Aerospace Science and Technology 24 (2013) 264–274

Contents lists available at SciVerse ScienceDirect

Aerospace Science and Technology

www.elsevier.com/locate/aescte

Thermochemical exploration of hydrogen combustion in generic scramjet

combustor
Malsur Dharavath, P. Manna, Debasis Chakraborty ∗
Computational Combustion Dynamics Division, Directorate of Computational Dynamics, Defence Research & Development Laboratory, P.O. – Kanchanbagh, Hyderabad, India

a r t i c l e i n f o a b s t r a c t

Article history: Numerical exploration of non-reacting and reacting flow field of hydrogen fueled scramjet combustor is
Received 16 December 2010 presented. Three-dimensional Navier–Stokes equations with K –ε turbulence model are solved using a
Received in revised form 16 November 2011 commercial CFD solver. A combination of Eddy Dissipation (ED) and Finite Rate Chemistry (FRC) models
Accepted 24 November 2011
is used to model combustion. The complex shock and expansion wave structures and their interactions
Available online 2 December 2011
are well captured in the simulation. The computed wall pressures match very well with the experimental
Keywords: values. Computed velocity profiles match nicely with the experimental results in the near wake region
Scramjet but overpredict the values in the far field region. It has been observed that the periodic boundary
Reacting flow condition predicts early onset of reaction but the reaction zone is much boarder and intense for the
Turbulence–chemistry interaction full combustor simulation. Full combustor simulations predict the temperature profile more accurately
and maximum deviation is of the order of 12%. Detailed H2 –air chemistry is required to get better match
of temperature field in the near wake regions.
© 2011 Elsevier Masson SAS. All rights reserved.

1. Introduction increase in momentum are very important for efficient design of

scramjet combustor. Parallel injection of hydrogen fuel has often
The development of the efficient propulsion systems capable of been contemplated in high speed operating envelope of scramjet
producing large thrust is the key for the success of the hypersonic combustor to reduce the injection losses and to take advantage of
vehicle development programs. The combustion must take place incremental momentum of the fuel stream. A good number of ex-
at supersonic speed when the flight Mach number is above 6.0 to perimental and numerical studies [6,7,12–14,18,19] were reported
maintain permissible temperature and pressure at the combustion in the literature to focus on various aspects of flow phenomena
chamber with reasonable combustion efficiency. Scramjet propul- including drag losses, mixing, combustion, intake combustor in-
sion system with hydrogen fuel is one of most promising candi- teraction, etc., in strut based scramjet combustors with hydrogen
dates for providing thrust for these vehicles. The flow field inside fuel. These studies mostly measure and compare the wall proper-
a scramjet combustor is highly complex. The mixing of reactants, ties (surface pressure and heat flux) and exit profiles for various
flame holding, and stability and completion of combustion are the flow parameters. Detailed diagnostics of flow distribution, namely,
major concerns in supersonic speed in the combustion chamber. temperature and species mass fraction across a cross section in-
Heiser and Pratt [8] provided an overview of the preliminary side the combustor is very limited. Scramjet experiments with
concept of fuel–air mixing and mixing controlled supersonic com- hydrogen fuel in a simple geometry at the Institute for Chemi-
bustion. Considerable efforts have been focused on different in- cal Propulsion of German Aerospace Center at DLR [20–22] pro-
jection schemes like cavity, strut, pylon for different geometrical vide important wall pressure distribution as well as velocity and
configurations and flow conditions in the past two decades. Se- temperature profiles at different cross sections in the combus-
lected methods that have been used to enhance the mixing process tor. In these experiments, hydrogen was injected from the base
in the scramjet engines are summarized and reported in Ref. [17]. of a wedge-shaped strut at sonic speed parallel to an airstream
The problem of slow lateral fuel transport in the air stream can of M = 2.0. Measured wall pressures and velocity and tempera-
be circumvented by injecting the fuel in the core region of the ture profiles across the cross section can act as an important data
flow by means of struts and or pylons. The oblique shocks gener- base for CFD model validation in scramjet combustor flow. RANS
ated from the struts also augment the mixing which is very much [15,23] and LES [2,5] calculations for this experimental condition
needed in high speed propulsion devices. Reduction of losses and are presented in the literature to address the effect of combus-
tion model, unstructured grid on the flow development. Wepler et
al. [23] presented RANS calculations with K –ε turbulence model
* Corresponding author. Tel.: +91 40 24583323; fax: +91 40 24340037. and Probabilistic Euler Lagrangian (PEuL) based combustion model
E-mail address: [email protected] (D. Chakraborty). while Oevermann [15] used a two equation K –ε turbulence model

