Académique Documents
Professionnel Documents
Culture Documents
Numerical Analysis of the Turbulent Flow and Alumina Particle Trajectories Solid Rocket Motors
Alessandro Ciucci*, Gianluca laccarino CIRA, Italian Aerospace Research Center 81043 Capua (CE), Italy
Abstract numerical analysis two-phase internal flow a solid rocket motor has been carried out employing an Eulerian-Lagrangian approach. The full NavierStokes equations solved numerically phase. The particulate phase is simulated through a Lagrangian deterministic model. Two computational multiblock grids, with different block arrangements in the aft-dome region have been employed. Also, different turbulence models (algebraic, standard Reynolds k-e and modified k-e) have been adopted to investigate their influence flow patterns. Calculations have been performed realistic segmented solid rocket motor. Results show a rather strong effect of the grid on the accuracy of the flow solution. Two-equation turbulence models provide much better results than algebraic models; also, no appreciable difference is observed in the solutions obtained using different low Reynolds formulations adopted to account for wall injection for the motor geometry considered. Alumina droplet trajectories have been computed different diameters; noticeable difference exists only between particle paths calculated from laminar turbulent flow fields. ofthe in are forthegas low onthe ona for a and
in
VT friction velocity y+ " wall coordinate e = turbulence dissipation rate li viscosity p = density Subscripts g - gas inj = injection p particle t turbulent Introduction Slag accumulation molten aluminum oxide inside solid rocket motor combustion chambers has been observed both spin-stabilized space vehicles launcher boosters of large length-to-diameter ratio with segmented grain submerged nozzle. Liquid droplets are produced by the combustion of the aluminized propellant usually employed improve performance and suppress high frequency combustion instabilities. Some these droplets remain motor during its operation and collect in the aft-dome region of the booster. The alumina slag deposition at motor aft-end results in motor performance loss, damage of thermal protection due to overheating, and possible sloshing and ejection of liquid agglomerates through the nozzle, which may cause pressure disturbances thrust imbalance. Accurate prediction of the booster turbulent, two-phase flow field, especially in the motor aftdome, is of great importance to the correct estimation of slag accumulation rate and, ultimately, motor performance. An in-depth analysis of the alumina deposition process is hampered by the complex flow features, involving complex combustion chamber geometry, turbulent flows with mass injection, regions of high recirculation, two-phase flow phenomena; in of and of and inthe to andi
Nomenclature CD = droplet drag coefficient Cf = skin friction coefficient droplet diameter /2,/n damping functions g = acceleration of gravity k = kinetic energy of turbulence Re = Reynolds number / = time T = temperature v = velocity v+ = dimensionless injection velocity
Aerospace Engineer, Propulsion Unit, Head # Aerospace Engineer, Computational Methods Unit Copyright 1997 by CIRA. Published by the American Institute of Aeronautics and Astronautics, Inc. with permission
D=
=gas
been adopted9. Both k-<10' " and k-e12 turbulence models have been employed; more recently, a modified form of the low Reynolds k-e model has been proposed for injection driven flows13' 14. With regard liquid phase, both Eulerian model8 and, more frequently, Lagrangian approach, either deterministic6' " or stochastic10'13, have been adopted. Some models that incorporate aluminum combustion and particle break-up effects have also been proposed10'15. Most of the simulations have regarded only the motor aft-end region8' 13 in order to save computer time, only computations have solved the full motor domain10' ". Despite these numerous efforts, many uncertainties remain in the flow field computation slag mass prediction because limitations current physical models available droplet data. A numerical analysis booster internal flow has been carried out with the objective to improve the current understanding and modeling capabilities of the complex flow characteristics encountered in many solid motor chambers. particular, simulations have been performed with computational grids different turbulence models. The main focus of the present analysis was not the estimation of the amount of slag deposited, rather investigation effects computational grid turbulence model on the flow patterns. The configuration considered this analysis realistic motor geometry towards combustion time, typical of large segmented solid rocket boosters. The gas phase has been treated with an Eulerian approach solving the full, compressible Navier-Stokes equations with an associated turbulence model. A Lagrangian model has been adopted for the discrete phase simulate internal two-phase flow. Greater details on the numerical model are given in the following.
