You are on page 1of 10

15th Annual CFD Symposium, August 9-10, 2013, Bangalore

2-D Numerical simulation of Hydrogen combustion in DLR SCRAMJET


Combustion chamber
Syam Sundar Vangara1, Chandraprakash Tourani2, Vijay Kumar Kannan3

Abstract
Computational investigations of flow with and without combustion have been carried out in 2-D DLR
SCRAMJET combustor geometry using RANS equations. The combustor has hydrogen injection parallel to the
main flow from base of the wedge. The simulations have been carried out using STAR-CCM+. For turbulence
closure k-Omega SST (shear-stress transport) model has been used. The Hydrogen-Air combustion modeling is
performed using Eddy Break-up (EBU) model with 1-step global reaction chemistry. Numerical simulation
results have been compared both qualitatively and quantitatively with the published 3-D computational results
and experimental data.

Keywords
Supersonic combustion, DLR SRCAMJET, STAR-CCM+, SST K-Omega model, Hydrogen
combustion, EBU.

Nomenclature
M = Mach number
EBU = Eddy Break-up
= Density
P = Pressure
T = Temperature
Y = Mass fraction
k = Turbulent kinetic energy
= Specific dissipation rate
CFD = Computational Fluid Dynamics
SST = Shear stress transport

Introduction
Supersonic combustion leads to very short residence time in combustor, in which fuel and air have to
mix and burn completely for efficient combustion. Numerous experimental investigations were conducted to
study the efficiency of supersonic combustion. Various computational studies have also been carried out with
different injector & combustor configurations, and fuels to obtain efficient combustion. The complex flow
phenomenon involving shock-waves along with shock-shear layer interactions and shock reflections from wall,
pose a challenge for numerical codes to accurately model or capture the flow physics in such scenarios.
The present work is focused towards validating the experimental investigation of hydrogen combustion
in DLR SCRAMJET combustor using STAR-CCM+.

1
Undergraduate Student, Department of Aerospace Engineering, IIT Madras, Chennai, India
2
Application Engineer, CD-adapco, 7th Floor, Navigator Bldg, I.T.P.L., Whitefield, Bangalore, India ;
E-mail: chandraprakash.tourani@cd-adapco.com
3
Sector Head - Aerospace, Turbo-machinery, & Wind Power, CD-adapco, Bangalore, India
Physical model
The combustor geometry of the Scramjet Combustion chamber at DLR is same as given by Oevermann
[1] and is shown in Fig 1. The chamber has a one-sided divergent channel with inlet cross section of 45 50 mm
and length of 300 mm. The tip of the wedge has coordinates of X=35 mm and Y=25 mm, where lower wall is at
Y=0 mm and air inlet is at X=0 mm. Vitiated air enters the inlet at Mach 2 and hydrogen gas is injected from the
base of the wedge through 15 holes, each with 1 mm diameter. Half angle of the wedge is 6 degrees.

Figure 1. Geometry of the DLR Scramjet combustor [1].

Computational model
The 3-D CAD model with a width of 2 mm has been constructed using CAD modeler built within
STAR-CCM+. The mesh topology consisting of Polyhedral cells was created in STAR-CCM+. The 3-D meshed
domain, is converted into a 2-D domain consisting of approximately 0.2 Million cells, as shown in Fig. 2. Mesh
has been refined in the wake region behind the wedge, to resolve shock-shear layer interactions, as well as the
mixing and diffusion effects of hydrogen effectively. Near wall region is also refined to capture the shock-
boundary layer interactions. The 2-D computational model has an air inlet height of 50 mm and the fuel injection
height of 1 mm at the base of the wedge.

Figure 2. 2-D Mesh consisting of 2,00,801 cells.

Boundary conditions
The inlet conditions for vitiated air and hydrogen injection are taken from Oevermann [1] and are given
in Table 1. The stagnation temperature and pressure of vitiated air at the inlet are 612 K and 7.825 bars,
respectively. The mass flow rate of the Hydrogen inlet was adjusted to maintain the equivalence ratio of 0.0125,
as given in [2]. All outlet boundary flow variables are extrapolated from the interior of the domain, as the flow is
supersonic. Walls are considered as adiabatic with no-slip condition.
Table 1. Conditions at Air and Hydrogen inlets

Condition Air inlet Hydrogen inlet


T [K] 340 250
P [1e5 Pa] 1 1
[kg/m3] 1.002 0.097
YO2 0.232 0
YN2 0.736 0
YH2O 0.032 0
YH2 0 1
k [m2/s2] 10 2400
[s-1] 65 41666.67