1270-9638/$ – see front matter © 2011 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ast.2011.11.014
 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274 265

Fig. 1. Schematic sketch of the DLR scramjet combustion chamber.

combined with stretched laminar flamelet model for turbulent dif- Table 1
fusion flame. Coupling between turbulence and non-equilibrium Inflow conditions of the air stream and the hydrogen jet.
chemistry was achieved via a statistical description of the mixture Parameters Air Hydrogen
fraction with a presumed β -Probability Density Function (PDF). Al- Mach number 2.0 1.0
though LES study of Berglund and Fureby [2] showed a reasonable Axial velocity (m/s) 730 1200
match with the experimental values, it reported some discrepan- Static temperature (K) 340 250
cies concerning the volumetric expansion and time averaged veloc- Static pressure (bar) 1 1
Density (kg/m3 ) 1.002 0.097
ities. Genin et al. [5] use Large Eddy Simulation and Linear Eddy
O2 mass fraction 0.232 0
Model (LES-LEM) approach to simulate the Oevermann experimen- H2 O mass fraction 0.032 0
tal condition. Although, the use of LES-LEM model has improved N2 mass fraction 0.736 0
the temperature profile prediction compared to RANS based model, H2 mass fraction 0.0 1
there is significant difference between the prediction and experi-
mental results. It is clear that modeling issues in high speed tur-
adjacent region is almost symmetrical (except the two side wall-
bulent reactive flows need further investigation. The higher order
adjacent holes). Hence, one injection hole along with both side
turbulence and combustion models are computationally prohibitive
adjacent regions (up to middle of the two adjacent injection holes)
for practical engineering applications.
has been considered for the computational domain. The computa-
In this work, the experimental investigations of Waidmann
tional domain and typical grid structure are shown in Fig. 2.
et al. [20–22] have been explored numerically with K –ε turbulence
In the simulation, X -axis is taken along the length, Y -axis along
model, simple chemical kinetic scheme and simple turbulence–
the height and Z -axis along the width with the origin being placed
chemistry interaction model to find the capability of these stan-
at middle of the bottom wall. Structured grid with the typical size
dard engineering tools to predict the mixing and combustion
540 × 75 × 30 is generated for the computational domain using
characteristics in a scramjet combustor. Three-dimensional Navier–
ICEM-CFD [1]. The grids are clustered towards the upper, lower,
Stokes equations are solved using commercial CFD software. Com-
and strut wall as well as in the wake region of the strut to capture
puted thermochemical behaviors of the flow field are compared
high gradient flows in these regions. In the present study four dif-
with experimental results.
ferent types of boundary conditions are applied. Supersonic inflow
2. Experimental conditions and computational details has been imposed at the inlet of the combustor and H2 injection
hole and supersonic outflow has been considered at the outlet of
Experimental conditions of Waidmann et al. [20–22] have been the combustor. Adiabatic wall condition is imposed on top and bot-
discussed adequately in the literature. In the experiment, pre- tom wall of combustor and strut surface. Symmetry condition is
heated air (heated by combustion of hydrogen with air in a heater) imposed on two sides of the domain.
is expanded through a Laval nozzle and enters in the combustion
inlet at M a = 2.0. The combustor has a constant area section of 3. Computational methodology
0.058 m from the combustor inlet. This is followed by a diver-
gence section (one sided divergent combustor) of 0.242 m length CFX-11 [1] is an integrated software system capable of solving
with 3◦ divergent angle at upper wall, provided to compensate for diverse and complex multi-dimensional fluid flow problems. The
the expansion of the boundary layer. The width and the length of software solves 3D Reynolds Averaged Navier–Stokes (RANS) equa-
the combustor are 0.045 m and 0.3 m respectively. The heights tions in a fully implicit manner. It is a finite volume method and is
of the combustor at inlet and outlet are 0.05 m and 0.068 m re- based on a finite element approach to represent the geometry. The
spectively. A wedge-shaped strut is placed at the middle in the method retains much of the geometric flexibility of finite element
combustion chamber at 0.035 m downstream from the combustor methods as well as the important conservation properties of the
inlet. The length and half-wedge angle of the strut are 0.032 m finite volume method. It utilizes numerical upwind schemes to en-
and 6◦ respectively. Hydrogen fuel is injected parallel to the air sure global convergence of mass, momentum, energy and species.
stream (vitiated air) through 15 numbers of holes with a diameter In the present study, the discretization of the convective terms is
of 0.001 m placed 0.0028 m apart (along the width) at the mid- done by the first-order upwind difference scheme. Local time step-
dle of the strut base. Fig. 1 shows the schematic of the combustor ping has been used to obtain steady-state solutions. The turbulence
geometry with the physical dimensions. The flow conditions of the model used was K –ε model with wall functions. The combus-
incoming air stream and the hydrogen fuel are taken from Ref. [20] tion is modeled using combination of infinitely fast rate kinetics
and are given in Table 1. Combustion was initiated by pre-burning based on Eddy Dissipation (ED) and Finite Rate Chemistry (FRC)
of a small amount of O2 in a H2 tube by a spark. models. The details of the governing equations, thermodynamics,
Since the injection holes are equal in size and placed in equi- combustion models and the discretization schemes are given in
distance along the width, the flow behavior of the holes and the following subsections. To find out the accuracy and the range
266 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274