tohe
a
furthermore, the uncertainties in particle density and particle size distribution represent major problem for a reliable estimation of the deposited slag. Several works concerning the numerical investigation of the internal flow field in a solid rocket motor have been presented in the past years1' 2' 3. More specifically, the problem of alumina slag deposition during motor operation been addressed by many researchers4. many cases flow been assumed as potential5"7 or inviscid rotational8, in other cases compressible viscous flow models have a
In the
approach has been adopted as in Cesco et al. : the continuous phase computations have regarded "equivalent gas", which density includes alumina mass fraction, then particle trajectories determined using flow field derived from computed equivalent field. in the the thegas
Gas flow field The computational tool used (Zonal Euler/Navier-Stokes) Simulation System which four main software packages: a- ZENDoMo (ZEN Domain Modeler): CADlike graphical interface definition computational domain; ! ZENGrid (ZEN Grid generator): a multiblock structured grid generator; ZENFlow (ZEN Row solver): zonal Euler/Navier-Stokes solver with algebraic and two-equation turbulence models; > A set of flow visualization tools including a particle tracer package.
has
has
gas
istheZEN
the
forthe
an
ofthe
The code based multiblock16, structured grid, finite volume approach with Jameson-like adaptive dissipation model using a TVD switch; residual averaging, multigrid, local time stepping and semi-implicit treatment turbulence source terms available features code17' ls. Both vectorized parallel version code is ona
are ofthe
and
afew
ofthe
inthe
andithe
and
ofthe
ofthe
and
ofthe
available. The code has been fully validated and applied in a variety of applications19'20. The present model is based on the Reynolds Averaged Navier-Stokes equations for a non reacting thermodynamic equilibrium. algebraic Baldwin-Lomax turbulence model21 lowReynolds two-equation model Myong Kasagi22 (MK) available code. Furthermore, modified two-equation, k-e, turbulence model has been added to account for the wall turbulent injection; following Sabnis al.23 standard damping functions used model have been suitably modified to account for the mass injection effects at the wall. The turbulence intensity of the incoming flow is accounted for by simply assigning a prescribed value of turbulent kinetic energy porous surface. More details given in the next section.
The
two
In
and
gasin
are
k-e
of
andthe
but
an
ofthe
are
ofthe
andofthe
inthe
et
and
in
thendof
isa
athe
inthek-
to
the
The
for
Numerical Model two-phase flow been simulated with combined Eulerian-Lagrangian approach. The full Navier-Stokes equations have been solved numerically to compute the continuous phase flow field; a discrete particle model based on the calculation alumina droplet trajectories been used discrete phase analysis. quasi-coupled
last years from both experimental numerical scientific community. From a physical point view necessary distinguish between weak strong injection defining nondimensional blowing ratio, v+, as the injection velocity vlnj over the friction velocity VT. When v+ is up to 0(1/10) the two-layer structure of the turbulent the
of its
The
has
and
by
to
of
forthe
has
andthe
inthe
to
are
the
has
are
an
injection case has been deeply analyzed; Cebeci29 proposed modification Driest viscous damping with accounting reduction of the laminar sublayer. The same modification has been introduced by Sabnis et al.23 into a Low Reynolds k-e model. More recently Piomelli et al.30
a oftheVan theaimof forthe
presented some results that proved existence of the two-layer formulation and the distortion near-wall turbulent structures. The strong injection case need to be more carefully analyzed because, best authors knowledge, theoretical results data available now. viscous damping terms have to be reformulated in order to apply the MK Low Reynolds model simulation flow over an injecting wall. weak injections easy introduce a modification following the work by Cebeci fashion similar Sabnis al.; proposed functions are: DNSandLE the
boundary layer retained existence universal law of the wall (even if modified) has been proved Stevenson28; particular, laminar sublayer thinner than corresponding solid surface. In this case, the turbulence intensity of the incoming flow can be neglected. On the other hand, when v* 0(1) turbulent fluctuations near the injecting wall must taken into account, leading to a different structure of the boundary layer. From modeling point view, weak is andthe by in is theon the toa ofa
is the a be of the
This new function has been reported in Fig. 1 together with the modified /2 function for weak injection and the original MK damping function. Note that the /^ has not been modified because a damping is provided even high v~. /2 function correct behavior wall maximum damping effect is reached region (the viscous sublayer) that displaced from the wall and thinner than the corresponding over a solid wall. It necessary outline that /a function and, eventually the new function /a, must be tuned and correlated experimental data
at
Thenw
athe
ina
andthe
haste
is
to
toDNSr
thenw
available today.
