divert Jets
1. Particulate Extensions
The one-way coupled version of PDAC implements the same empirical drag and heat transfer models as CRAFT
CFD
, and details of the PDAC implementation are given in Reference [5]. PDAC performs post-DSMC analysis
of the particulate transport based on a previous hybrid continuum-DSMC simulation.
Two-way coupled methods that allow particle to gas momentum and energy transfer as well have been explored.
Reference [10] describes a heat-bath model and compares it to the model by Burt and Boyd (BBG)
11
using a 1D
code for a simple two-phase equilibrium test problem. It is assumed in both of these methods that the number
density of solid particulates is much different than the number density of gas molecules such that different scale
factors, i.e., number of real particles represented by each simulation particle, are required to scale the simulated
densities to the real densities. For typical mass loadings (typically a few percent or less) of micron-sized Al
2
O
3
particles, this assumption is generally true. For example, the number densities of gas molecules are typically of the
order 10
10
times greater than the solid particle number densities due to the large difference in mass per particle
compared to the molecular mass. Therefore, direct simulations with a single number density scale factor for
molecules and particles would be impractical since they would require the total number of simulation molecules and
particles to be greater than 10
10
in order to provide adequate statistics.
The BBG model, described in Reference [11], uses the Gallis free-molecular analysis, modified to account for
internal degrees of freedom in the gas molecule, to transfer energy to the particle. An analogous procedure is used
for drag. The method to treat gas-particle collisions that transfer momentum and energy from the particle to the gas
is analogous to that used for gas-gas collisions in a typical DSMC calculation. However, with different gas and
particle weighting factors, it is shown to conserve energy and momentum on average. The heat-bath model is an
alternate approach that was developed to ensure detailed balance on a macroscopic level within each simulation cell
during each time step. In this method, the energy transferred from the gas to the particle is stored in a heat bath
and later distributed to (subtracted from) the gas molecules. During each time step, the energy transferred to the
solid particle is calculated in the same manner as in the BBG model. However, the total energy transferred to
particles within a cell is stored and the particle temperatures are updated as in the BBG model. A proportional
American Institute of Aeronautics and Astronautics
3
AIAA 2007-5704
amount of energy is subtracted from each gas molecule ( p
p
g
N
E
N
= + + +
& r
(1)
where is
1
2RT
, s is the speed ratio *u
e
, is the angle between u
e
and x, and is the angle between the nozzle
exit plane and x
16
. Similarly, other flow properties such as velocity and temperature can be obtained from various
moments of solution. Thus, the mass, momentum, and energy fluxes across a DSMC or CFD domain boundary can
be used for a FM calculation of the farfield plume. A FM code that calculates these properties on a Cartesian grid
using the projection of an inflow in the DAC surface format (which is described by properties on a triangulated
surface) has been used in this study.
III. Results
As a demonstration of the methodology described in the previous section, a generic missile configuration at an
altitude where the freestream is sufficiently rarefied is chosen. Thus, the direct simulation Monte Carlo technique is
appropriate for plume and freestream interactions, while the plume near the missile is captured using the continuum
method. A zero angle of attack case is simulated, so that axisymmetric calculations can be performed. A more
detailed discussion of axisymmetric results has been reported previously 5 Additional simulations were performed
for a moderate angle of attack in order to demonstrate the application to a 3D problem. For these simulations, the
inner PDAC solutions were then coupled to a free-molecular code and to an outer PDAC calculation.
The axisymmetric results are shown in Figure 7a and Figure 7b. The upper parts of both figures include the
CRAFT CFD
solution inside the interface boundary (highlighted in black) along with the PDAC solution that uses
the values on the interface boundary as input. Results for both translational temperature and number density show a
continuity of solution across the boundary. Discrepancies in temperature reflect the attention given to modeling the
shock layer and its interaction with the plume in the DSMC calculation. It was found that at high altitudes that the
plume core calculated by CFD is insignificantly influenced by the freestream. The number density contours in
Figure 7b are in good agreement.
a) Translational temperature: upper PDAC /lower
CRAFT CFD
b) Number density: upper PDAC /lower CRAFT
CFD
Figure 7. Temperature And Number Density Contours Comparing CRAFT CFD
and PDAC results for the 3D case are shown in Figure 8a and Figure 8b. The
interface boundary is included, and the CFD solution is enclosed within it. Again the continuity of solution across
the boundary is maintained. A slight decrease in density across this boundary near the missiles nozzle can be seen.