Fig. 2. (a) Computational domain, and (b) grid distribution on combustor with zoomed at strut region.

of applications, the software has been validated for various react- Turbulent kinetic energy (K ) equation:
ing and non-reacting flows pertaining to the scramjet combustor   
including transverse sonic injection in supersonic flow [10], trans- ∂ ∂ ∂ μl μt ∂ K
(ρ K ) + (ρ uk K ) = + + SK
verse H2 injection in constant area duct [11], staged H2 injection ∂t ∂ xk ∂ xk Pr σ K ∂ xk
from struts [16] and pylon injectors [9]. All these validation ex-
Rate of dissipation of turbulent kinetic energy (ε ) equation:
ercises have revealed that although, the computed pressures over-
  
predict the experimental values in the injection zone, the com- ∂ ∂ ∂ μl μt ∂ ε
putational and experimental values of the flow parameters match (ρε ) + (ρ uk ε ) = + + Sε
∂t ∂ xk ∂ xk Pr σε ∂ xk
fairly well in the divergent portion of the combustor where the
major portion of thrust is produced. Species mass fraction ( Z ):
  
∂ ∂ ∂ μl μt ∂ Z
3.1. Governing equations (ρ Z ) + (ρ uk Z ) = +
∂t ∂ xk ∂ xk Pr σc ∂ xk
The appropriate system of equations governs the turbulent flow where ρ , u i , p , H are the density, velocity components, pressure
of a compressible gas may be written as and total enthalpy respectively and μ = μl + μt is the total vis-
cosity; μl , μt being the laminar and turbulent viscosity and Pr is
Continuity equation: the Prandtl number. The source terms S k and S ε of the K and ε
equation are defined as
∂ρ ∂
+ (ρ uk ) = 0, k = 1, 2, 3 ∂ ui ∂ ui ρε2
∂t ∂ xk S K = τik − ρε and S ε = C ε1 τik − C ε2
∂ xk ∂ xk K
Momentum equation:
where turbulent shear stress is defined as
∂ ∂ ∂P ∂(τik )  
(ρ u i ) + (ρ u i uk ) + = , i , k = 1, 2, 3 ∂ ui ∂ uk
∂t ∂ xk ∂ xi ∂ xk τik = μt +
∂ xk ∂ xi
Energy equation: Laminar viscosity (μl ) is calculated from the Sutherland law as
 3/2  
∂ ∂ ∂ ∂ qk T T ref + S
(ρ H ) + (ρ uk H ) = − (u j τ jk ) + , j , k = 1, 2, 3 μl = μref
∂t ∂ xk ∂ xk ∂ xk T ref T+S
 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274 267