Particle trajectory The booster two-phase flow is simulated by following the motion discrete particles inside combustion chamber. model adopted here, particles assumed have spherical shape, non-rotating and no collisions between particles and breakups occur. Furthermore, distributed combustion considered, alumina droplets assumed injected from propellant surface. Also, mass and heat transfer occurs between the two-phases. For the sake of simplicity, and being the main emphasis of
of Inthe the the to a are are
ofthe
tohe
ofthe
not
no
orDNS
are
and
are
upto
The
the
k-e
tohe
ofthe
this work on the gas flow field characteristics, a deterministic approach has been adopted, that is the turbulent dispersion on particle trajectories is not accounted for. With these assumptions, the particle momentum equation can be integrated to determine the droplet trajectory:
For
its
andi
to
et
the
to
no
tobe
dt
worth noting that turbulent fluctuations zero wall asymptotic behavior turbulent quantities same that obtained solid surfaces32 and, therefore, asfrthe are athe ofthe the areth
for
"
Itis
behavior fa /2 y->0 modified (/u=O(l/y) and f2=O(y)). For strong injections the of and for arenot
/p. function is only slightly affected, its behavior being dominated by the hyperbolic tangent, while the proposed function /2 longer effective returns isno
2
where pg, pp, vg, vp are the gas and particle densities and velocities, respectively, g is acceleration of gravity, droplet diameter CD drag
as
the
Disthe
and
isthe
Rep<1000
always 1) damping provided. However, low Reynolds effects are not negligible; in fact, in our computations, the turbulent Reynolds number of wall can be estimated as 0(10) and, therefore, the high Reynolds assumption applicable31. Furthermore, the study of the asymptotic behavior of the turbulent quantities shows that/u=0(l) /2=0(1). order Reynolds formulation ando is the isnot
and
(it
Cn =0.438 isthe
Re^MOOO by:
some new damping functions must be designed; in particular, a new /2 function is here proposed as:
Rep =
velocity directly obtained from computed flow field, whereas density derived from computed value after adjustment the alumina mass fraction. droplet bulk density is
thegas
In
touseaLw
k-e
Thegas
the
the
is
The
is
for
the
isnot
is
Results and Discussion numerical model been applied simulate typical solid rocket motor internal flow, focusing on the influence of the computational grid on the
figure generated ONERA using blocks about 18000 cells; this grid will referred grid because solution adopted mesh generation aft-end region. grid shown the upper part of Fig. 3 was developed by CIRA using the same block decomposition as the ONERA grid except motor aft-dome where C-type decomposition was adopted, yielding a total number of blocks 21000 cells. This grid generated after obtaining first results with grid which showed reattachment point located region of highly skewed cells. Hence, in the attempt to was by ten be and ofthe inthe The forthe forthe 13 and a
The
has
andTiK.
to
the
was
in
theH-
toasH-
on
particle trajectories.
Case description solid rocket motor geometry selected reference configuration in the present investigation is shown in Fig. 2. This is representative of a large segmented solid rocket motor; should noticed that the entire motor domain has been considered. It is assumed that the burn out of the first segment has already occurred and an axisymmetric configuration can be considered, with mass addition only from the second third segments. This motor geometry presents main flow features encountered an actual solid rocket motor, including both entire combustion chamber and the submerged nozzle, and ablated inhibitor rings first second intersegments; particular, static deformation thermal protection (second inhibitor) has been taken into account25, alumina pool motor aft-end has been included10 (as indicated by the vertical straight line closing aft-end cavity Fig.
asthe
improve quality solution, C-type grid was chosen, which indeed led to more regular, bodyconforming cells. Furthermore, a stronger clustering of grid nodes at the walls in the aft-dome and nozzle regions employed satisfy requirements the turbulence models. On the other hand, in order to keep the number of grid cells at an acceptable level, in the other regions of the motor the grid was constructed in a fashion similar to the H-grid, except for increased clustering inhibitor walls. Both laminar turbulent computations were carried out with the two grids. The main results are discussed in the following. the ofthe
The
was
to
the
the
it
be
an
alofthe
and
in
the accurate determination of the gas flow field and an assessment of the effects of such important elements of the numerical simulation as the computational grid and the turbulence model adopted on the solution.