Numerical Approach
The k-Omega SST model has been used for modeling turbulence. The values of turbulent kinetic energy
(k) and specific dissipation rate () are given in the Table 1. The turbulent Prandtl and Schmidt numbers are 1.2
and 0.7, respectively as given in [2]. Hydrogen-Air combustion modeling is performed using Standard EBU
model with reversible 1-step global reaction chemistry. The Standard EBU model assumes that the reaction rate
is dictated solely by the turbulent mixing time scale. The properties of Hydrogen-Air mixture have been
determined using mass-averaged mixing laws for mixture of perfect gases. The properties like viscosity, specific
heat capacities, and thermal conductivity for individual species are computed using Sutherland's law, NASA
thermodynamic polynomials, and power law, respectively. Coupled implicit solver scheme with second order
discretization is implemented in STAR-CCM+ for solving RANS equations. Initialization of the fluid domain is
done using grid-sequencing (GS) expert initialization within coupled solver in STAR-CCM+. This technique
initializes the fluid domain with invisid solution of the flow problem, which is closer in reality to the actual
viscous solution, as compared to the normal initialization as specified in initial conditions. The usage of expert
initialization helps cut down the number of iterations needed to converge the solution.

Results and Discussion


In this section, 2-D numerical simulation results for the DLR scramjet combustor have been presented
for two cases: (1) Hydrogen injection without combustion or cold-flow and (2) Hydrogen injection with
combustion. The obtained numerical results have been qualitatively and quantitatively compared with the 3-D
computational results published in [2] and experimental data [1].

1. Hydrogen injection without combustion (cold-flow)


Fig. 3 qualitatively compares the experimentally obtained density-variation capturing schlieren image,
with 2-D density contours obtained computationally for cold-flow. The density contour plot closely captures the
overall complex flow structures, as seen in experiments. The generation of shock waves from the tip of the
wedge and their reflection from the top and bottom wall, along with their interaction with the injected hydrogen
jet, has been captured in detail. Due to slight angle in the upper combustor wall, the shock wave getting reflected
off the upper wall hits the wake slightly further, as compared to the one getting reflected off the bottom wall.
This causes the wake to slightly expand, thus acting like a compression ramp for the flow on the upper side of
the wake, and thereby creating a shock wave which eventually merges with the upper re-compression shock, at
the upper wall.
Fig. 4 compares 2-D numerical simulation results (present work) for cross-stream velocity profiles at
four different streamwise locations, with published 3-D computational results and experimental data. The 2-D
results overall compare well with the 3-D results & experimental data. It can be noticed that, the flow in the
wake region just behind the wedge, decelerates to a subsonic velocity and then further speeds up back to
supersonic velocity, which achieves a near constant value close to the end of the combustor's length. However, at
X = 78 mm, along the centerline, i.e. Y = 25 mm, the deceleration in the streamwise velocity is largely over-
predicted by 2-D results, thus showing the presence of a re-circulation zone, as compared to 3-D results and
experimental data.
Fig. 5-6 show numerical and experimental results for static pressure distribution along the bottom wall
and centerline of the combustor. 2-D results provide a good overall agreement with the 3-D computational results
and experimental data. In Fig. 5, first pressure peak as shown by the experimental data is captured well by the 2-
D results, which on the other hand is slightly over-predicted by 3-D results. However, both 2-D and 3-D results
fail to capture the location of the second peak in pressure values along the bottom wall. In Fig. 6, the pressure
jump profile is captured well by the 2-D results as compared to 3-D results. But, after X = 175 mm, the plateau
as can be seen in the experimental data, is not captured well by the 2-D results, as compared to the 3-D results.
The reason for the above discrepancies between 2-D results with respect to 3-D CFD and experimental ones,
could be due to the influence of three dimensional effects which are neglected in the 2-D scenario. In the
experiment, the hydrogen is injected through 15 holes resulting in a truly three-dimensional flow field.
Furthermore, the pressure measurements were taken near the side walls of the channel where corner boundary
layer effects are present [1].

Figure 3. Comparison of shadow picture (experimental) with Density contour for cold flo
Streamwise velocity profile at X=78 mm Streamwise velocity profile at X=125 mm
Experimental [1] 2D results (Present) Experimental [1] 2D results (Present)
3D results [2] 3D results [2]
50 50

Position [0.0, 1.0, 0.0] mm


Position [0.0, 1.0, 0.0] mm

40 40

30 30

20 20

10 10

0 0
-100 0 100 200 300 400 500 600 700 800 900 400 500 600 700 800 900
Streamwise velocity (m/s)
Streamwise Velocity (m/s)