where T is the temperature and μref , T ref and S are known coef- bulent flows, this mixing time is dominated by the eddy properties
ficients. The turbulent viscosity μt is calculated as and, therefore, the rate is proportional to a mixing time defined by
the turbulent kinetic energy (K ) and dissipation (ε ), i.e., reaction
ρK2 rate αε / K . The mixing rate determined from the ED model is given
μt = cμ
ε as
The coefficients involved in the calculation of μt are taken as  
ε Yo Yp
R H2 ,edm = A ed ρ min Y f , , B ed
c μ = 0.09, C ε1 = 1.44, C ε2 = 1.92 k νH2 1 + νH2
σ K = 1.0, σε = 1.3, σc = 0.9 where Y f , Y o and Y p are the mass fraction of fuel, oxidant and
The heat flux qk is calculated as qk = −λ ∂∂xT , λ is the thermal con- products respectively, A ed and B ed are model constants, νH2 is sto-
k
ductivity. ichiometric coefficients of H2 reaction. In FRC model the kinetic
rate of change of any species (H2 , O2 and H2 O) in a reaction is
3.2. Thermodynamics model generally described by Arrhenius expression involving an exponen-
tial dependence on temperature and power law dependence on the
A thermally perfect gas is assumed in the present study and, concentrations of the reacting chemical species. The rate of reac-
consequently, the specific heats for all species are function of tem- tion of R H2 ,frc (in kg mol/m3 s) is given by the expression [3,4]
perature only. The specific heats are calculated using a fourth-order 
polynomial at the interval of fluid temperature 300–5000 K. In R H2 ,frc = −2 1.102 × 1019 ∗ exp(−8052/ T )c H
2 2
c − κb c H
2 O2 2O
each interval, the same form for the polynomials is used but dif-
ferent coefficients can be used. where c is the molar concentration (in g mol/cm3 ) and κb , the rate
constant of backward reaction, is obtained from the forward rate
C pi
= Ai + B i T + Ci T 2 + D i T 3 + E i T 4 constant and equilibrium constant (κb = κ f /κe ), where κ f and κe
R are forward rate constant and the equilibrium coefficient respec-
where A i , B i , C i , D i and E i are curvefit constants [15] and T is the tively, it can be written as κe = ( R T )−2 exp(−2g i / R T ).
fluid static temperature. C p i is linearly extrapolated when the fluid The rate of reaction is then determined from the minimum of
temperature T > 300 K or T < 5000 K. Then, the static enthalpy h the mixing and kinetic net rate and is expressed as
n
is calculated as h = i =1 αi h i ( T ) and the static enthalpy of each
species h i ( T ) is R H2 = min( R H2 ,edm , R H2 ,frc )
T The finite rate turbulence–chemistry interaction model through
h i ( T ) =
hof i + C p i ( T ) dT source term is difficult and need extra models. Such interactions
T0
are not considered in the present study.

where
hof i
is the standard heat of formation of species i, defined
3.4. Discretization of governing equations
as the heat evolved when one mole of substance is formed from
its elements in their respective standard states at 298.15 K and
The CFX-11 solver utilizes a finite volume approach, in which
1.0 atmosphere. The fluid temperature is calculated based on the
the conservation equations in differential form are integrated over
solution of the fluid enthalpy using a Newton’s iteration method
a control volume described around a node, to obtain an integral
for finding the roots of the polynomials. An equation of state of
equation. The pressure integral terms in the momentum integral
the following form for a multi-component is used to calculate fluid
equation and the spatial derivative terms in the integral equa-
density ρ = P /( R T / M w ), where the mixture
molecular weight is
tions are evaluated using finite element approach. An element is
obtained by the following equation, M w = ( i =1 (αi / M w ))−1 and
n
described with eight neighboring nodes. The advective term is
R is the universal gas constant.
evaluated using upwind differencing with physical advection cor-
The Gibbs free energy is required to determine the equilibrium
rection. The set of discretized equations form a set of algebraic
constants for the combined eddy dissipation and finite rate chem-
equations: A x = b, where x is the solution vector. The solver uses
istry models. It is obtained for a constant pressure process by
an iterative procedure to update an approximated xn (solution of
gi Bi Ci Di Ei x at nth time level) by solving for an approximate correction x
= A i ( T − ln T ) − T2 − T3 − T4 − T 5 + Fi − Gi T  where R = b − A xn is the residual at
R 2 6 12 20 from the equation A x = R,
where G i is an additional curvefit constant. nth time level. The equation A x = R  is solved approximately using
an approach called Incomplete Lower Upper factorization method.
3.3. Combustion modeling An algebraic multigrid method is implemented to reduce low fre-
quency errors in the solution of the algebraic equations. Maximum
The combustion model used for the simulation is a combina- residual (= ϕ nj +1 − f (ϕ nj +1 , ϕ nj )) < 10−4 is taken as convergence
tion of Eddy Dissipation (ED) model and Finite Rate Chemistry criteria.
(FRC) model to calculate effective reaction rate. The reaction rate
is calculated by using single step chemistry, and for the following 4. Results and discussion
reversible reaction has been chosen.
As mentioned earlier, the injection holes are equal in size and
2H2 + O2 ⇔ 2H2 O placed in equi-distance along the width, the combustor geometry
The combined model is computed to be the minimum of the is symmetry around the holes and adjacent region. However, this
ED rate and the FRC rate. In the ED model, the chemical reaction is geometrical symmetry is not present the wall-adjacent holes. One
fast relative to the transport processes in the flow. When reactants injection hole along with both side adjacent regions (up to mid-
mix at the molecular level, they instantaneously form products. dle of the two adjacent injection holes) has been considered for
The model assumes that the reaction rate may be related directly the computational domain. The supersonic inflow condition cor-
to the time required to mix reactants at the molecular level. In tur- responding to Mach 2.0 is imposed at the inflow boundary. Both
268 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274