Computational grid effects
Simulations conducted in the preliminary phase of this study27 has indicated a certain influence of the computational grid solution, especially aft-end region of the motor and within the nozzle where complex flow patterns high gradients flow variables occur. Therefore, decided perform detailed study grid effects flow solution; this end, different multiblock grids were employed, which differed mainly in the topology adopted in the motor aft-dome. The different block decomposition together with some details computational grids shown Fig. grid illustrated lower part this
Laminar flow The convergence history laminar cases are shown in Fig. 4. A one-dimensional isentropic solution adopted initial flow field. quite different convergence is achieved with the two grids; the C-grid yields smooth convergence, where residual oscillations are progressively damped; on the contrary, stable limit cycle appears after about 45000 iterations with H-grid, implying periodic oscillatory behavior flow. attempt reach a steady solution, first the CFL number was decreased, with appreciable beneficial effects, then the artificial dissipation of the numerical scheme was increased, again with no improvements. Thus, a steady state attained flow solution 50000 iterations (just before increasing number) chosen comparison. Mean flow oscillations were observed behind second inhibitor with the H-grid; on the contrary, a steady solution obtained with C-grid. flow structure motor aft-end revealed greater detail by the streamlines depicted in Fig. 5. These streamlines clearly reveal differences between C-grid H-grid solutions; reattachment point is located at the same position along the back face of the nozzle, but the large recirculation has a somewhat different structure, with the core of the vortex having approximately the same axial position smaller radial position (closer the nozzle back face) in the C-grid case. Also, a secondary recirculation lower right corner obtained with the H-grid, but not with the C-grid; this
was forthe forthe
a
the
at
in
the
and
ofthe
the
and
athe
andthe
at
of
a
the
in
2).
ofthe
the
Inthe
The
ofthe
was
wasnot
no
andthe
at
to
was
Inay
for
theCFL
the
inthe
was
the
inthe
is
onthe
inthe
and
ofthe
the
and
itwas
ofthe
to
two
onthe
to
buta
the
the
in
The
athe
ofthe
are
3.The
inthe
of
in
is
to
the
is probably different node wall clustering between grids, being stronger C-grid. A comparison between u-velocity profiles obtained with H-grid with C-grid three axial locations in the third segment of the motor is reported Fig. agreement good first cross section behind inhibitor ring, significant discrepancy exists second cross section toward third segment; injecting flow is captured well with both grids but a thewo duetoh inthe
Turbulence model effects In order assess influence turbulence model on the flow solution, the numerical results obtained for laminar computations and turbulent computations using the Baldwin-Lomax turbulence model, the standard low Reynolds k-e model and the
to
the
the
and
the
the
in
6.The
is
athe
at
velocity deficiency at the motor axis is attained with the H-grid, reflecting unsteady, oscillatory behavior which gives rise to the residual history of Fig. Large differences seem exist also within motor aft-dome, being the velocity magnitude higher with H-grid; should noticed, however, that the velocity recirculation region case
the
very low.
Turbulent flow The evaluation of the effects of the computational grid been completed comparing the flow solutions obtained with H-grid with C-grid under assumption turbulent flow, employing two-equation, Reynolds turbulence model standard form. turbulent computations were started from the corresponding converged laminar solutions (solution 50000 iterations for the H-grid) and a full convergence was achieved both cases. Convergence problems with and were initially experienced with H-grid wall of divergent part of the nozzle, probably due to the poor grid resolution this region. order overcome this difficulty, a slip boundary condition was adopted for the nozzle divergent wall only; this does not affect the solution upstream within the motor combustion chamber. Overall, flow fields appear very similar. However, some differences observed aftend region of the motor. The streamlines for the Cgrid and H-grid turbulent solutions are shown in Fig. reattachment point moved closer nozzle nose in the H-grid solution, and the main recirculation different shape with different position center cases; secondary recirculation is observed in either solution. To locate the reattachment point with greater accuracy, the distributions of the skin friction coefficient along the nozzle back face for both the Hgrid and the C-grid are illustrated in Fig. 8; the corresponding C/ laminar distributions are also reported for comparison. The turbulent H-grid and Cgrid distributions very different, with former having much higher C/values latter being very close laminar distributions. This expected, other than different node clustering the wall (higher in the C-grid) no explanation has been found.