Streamwise velocity profile at X=157 mm Streamwise velocity profile at X=233 mm


Experimental [1] 2D results (Present) Experimental [1] 2D results (Present)
3D results [2] 3D results [2]
50 50
Position [0.0, 1.0, 0.0] mm
Position [0.0, 1.0, 0.0] mm

40 40

30 30

20 20

10 10

0 0
400 500 600 700 800 400 500 600 700 800
Streamwise velocity (m/s) Streamwise Velocity (m/s)

Figure 4. Cross-stream velocity profiles at different streamwise locations

Static pressure distribution along the bottom wall


Experimental [1] 2D results (Present)
3D results [2]
2.50E+05

2.00E+05
Static Pressuure Pa

1.50E+05

1.00E+05

5.00E+04

0.00E+00
0 50 100 150 200
Position [1.0, 0.0, 0.0] mm

Figure 5. Static pressure distribution along the bottom wall


Static pressure variation along the line Y=25 mm
Experimental [1] 2D results (Present)
3D results [2]
2.00E+05
1.80E+05
Static pressure Pa

1.60E+05
1.40E+05
1.20E+05
1.00E+05
8.00E+04
6.00E+04
4.00E+04
50 100 150 200
Position [1.0, 0.0, 0.0] mm

Figure 6. Static pressure variation along combustor centerline - Y=25 mm

2. Hydrogen injection with combustion


The results of the 2-D computational simulations of the cold flow with combustion are presented in this
section. The experimental schlieren image of the flow field and contours of density are compared in Fig. 7. The
overall flow physics, as depicted through the schlieren image, has been capture well by 2-D numerical
simulation. The same can be seen in the density contours of Fig. 7. The widening of the reaction zone followed
by a constriction has been captured well by the 2-D numerical simulation. The axial location of the narrow
combustion zone as seen in the 2-D density contours, closely matches with the one shown by the experimental
shadow graph image. The presence of this narrow combustion zone, acts like a kick to the subsonic flow within
the wake region, thus accelerating it back to supersonic speeds.
The profiles of the streamwise velocity components at three different axial locations along the length of
combustion chamber have been compared with 3-D computational and experimental results in Fig. 8. The 2-D
results are able to reasonably capture the cross-stream axial velocity flow profiles as shown by the experimental
data. Both, 2-D and 3-D results fail to predict the actual location as given by the experimental data in particular
along the combustor centerline, i.e. Y = 25 mm. For e.g. at X = 78 mm, both 2-D and 3-D results grossly over-
predict the velocity deceleration, leading to a negative velocity or a re-circulation zone due to combustion along
the centerline just behind the wedge base. Similarly, at X = 125 mm, both 2-D and 3-D over-predicts the
acceleration in axial velocity along the centerline in the wake region, which is not shown by the experimental
data. It can be noticed from the experimental data that, from X = 78 mm till X = 125 mm, the axial velocity in
the wake region remains more or less same, i.e. around 200 m/s. This tacitly communicates the extent of the
combustion zone within the wake region.
Fig. 9 shows streamwise axial velocity profile along the combustor's centerline within the wake region.
The same is compared to 3-D numerical results and experimental data. As can be noticed, 2-D result reasonably
captures the velocity trend, as shown by the experimental data. But, both 2-D and 3-D numerical results
demonstrate the presence of a re-circulation zone i.e. negative axial velocity in the wake region behind wedge
base. The reason for this anomaly could be attributed to the assumption of 1-step global reaction chemistry for
Hydrogen-Air combustion, which could be causing a rapid heat release versus gradual heat release pattern seen
during experiments. Thus the extent of combustion zone is longer in case of experiments, than what is noticed in
2-D CFD results. The same reason is the cause of an abrupt acceleration of flow within the wake region from X
= 78 mm to X = 207 mm, as seen in Fig. 8.
Fig. 10 shows the cross-stream temperature profiles at different streamwise locations. Here too, the 2-D
CFD results are able to capture the trends as shown by the experimental data. But, here also, the rapid heat
release due to the assumption of 1-step global reaction chemistry for Hydrogen-Air combustion, causes 2-D
results to largely over-predict the experimental temperature values near the base of the wedge, i.e. X = 78 mm.
However, as we move in the downstream direction, the 2-D results capture the cross-stream temperature profile
much accurately. The rapid heat release caused due to 1-step reaction mechanism, calls for the implementation of
a multi-step reaction mechanism for Hydrogen-Air combustion, so as to accurately model the heat release pattern
as encountered in reality.