Fig. 3. Mach number distribution on mid-plane of combustor width ( Z = 0.0).

Fig. 4. Density distribution on mid-plane of combustor width ( Z = 0.0).

Fig. 5. Static pressure distribution on mid-plane of combustor width ( Z = 0.0).

non-reacting (with mass addition) and reacting simulations are complex flow pattern in the wake region of the strut is seen in the
carried out. mid-plane pressure and density distribution presented in Figs. 4
and 5 respectively.
4.1. Non-reacting flow simulation Experimental shadowgraph and numerical schlieren (density
gradient) are compared in Fig. 6 to depict the position of the
In the simulation, sonic hydrogen gas is injected parallel to air shocks and their interaction and the numerical simulation exhibits
flow (M a = 2.0) from the base of the strut. Hydrogen is considered all detailed flow structures as seen in the experiment. The com-
as inert gas without any combustion. The Mach number distribu- puted static pressure distributions at bottom wall are compared
tion at the mid-plane of the combustor width ( Z = 0) is shown in in Fig. 7(a). The oblique shockwave generated from the leading
Fig. 3. edge is hitting the lower wall at X = 0.07 m from the combustor
The complex shock structures from the leading edge and the entry. Both the magnitude and location of the pressure rise are cor-
expansion waves from the base of the struts and their reflections rectly matched with the experimental result. Although, the loca-
from the walls are crisply captured in the simulation. Thin bound- tion ( X ∼ 0.136 m) of the second reflection point is predicted cor-
ary layers at both upper and lower walls are seen in the figure. The rectly, the computation underpredicts the second shock strength.
 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274 269

Fig. 6. Comparison of (a) experimental shadowgraph with (b) computational density gradient.

Fig. 7. Comparison of wall pressures: (a) experiment & CFD (lower wall), (b) upper & lower wall (CFD).

Computed static pressures at the upper and lower walls are com-
pared in Fig. 7(b). Both the pressures are following the same trend
except at the first reflection point where the lower wall exhibits
about 11% higher pressure.
The comparison of static pressure at the center line of the
combustor (line passing through the center of the H2 injector,
Y = 0.025 m) is shown in Fig. 8. Overall good match is observed
between the experiment and present computation except at X =
0.11 m (0.043 m downstream of hydrogen injection plane) where
experimental value is higher than the computed results. The pro-
files of the axial velocity (u velocity) component at four different
stream-wise locations ( X = 0.078, 0.125, 0.157 and 0.233 m) are
compared with experiment results in Fig. 9. A very good match is
observed between the computed and the experimental results. The
magnitude of the velocity defect in the strut wake region is dimin-
ishing as we proceed downstream.
Fig. 8. Comparison of static pressure at middle of combustor height (Y = 0.025 m).
The performance of the combustor in non-reacting flow with
mass addition is characterized by mixing efficiency which is de-
fined as: where ρgas is the gas density, Y H2 is the mass fraction of hydro-

 gen with Y H2 = ρH2 /ρgas , ρH2 is the density of hydrogen, A is the
A
αρgas Y H2 u d A 1, φ<1 cross-sectional area, and u is the axial velocity. φ is the local equiv-
ηm = , with α = alence ratio which is defined as φ = ( M O2 Y H2 /2M H2 Y O2 ) where
ṁH2 1/φ, φ  1
270 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274

Fig. 9. Comparison of axial velocity (u) profile at various axial locations: (a) X = 0.078 m, (b) X = 0.125 m, (c) X = 0.157 m and (d) X = 0.233 m.