modified form model were compared. application of the algebraic turbulence model in this case is questionable due to the large recirculations regions and the complex geometry in the motor aftdome, where length scale cannot uniquely defined; criterion adopted here take distance from nearest wall grid node. All computations were performed using same, C-type grid. This grid is certainly well suited for laminar conditions; at the same time, it is sufficiently clustered normal direction) injection walls, being the first inner node at y+~10, but probably suitable capture turbulent near wall effects slip surfaces. However, these effects were beyond the scope of the present activity. A first comparison made among laminar results, solution obtained with Baldwin-Lomax model and the solution computed using the standard Reynolds model. Successively, flow fields determined with different forms turbulence model (standard low Reynolds, modified Reynolds weak injection, modified Reynolds strong injection) compared discussed. The details of the aft-dome flow field computed under laminar, Baldwin-Lomax hypothesis revealed streamlines shown Figs. large differences flow structure appear quite evident: the laminar and k-e results show a similar shape of the recirculation, but with a different position of the reacttachment point and of the recirculation center. The Baldwin-Lomax solution is significantly different, presenting a fairly complex flow structure with additional, secondary recirculations, most likely due to the inaccuracy of an algebraic turbulence model. u-velocity profiles three axial locations third segment shown Fig. agreement quite satisfactory cross section 1, except for a small discrepancy in the region behind second inhibitor ring. differences more pronounced cross section laminar turbulent k-e profiles agree well in the injection region motor centerline, exhibit different behavior within shear layer; BaldwinLomax profile looks similar fully developed turbulent profile, with clear evidence shear layer, incorrect distribution approaching wall and a lower value axial velocity at motor axis. The velocity profiles in the recirculation region are very similar in shape despite the different flow structures observed in Figs. 5,7 and 9; the velocity magnitude is highest for laminar flow and lowest for the Baldwinoftheka
the
buta
ofthe
thendof
the
athe
the
wasto
the
4.
to
the
atny
the
be
The
the
it
inthe
be
isnay
(inthe
not
to
has
by
atno
is
the
the
and
the
the
the
low
of
k-e
low
k-e
the
ints
The
the
at
low
for
for
are
in
the
athe
the
of
In
to
are
bythe
in
andk-e
and9.The
thewo
are
inthe
inthe
7.The
has
tohe
two
The
5,7
and
low
ofthek-
the
atl
inthe
are
has
10.The
is
ofits
inthewo
no
the
The
at
at
2;the
andthe
but
the
no
toa
are
the
an
ofthe
the
tohe
andthe
and
wasnot
at
the
and
are
in
at
the
Lomax solution; case, their absolute values very small. contour turbulent kinetic energy computed using model) reported Fig. 11; highest values attained the wake second inhibitor that impinges aft-end region. visualization entire flow field is provided using a LIC (Line Integral Convolution) technique Fig. 12; details recirculation the aft-dome region clearly evidenced Fig. sensitivity flow solution nearwall modeling turbulence analyzed comparing inay
The mapofthe
value corresponding turbulent intensity (with respect to the injection velocity) of 10%. The result of this investigation shows that no significant differences are obtained among the computations in terms mean flow quantities; this shown and v-velocity profiles third segment reported in Fig. 14. The turbulent kinetic energy profiles at the same location, illustrated in Fig. 15, ofkande toa
of is bytheuathendof
show small alterations due to the different damping functions adopted: highest values close to the
computed by means of the standard MK k-e model are reported; only limited number injection points located towards the end of the grain has been considered. Close to the propellant surface, larger particles less dragged mean flow than lighter particles, but they tend to remain trapped in the aft-dome region; in particular, for a diameter of 20 ujn all the particles exit through the nozzle, while for 140 urn all of them impinge on the nozzle back face, as clearly shown in Fig. 18. The comparison of the particle trajectories different flow fields discussed above is reported in Fig. 19 for a diameter of 20 u,m; the trajectories corresponding to the turbulent flow fields are practically overlapped, while those corresponding laminar solution differ only slightly. These discrepancies are clearly evidenced in Fig. where particle trajectories three diameters considered are computed using the laminar a are 20, tohe
forthe
are
(as
theMKk-
in
the
are
is
ofthe
inthe
in
in
the
ofthe
ofthe
is
in
in
The
ofthe
of
are
is
tohe
by
13.
bythe
andthe
propellant surface are obtained using the model for strong injection. must pointed that inhibitor wake dominates flow field and, particular, aft-dome region; peak reached within shear layer attains values much higher than level klnj assumed wall. further evidence small sensitivity solution turbulence model, contour lines turbulent kinetic energy reported Fig. 16. conclusion, differences in the mean flow and turbulent quantities induced by the different modeling of the turbulent structures near the wall are not very significant in this case. Further analysis is required using a simpler geometry before any definite conclusion can be drawn on role turbulence modeling injecting walls.