The one-way coupling assumption is less accurate when the flow is mostly tangential to the boundary, which is the
case in this region. However, this difference becomes negligible for farfield values.
a) Translational temperature b) Number density
Figure 8. Temperature And Number Density Contours Comparing CRAFT CFD
continuum flow solver and the DAC DSMC flow solver has been
demonstrated for a 3D application. The development of the Automatic Efficient Generalized Interface Surface
Toolkit (AEGIS Toolkit) has enabled 3D and/or complex interface surfaces to be created and appropriate properties
output for the surface with minimal user intervention. The coupling methodology has been applied to a high altitude
missile configuration. To capture the farfield plume structure, these solutions are then coupled to an additional
DSMC simulation or a free-molecular calculation using scripts to automate the process. Thus, a method for
modeling the nearfield to farfield plume expansion using the best appropriate model in each region has been
developed. Extensions for particulates and unsteady flow phenomena are also available which can be important
when simulating high altitude plume flowfields.
Acknowledgments
Portions of this work were funded by ARDECOM.
References
1
Papp, J. L., Yellin, K. A., and Dash, S. M., Plume and Divert Jet Effects on Missile Flowfields at Higher
Altitudes, AIAA 2004-1110, 42nd Aerospace Sciences Meeting and Exhibit, Reno, NV, Jan. 2004.
2
York, B.J., Kenzakowski, D.C., Lee, R.A., and Dash, S.M., "Upgrades to CRAFT Code for Simulation of
Ducted Launcher/Plume Interactions," 1997 JANNAF EPTS & SPIRITS Users Group Joint Meeting, Lockheed
Martin Corp., Sunnyvale, CA, April 1997.
3
York, B.J., Lee, R.A., Sinha, N., and Dash, S.M., "Progress in the Simulation of Particulate Interactions in Solid
Propellant Rocket Exhausts, AIAA-2001-3590, 37th AIAA/ASME/SAE/ASEE Joint Propulsion Conference, Salt
Lake City, UT, July 2001.
4
LeBeau, G.J. and Lumpkin III, F.E., Application Highlights of the DSMC Analysis Code (DAC) Software for
Simulating Rarefied Flows, Computer Methods in Applied Mechanics and Engineering, 191 (6-7), pp. 595-609,
2001.
American Institute of Aeronautics and Astronautics
12
AIAA 2007-5704
American Institute of Aeronautics and Astronautics
13
5
Papp, J., Wilmoth, R., Chartrand, C., and Dash, S., Simulation of High-Altitude Plume Flow Fields Using a
Hybrid Continuum CFD/DSMC Approach, 42nd AIAA/ASME/SAE/ASEE Joint Propulsion Conference, AIAA-
2006-4412, July 2006.
6
Millikan, R. C., and White, D. R., Systematics of Vibrational Relaxation, J. of Chemical Phys., Vol. 30, No
12, Dec. 1963.
7
Bird, G.A., Molecular Gas dynamics and the Direct Simulation of Gas Flows, Oxford Science Publications,
Clarendon Press, Oxford, 1994.
8
Wilmoth, R.G., Carlson, A.B., LeBeau, G.J., DSMC Grid Methodologies For Computing Low-Density,
Hypersonic Flows About Reusable Launch Vehicles, AIAA-96-1812, 31
st
Thermophysics Conference, New
Orleans, LA, June 1996.
9
Boyles, K.A., et al., The Use of Virtual Sub-Cells In DSMC Analysis of Orbiter Aerodynamics at High
Altitudes Upon Reentry, AIAA-2003-1030, 41st AIAA Aerospace Sciences Meeting and Exhibit, Reno, NV, Jan.
2003.
10
Wilmoth, R.G., and Papp, J.L., A Simple Fully-Coupled Two-Phase Flow Model for DSMC Plume
Simulations, JANNAF 29
th
EPTS & 11
th
SPIRITS Users Group Joint Meeting, Littleton, CO, June 2006.
11
Burt, J. M., and Boyd, I. D., Development of a Two-Way Coupled Model for Two Phase Rarefied Flows,
AIAA 2004-1351, 42nd Aerospace Sciences Meeting and Exhibit, Reno, NV, Jan. 2004.
12
Wilmoth, R.G., Yellin, K.A., Papp, J.L., and Dash, S.M, Continuum-DSMC Coupling Issues for Steady and
Unsteady Two-Phase Plumes, JANNAF 28
th
EPTS & 10th SPIRITS Users Group Joint Meeting, San Diego, CA,
Nov. 2004.
13
Miller, J. V., On GDMs: Geometrically Deformed Models for Extraction of Closed Shapes from Volume
Data, Masters Thesis, Rensselaer Polytechnic Institute, 1990.
14
GNU Triangulated Surface Library. http://gts.sourceforge.net/
15
VanGilder, D.B., Wilmoth, R.G., and Papp, J.L., A Hybrid Continuum/DSMC Methodology with Application
to Aerocapture Systems, 54
th
JANNAF Propulsion Meeting, Denver, CO, May 2007.
16
Woronowicz, M.S. and Rault, D.F.G., On Plume Flowfield Analysis and Simulation Techniques, AIAA-94-
2048, AIAA Joint Thermophysics and Heat Transfer Conference, Colorado Springs, CO, June 1994.