Comparing the density field with the non-reacting case (Fig. 4),
we can observe that the heat release due to reaction has reduced
the density in the wake region of the strut. Also, the shock struc-
tures are weakened significantly in the downstream due to the
combustion of hydrogen fuel.
The distribution of H2 O vapor mass fraction at the mid-plane of
the combustor width ( Z = 0) presented in Fig. 12 clearly depicts
the reaction zone in the combustor. It is seen that combustion has
occurred in a narrow zone from the base of the strut and the fuel
has not spread enough in the lateral direction. Both the near wall
zones remain unaffected even after the combustion is completed
mainly because the injection is parallel and the low fuel mass flow
rate (φ = 0.0276).
The comparison of computed axial velocity (u velocity) distri-
bution with experiment at Y = 0.025 m is presented in Fig. 13.
Fig. 10. Mixing efficiency for non-reacting flow. The high speed hydrogen fuel sharply decelerated from 1200 m/s
(injection velocity of hydrogen) to about 200 m/s within short dis-
M O2 and M H2 are the molecular weights of oxygen and hydrogen tance of ∼ 0.003 m. The axial velocity remains almost constant in
respectively, and Y O2 is the mass fraction of oxygen. The distribu- the reaction intense zone (up to X = 0.11 m), afterwards, the ve-
tion of the mixing efficiency along the length of the combustor is locity gradually increases in the divergent combustor. The trend of
shown in Fig. 10. the present results matches well with the experiment, however,
computed axial velocity is higher. Comparisons of axial velocities
4.2. Reacting flow simulation at different cross sections (at x = 0.078, 0.125, and 0.207 m) pre-
sented in Fig. 14 reveal that the computed velocity profile near
Reacting flow simulation has been carried out with the reac- wake region is matching well with the experimental values, while
tion of hydrogen fuel in vitiated air. Based on mass flow of air and it overpredicts the velocity in the far wake region. A better grid
hydrogen, the stoichiometric ratio is calculated to φ = 0.0276. Sin- density may be required to resolve the differences.
gle step reaction considering H2 and O2 as reactant and H2 O as The computed temperature profiles at the same axial locations
product is considered for the simulation. The density field at the ( X = 0.078, 0.125 and 0.233 m) are compared with the experi-
mid-plane of the combustor width ( Z = 0) is shown in Fig. 11. mental result in Fig. 15. Clearly the reaction is confined at the
 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274 271

Fig. 11. Density contour distribution at mid-plane of combustor width ( Z = 0.0).

Fig. 12. H2 O mass fraction distribution at mid-plane of combustor width ( Z = 0.0).

The distribution of combustion efficiency along the length of

the combustor has been shown in Fig. 16. Combustion efficiency is
defined by the following formula:

A ρgas Y H2 u d A
ηc = 1 −
ṁH2,inj

which describes how much of the injected fuel has been consumed
since injection.
From the results it has been observed that the combustion has
been completed within an axial distance of 0.063 m from the base
of the strut (H2 mass ∼ 0.02% of total H2 ).

4.3. The effect of periodic boundary condition on the flow field

New simulations were carried out by considering the full com-

Fig. 13. Comparison of axial velocity at middle of the combustor height (Y = bustor. Taking advantage of the symmetry, only one half of the
0.025 m). combustor is simulated along with seven and half injectors. A to-
tal 420 × 65 × 78 grids consisting of 2.13 millions computational
volumes are used in the simulation. All the wall boundary layers
central zone of the combustor downstream of the strut. The ex- (top, bottom and side) were well resolved in the numerical simula-
tent of the combustion zones is about 10d (d is diameter of in- tion. Minimum y + at the wall is about 10. Since no information on
jection hole) in the middle along the height of the combustor. the boundary layer at the combustor entry is available, additional
The zone of combustion slightly moves upwards (towards the top simulation is carried out with initial boundary layer thickness of
wall) in the downstream region. A good overall trend between 5 mm and the results are compared with no initial boundary layer
the experiment and the computation is predicted. The computed case. 1/7th power law is fitted for the velocity profile at the com-
value of temperature distribution in the reaction zone in the near bustor entry. The computed water mass fraction distributions for
wake region ( X = 0.078 m) is higher compared to the experimen- these two cases are presented in Fig. 17.
tal data. This is due to instantaneous heat release due to fast Comparing the water mass fraction distribution with the pe-
chemistry assumption. A detailed H2 –air chemical kinetics is nec- riodic boundary case (Fig. 12) we could see although the peri-
essary to predict the finer details of the flow field in the near wake odic boundary case is showing reaction closer to the strut base,
region. a boarder and intense reaction zone is visible in the downstream
272 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274

Fig. 14. Comparison of axial velocity profiles: (a) X = 0.078 m, (b) X = 0.125 m, (c) X = 0.207 m.