It be out the the in
Conclusions A numerical investigation turbulent flow typical solid rocket motor been carried this work. The continuous gas phase has been simulated solving the full Navier-Stokes equations. The particulate phase is simulated using a Lagrangian deterministic model. First, grid sensitivity analysis been performed using two computational multiblock grids, with different block arrangements in the aft-dome region. Laminar turbulent computations have been carried with both grids. Considerable differences in the flow solutions have been attained. In a second phase, different turbulence models (algebraic and two-equation) have been adopted to investigate their influence flow patterns.
ofthe
forthe
of
has
the
thek
the
the
of
and
athe
Asa
is
ofthe
ofthe
are
in
In
ofthe
tohe
out
and
onthe
has
outin
andew
k-e
the
ofthe
for
The lack of accurate particle data represents one of the major limitations for realistic alumina droplet trajectory simulation. As far as the particle size distribution is concerned, the work of Traineau et al.26 has been considered here. According to their results the particles are distributed following a bi-modal law, with a first peak at 1 Jim and a second peak around 70 Urn; therefore, three different particle diameters have been considered: 140 (im. Furthermore, assumed that droplets injected perpendicular propellant surface speed m/sec. alumina particle trajectories predicted using computed flow field employing Lagrangian deterministic approach. Fig. trajectories obtained using the turbulent flow field
20,8and it
account wall injection yield significant changes solution motor geometry considered. This is most likely due to the fact that the for
inthe
flow field dominated near wall effects. These models should applied more simpler isnot be by
configurations and a comparison with DNS or experimental data should be performed before any conclusion on the validity of these formulations may drawn. Finally, alumina droplet trajectories have been
forthe
dont
is
the
are
be
to
tohe
athe
of1
The
the
are
In
17,
for
In
ina
Acknowledgements The authors would like to acknowledge the support of ONERA which provided computational grid Y. Fabignon for his useful input. Many thanks are due to M. Amato for his precious contribution in the development of the computer code, to P. Leoncini for his help with particle tracker and to B. Sikorsky for contribution generation LIC images.
Rocket Motor Internal Flows", Propulsion and Power, Vol. 1989 13.R. Madabhushi, Sabnis, Jong, Gibeling, "Calculation Two-Phase AftDome Flowfield in Solid Rocket Motors", J. of Propulsion Power, Vol. 1991 14.R. H. Whitesides, R. A. Dill, and D. C. Purinton, Sambamurthi, "Design Subscale
K. J.S
andMr.
5,No.6
F.Jde
H.J
and
his
inthe
ofthe
J.K
References
1. Salita, "Vorstrem: Simple Versatile Viscous Solver Chamber Flow Solid Rocket Motor", AIAA Paper AIAA-94-2779, 30th Joint Propulsion Conference, Indianapolis, IN, June 1994 Johnston, Computational Fluid Dynamics Analysis of the Internal Flow in a Titan SRMU", AIAA Paper AIAA-90-2079, 26th Joint Propulsion Conference, Orlando, FL, July 1990 3. Ciucci, R. M. Jenkins, W. A. Foster Jr., "Analysis a and
for ina
Propellant Slag Evaluation Motor Using TwoPhase Fluid Dynamic Analysis", AIAA Paper 2780, 32nd Joint Propulsion Conference, Lake Buena Vista, FL, July 1996 15.Liaw, and Chen, "Numerical Investigation of Slag Behaviour with Com/Breakup/VOF models solid Rocket Motors", AIAA Paper 95-2726, 31st Joint Propulsion Conference, San Diego, CA, July 1995
2D
ofa
7,No.2
ofthe
and
J.of
2.