Figure 7. Comparison of shadow picture with Density contour for combustion


Streamwise velocity profile at X=78 mm Streamwise velocity profile at X=125 mm
Experimental [1] 2D results (Present) Experimental [1] 2D Results (Present)
3D results [2] 3D Results [2]
50 50
Posiiton [0.0, 1.0, 0.0] mm

Posiiton [0.0, 1.0, 0.0] mm


40 40

30 30

20 20

10 10

0 0
-200 0 200 400 600 800 0 200 400 600 800
Streamwise velocity m/s Streamwise velocity m/s

Streamwise velocity profile at X=207 mm


Experimental [1] 2D results (Present)
3D results [2]
50
Posiiton [0.0, 1.0, 0.0] mm

40

30

20

10

0
400 500 600 700 800
Streamwise velocity m/s

Figure 8. Cross-stream velocity profiles at different streamwise locations

Streamwise velocity along the line Y=25 mm


Experimental [1] 2D results (Present)
3D results [2]
1400
1200
Streamwise Velocity m/s

1000
800
600
400
200
0
-200
50 75 100 125 150 175 200
Position [1.0, 0.0, 0.0] mm

Figure 9. Streamwise velocity plot along the line Y=25 mm


Temperature profile at X=78 Temperature profile at Temperature profile at
mm X=125 mm X=233 mm
Experimental [1] Experimental [1] Experimental [1]
2D results (Present) 2D results (Present) 2D results (Present)
3D results [2] 3D results [2] 3D results [2]
50 50 50

Position [0.0, 1.0, 0.0] mm

Position [0.0, 1.0, 0.0] mm


Position [0.0, 1.0, 0.0] mm

40 40 40

30 30 30

20 20 20

10 10 10

0 0 0
0 1000 2000 0 1000 2000 0 500 1000
Temperature K Temperature K Temperature K

Figure 10. Cross-stream temperature profiles at different streamwise locations

Summary and Conclusions


2-D numerical simulations of Hydrogen-Air combustion within the DLR combustor, have been
performed using STAR-CCM+ for two cases, i.e. hydrogen injection without combustion, and further with
combustion. For turbulence closure, the k-Omega SST (shear-stress transport) model has been used. The
Hydrogen-Air combustion modeling is performed using Eddy Break-up (EBU) model with 1-step global reaction
chemistry. Numerical simulation results have been compared both qualitatively and quantitatively with the
published 3-D computational results and experimental data. The 2-D CFD results from the present work show an
overall agreement with the published 3-D CFD results and experimental data. However, particularly near the
base of the wedge, i.e. X = 78 mm location, there is considerable disagreement between 2-D CFD results and
experimental values. One of the reasons could be the three dimensional effects, like three dimensional mixing of
fuel and three dimensional shock interactions with boundary layer and shear layer, are not captured by the two
dimensional simulation. Also, the presence of a re-circulation zone in the wake region behind the wedge base,
does not agree well with the experimental data. This could have been caused by the rapid heat release due to the
assumption of 1-step reaction chemistry for Hydrogen-Air. Hence, to realistically model the heat release pattern
for Hydrogen-Air combustion, a multi-step reaction mechanism should be used.

Future scope
To further improve results obtained above from 2-D CFD simulation, the chemistry could be modeled
using a multi-step reaction mechanism for Hydrogen-Air. A 7-step mechanism which is developed by Shang et al
[3], is shown to give better results for two-dimensional, steady-state, shock-induced combustion problem
according to Cluster [4]. Also, a Hybrid EBU combustion model could be used for turbulence-chemistry
interaction, which computes the reaction rate based on both mixing and chemical kinetic time-sales. All the
above recommendations are already being implemented and evaluated in an on-going project work at CD-
adapco, India.

Acknowledgements
I would like to thank CD-adapco for the supporting throughout the project.
References
[1] Michael Oevermann, Numerical Investigation of Turbulent Hydrogen Combustion in a Scramjet using
Flamelet Modeling, Aerospace Science and Technology, Vol. 4, No. 7, 2000,pp. 463-480.
[2] Dinde, P., Rajasekaran, A., and Babu, V., 3D Numerical Simulation of the Supersonic Combustion of
Hydrogen, The Aeronautical journal, 2006.
[3] Shang, H.M., Chen, Y.S., Liaw, P., Investigation of Chemical Kinetics Integration Algorithms for Reacting
flows, AIAA, 95-0806, Jan., 1995.
[4] Cluster, J.K., Mikolaitis, D.W. and Shyy, W., Effect of Reaction Mechanism in Shock-Induced Combustion
Simulations, AIAA Paper 98-0274, Jan., 1998.
[5] Menter, F. R., "Two-Equation Eddy-Viscosity Turbulence Models for Engineering Applications," AIAA
Journal, Vol. 32, No. 8, August 1994, pp. 1598-1605.

You might also like