Fig. 15. Comparison of static temperature profiles: (a) X = 0.078 m, (b) X = 0.125 m, (c) X = 0.207 m.

mental values in Fig. 19 which reveal that the profiles from full
combustor simulation are showing much better comparison (both
in magnitude and structure) with the experimental values com-
pared to the periodic boundary case. At the near strut location
( X = 0.078 m), full combustor simulation is capturing the peaks
and valleys of temperature profiles very well whereas the periodic
boundary case is showing much higher temperature in the core
of the wake region. This is due to the early onset of reaction in
the periodic boundary condition case as discussed earlier. At the
furthest location ( X = 0.233 m) the maximum deviation of tem-
perature is of the order of 12%. It may be recalled that the present
simulation is using infinitely fast rate kinetics which is likely to
yield higher computed temperature.

5. Concluding remarks

Fig. 16. Combustion efficiency for reacting flow. Mixing and combustion of hydrogen fuel injected parallely from
the struts into Mach 2 vitiated air stream in a generic scramjet
combustor are explored numerically. Both non-reacting and react-
regions for the full combustor simulation. It is observed that
ing flows are simulated. The generic scramjet combustor with hy-
higher boundary layer thickness at the combustor entry resulted in
drogen fuel injected from the base of the strut investigated at DLR
shorter reaction zone in the near field of the strut. The computed
in Germany is taken as the test case for validation. 3D Reynolds
centerline velocity distributions for the reacting case for all the
averaged Navier–Stokes equations are solved along with K –ε tur-
three cases are compared with experimental data in Fig. 18. With bulence model using a commercial CFD software. Combustion of
the full combustor simulation, velocity has reduced significantly in hydrogen gas with air was modeled with combined Eddy Dissi-
the reaction zone indicating more intense reaction which collab- pation Model (EDM)/single step Finite Rate Chemistry (FRC). The
orates well with the water mass fraction distribution. For 5 mm simulations crisply capture the complex flow structure including
initial boundary layer case, there is marginal difference in the ve- the leading shock waves generated from the leading edge of the
locity distribution in the zone 0.08 m < X < 1.1 m. fuel injection strut, expansion fan from the base of the strut and
The computed temperature profiles for different axial locations the interaction of shock reflections from the walls in non-reacting
(0.078 m, 0.125 m and 0.233 m) are compared with the experi- flow. The results of wall pressures and axial velocities match well
 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274 273

Fig. 17. H2 O mass fraction distribution at mid-plane of combustor width ( Z = 0.0): (a) 0 BL, (b) 5 mm BL.

with maximum deviation with experimental value is of the order

of 12%. A detailed H2 –air chemical kinetics is required for resolving
the difference. Higher boundary layer thickness at the combustor
entry resulted in shorter reaction zone in the near field of the
strut. The initial boundary layer does not have significant impact in
the flow development in the scramjet combustor. It has been ob-
served that while simple engineering turbulence and combustion
model is adequate for predicting the overall mixing and combus-
tion process of hydrogen combustion but a detailed description of
turbulence and chemistry is required for predicting the finer de-
tails of the flow field.

References

[1] ANSYS CFX, Release 11.0: Installation and Overview, January 2007.
[2] M. Berglund, C. Fureby, LES of supersonic combustion in a scramjet engine
model, Proceedings of the Combustion Institute 31 (2007) 2497–2504.
[3] Debasis Chakraborty, P.J. Paul, H.S. Mukunda, Evaluation of combustion models
for high speed H2 /air confined mixing layer using DNS data, Combustion and
Fig. 18. Comparison of centerline velocity. Flame 121 (2000) 195–209.
[4] J.P. Drummond, M.H. Carpenter, D.W. Riggins, in: S.N.B. Murthy, E.T. Curran
(Eds.), High Speed Propulsion System, in: Progress in Aeronautics and Astro-
with the experiment. In the reacting flow, the presence of reaction nautics, vol. 137, 1991.
has weakened the shock structure. Reaction of hydrogen fuel has [5] F. Genin, B. Chernyavsky, S. Menon, Large eddy simulation of scramjet combus-
occurred within a narrow zone (∼ 10d diameter) at the middle of tion using a subgrid mixing/combustion model, AIAA Paper 2003-7035, 2003.
the combustor. The computed velocity profile matches well in the [6] P. Gerlinger, P. Kasal, P. Stoll, D. Bruggemann, Experimental and theoretical in-
vestigation on 2D and 3D parallel hydrogen/air mixing in a supersonic flow,
near wake region, while the computation overpredicts the velocity
ISABE Paper 2001-1019, 2001.
in the far wake region. The effect of the periodic boundary condi- [7] D.D. Glawe, M. Saminy, A.S. Nejad, T.H. Cheng, Effects of nozzle geometry on
tion on the scramjet combustor is studied by carrying out the sim- parallel injection from base of an extended strut into a supersonic flow, AIAA
ulation with full geometry and comparing the results with periodic Paper 95-0522, 1995.
boundary case. It has been observed that the periodic boundary [8] W. Heiser, D. Pratt, Hypersonic Airbreathing Propulsion, Education Series, AIAA,
1994.
condition predicts early onset of reaction but the reaction zone is [9] A. Javed, D. Chakraborty, Numerical simulation of supersonic combustion of
much boarder and intense for the full combustor simulation. Full pylon injected hydrogen fuel in scramjet combustor, Journal of the Institution
combustor simulations predict the temperature profile accurately of Engineers (India) 87 (May 2006) 1–6.
274 M. Dharavath et al. / Aerospace Science and Technology 24 (2013) 264–274