"A
of Ignition and Flame Spreading in the Space Shuttle Head-End Star Grain", AIAA Paper 923272, 4. 28th Joint and Propulsion inthe Conference,
17.
of
J.
in
21,No.
and
fora
"A
for
J.of
G.
"Onthe
6-9
7.
F.And
8.
J.W
S.
P.T
"A
Approximation
Algebraic
and
of
Model
Separated Turbulent Flows", AIAA Paper 78-257, 16th Aerospace Science Meeting, Huntsville, AL, January 1978 Myong, H., Kasagi, N., Approach
and "ANew
in
2.
23. Sabnis, Madabhushi, Gibeling, McDonald, "On the Use of k-e Turbulence Model J. R. H.
for Computation Solid Rocket Internal Hows", AIAA Paper 89-2558, 25th Joint Propulsion
July 1996
11.L. Jacques, T. Pevergne , V. Vatel, "Prevision du
Propulsifs dans Systemes Transport Spatial, Bordeaux, France, September 1995 les
12.J. Sabnis, Gibeling, McDonald, "Navier-Stokes Analysis Solid Propellant S. H.J andH.
5",
of
de
of
andH.
to
for
for
96-
Ecoulements Propulsifs dans les Systemes de Transport Spatial, Bordeaux, France, September
1995
Traineau, Kuentzmann, Prevost, Tarrin, " Particle Size Distribution Measurements in a Subscale Motor for the Ariane 5 Solid Rocket Booster", 28th Joint Propulsion Conference, Nashville, TN July 1992 Ciucci. laccarino, Amato, "Development of Numerical Tool Investigation Turbulent Flow Field Solid Rocket Motor Application to the Ariane5-95 sec Test Case", CIRA Technical Report CIRA-TR-96-149, December 1996 28. T. Stevenson, "A Law of the Wall for Turbulent Boundary Layers with Suction or Injection", The College of Aeronautics, Cranfield, Rept. Aero. No. 166,1963. P. M. andP. G. a M. forthe ina ofthe -
29. Cebeci, "Behavior Turbulent Flow near Porous Wall with Pressure Gradient", AIAA J., Vol.8, 12,1976, 2152-2156. 30. Piomelli, Moin, Ferziger, "Large Eddy Simulation of the Flow in a Transpired Channel", J. Thermophysics, Vol. 5, No. 1, 1991, pp. 124T. of No. U. P.
128.
26.JC
31.F. Nicoud, B. Chaouat, "Turbulence au Voisinage d'une Paroi Debitante", Journee R&T CNES, 2628 June 1995. 32. B. Chaouat, "Modelisation et Simulation des Ecoulements Turbulents dans les Propulseurs a Propergol Solide", PhD Thesis, 1994, Onera Tech. Note 1995-2.
27.A
J.
p.
1.4
1.2
0.8
0.6
0.4
0.2
10
05
152034
Wall Units
First Segment
Second Segment
Third Segment
C-grid H-grid
Fig. Computational grids 3-
10000
20000
30000
60000
70000
80000
90000
C-Grid
H-Grid
r/R
0.75
0.86
0.87
0 . 8 8
0.89
0 . 9 0
0.91
0 . 8 6
0 . 8 7
0.88
0.90
0.91
x/L
x/L
(j)
06
C-Grid
H-Grid
r/R
0.75
0.86
0.87
0.88
Tt/L
0.90
0.91
0.86
0.87
0.88
0.90
0.91
x/L
o H
0.877
0.865
0.86
O.B7
0.88
jj/L
0.89
0 . 9 0
0 . 9 1
0.04
0.05 <X<*
0. 10
11
American Institute of Aeronautics and Astronautics
_ _j
.J
12
American Institute Aeronautics Astronautics of and
Fig. 13 - LIC flow visualization of the solid rocket motor - Close up view of aft-dome region
Standardtops -
0.02
0 . 0 4
x-voJotily component
0 . 0 6
-0.006
-0.005
0.004
l^vekicity component
4.003
-0.002
13
American Institute Aeronautics Astronautics of and
for
- k-*modlla1oritrongln)ctlon
O.SH
tia o.eso
0.397
0.02
o.se
o*n
UL
0 . 0 3
0 0.05 . 0 4
0 . 0 6
0.84 x/L
0 . 8 6
0.875 x/L
..
0.876
0.878 X/L
15
American Institute Aeronautics Astronautics of and