Fig. 19. Comparison of static temperature profiles at different axial locations: (a) X = 0.078 m, (b) X = 0.125 m, (c) X = 0.233 m.

[10] P. Manna, D. Chakraborty, Numerical investigation of transverse sonic injection [17] J.M. Seiner, S.M. Dash, D.C. Kenzakowski, Historical survey on enhanced mixing
in a nonreacting supersonic combustor, Proc. IMechE Part G: J. Aerospace Engi- in scramjet engines, Journal of Propulsion and Power 17 (6) (2001) 1273–1286.
neering 219 (2005) 205–215. [18] S. Tomioka, K. Kobayashi, K. Kudo, A. Murakami, T. Mitani, Effects of injec-
[11] P. Manna, D. Chakraborty, Numerical simulation of transverse H2 combustion tion configuration on performance of a stage supersonic combustor, Journal of
in supersonic airstream in a constant area duct, Journal of the Institution of Propulsion and Power 19 (5) (2003) 876–884.
Engineers (India) 86 (November 2005) 47–53. [19] S. Tomioka, A. Murakami, K. Kudo, T. Mitani, Combustion tests of a staged su-
[12] G. Masuya, T. Komuro, A. Murakami, N. Shinozaki, A. Nakamura, M. Murayama, personic combustor with a strut, Journal of Propulsion and Power 17 (2) (2001)
K. Ohwaki, Ignition and combustion performance of scramjet combustor with 293–300.
fuel injection struts, Journal of Propulsion and Power 11 (2) (1995) 301–307. [20] W. Waidmann, F. Alff, M. Bohm, U. Brummund, W. Clauss, M. Oschwald, Experi-
[13] T. Mitani, N. Chinzei, T. Kanda, Reaction and mixing-controlled combustion in mental investigation of hydrogen combustion process in a Supersonic Combus-
scramjet engines, Journal of Propulsion and Power 17 (2) (2001) 308–313. tion Ramjet (SCRAMJET), in: DGLR Jahrbuch, 1994, pp. 629–638.
[14] T. Mitani, T. Kanda, T. Hiraiwa, Y. Igarashi, T. Nakahashi, Drags in scramjet en- [21] W. Waidmann, F. Alff, M. Bohm, U. Brummund, W. Clauss, M. Oschwald, Su-
gine testing – Experimental and computational fluid dynamics studies, Journal personic combustion of hydrogen/air in a scramjet combustion chamber, Space
of Propulsion and Power 15 (4) (1999) 578–583. Technology 15 (6) (1995) 421–429.
[15] Michael Oevermann, Numerical investigation of turbulent hydrogen combus- [22] W. Waidmann, U. Brummund, J. Nuding, Experiments investigation of Su-
tion in a scramjet using flamelet modeling, Aerospace Science Technology 4 personic Ramjet Combustion (Scramjet), in: 8th International Symposium on
(2000) 463–480. Transport Phenomena in Combustion, San Francisco, USA, 1995.
[16] S. Saha, D. Chakraborty, Reacting flow computation of staged supersonic com- [23] Ulrich Wepler, W. Kaschel, Numerical investigation of turbulent reacting flows
bustor with strut injection, AIAA Paper 2006-3895, 2006. in a scramjet combustor model, AIAA Paper 2002-3572, 2002.