Dilute Gas
Dense Gas
K
n
=
0
.
1
K
n
=
0
.
0
1
Slip
Flow
Transitional
Flow
Free
Molecular
Flow
Continuum
Flow
Navier-Stokes
Equations
Figure 2.1: List of ow regimes that can be used to model gas microow
2.1 Governing Equations
For transient reacting ow the conservation of mass, momentum, energy and species
is used. Solution of complicated physics such as turbulent uid ow, heat transfer,
chemical reaction and others are required.
2.1.1 Navier-Stokes Equation
Conservation of mass or the continuity equation can be written in general as following:
t
+ (u) = 0 (2.1)
It states that the rate of mass into a uid is equal to the rate of mass out of the uid.
The term on the left is the rate of time of the density and the term on the right is the
convective term which shows the ow of mass out of a uid. Second law of motion, named
after Sir Isaac Newton states that the net forces acting on an element is equal to the mass
of a uid element with an acceleration. The forces acting on the uid element can be
Chapter 2. Governing Equations and Theory 8
split in body forces (gravity, magnetic, centrifugal and coriolis forces) and surface forces
(pressure or normal forces and viscous or shear forces). The conservation of momentum
is derived from the Newtons second law of motion and can be written in general as
follows in 3 dimensions:
Conservation of momentum, x component:
(u)
t
+ (uu) =
p
x
+ (u) +S
Mx
(2.2)
Conservation of momentum, y component:
(v)
t
+ (vu) =
p
y
+ (v) +S
My
(2.3)
Conservation of momentum, z component:
(w)
t
+ (wu) =
p
z
+ (w) +S
Mz
(2.4)
The conservation of energy describes the energy of a uid. The transport equation
describes the rate of change of the energy inside the uid as being equal to the ux of
heat into the element and the rate of work done on the uid particle due to the body and
the surface forces. The equation can be written in general for compressible unsteady 3
dimensional ow as:
Conservation of energy:
(i)
t
+ (iu) = p u + (kT) + +S
i
(2.5)
where equation of state is:
p = p(, T) and i = i(, T)
perfect gas p = RT and i = C
T
(2.6)
and the dissipation function is:
=
_
2
_
_
u
x
_
2
+
_
v
y
_
2
+
_
w
z
_
2
_
+
_
u
y
+
v
x
_
2
+
_
u
z
+
w
x
_
2
+
_
v
z
+
w
y
_
2
_
+( u)
2
(2.7)
Chapter 2. Governing Equations and Theory 9
where
S
M
= S
M
+ [S
M
] is the source term or the body force, e.g. gravity
= dynamic viscosity
= kinematic viscosity
k = materials conductivity
(2.8)
2.1.2 Navier-Stokes Integral Equation
The most common discretization method in commercial codes is the nite volume
method. Given is the transport equation for the property
()
t
+ (u) = () +S
(2.9)
where is the diusion coecient and is the conserved property.
To be useful for nite volume methods the transport equation is integrated over the
control volume as follows:
_
CV
()
t
dV +
_
CV
(u)dV =
_
CV
()dV +
_
CV
S
dV (2.10)
The rst term on the left hand side represents the rate of increase of the property and
the second term is the convective term. Right hand side is the diusive term and the
last term is the source term. Using the Gausss divergence theorem the equation above
is rewritten with the convective and diusive term integrated over the bounding surface
of the control volume.
t
_
_
_
CV
dV
_
_
+
_
A
n (u)dA =
_
A
n ()
_
CV
S
dV (2.11)
then for steady state (
t
= 0) problems the equation becomes:
_
A
n (u)dA =
_
A
n ()
_
CV
S
dV (2.12)
while for time-dependant problems the equation is:
Chapter 2. Governing Equations and Theory 10
_
t
t
_
_
_
CV
dV
_
_
dt+
_
t
_
A
n(u)dAdt =
_
t
_
A
n()dAdt+
_
t
_
CV
S
dV dt (2.13)
2.1.3 Turbulence
Almost all uid will experience some turbulence. In this study the so called k- model
and the large eddy simulation (LES) model were used to model the combustion physics.
Turbulence models used in computational uid dynamics are very dierent and can be
very accurate like Direct numerical simulation and large eddy simulation but will take
extensive amounts of time and computer resources.
2.1.3.1 k Turbulence Model
The k- model is based on two algebraic equation, k equation which can be derived
from the Navier-Stokes equation and the equation which on the other hand cannot
be derived from the Navier-Stokes. The k equation uses the following transport
equation[38, 39].
Turbulence Kinetic Energy:
k
t
+ (ku) =
_
_
+ 2
t
S
ij
S
ij
(2.14)
Dissipation Rate:
t
+ (u) =
_
k
k
_
+C
1
k
2
t
S
ij
S
ij
C
2
2
k
(2.15)
where
t
= C
k
2
is the Eddy Viscosity and the follow are the standard coecients
C
= 0.09
k
= 1.00
= 1.3
C
1
= 1.44
C
2
= 1.92
(2.16)
Chapter 2. Governing Equations and Theory 11
2.1.3.2 LES Turbulence Model
Unlike the k model the Large Eddy Simulation computes the turbulence directly
at least partially. It is computationally heavy and requires very ne mesh. In short
the LES computes the eddies that are larger than the mesh spacing and models the
eddies that are smaller than the mesh spacing. The model to calculate the small eddies
is called subgrid scale model (SGS). In 1999 a new LES model was proposed by [40]
called Wall-Adapting Local Eddy-viscosity (WALE). The main improvements over the
Smagorinsky model were to account for the contribution in regions where irrational strin
is dominated by vorticity. In near wall the Smagorinsky gives a non-zero value of
t
when
velocity gradient exist and this has been improved with the new WALE model [40]. The
proposed eddy-viscosity for the WALE model is written as follows:
t
= (C
w
)
2
(S
d
ij
S
d
ij
)
3/2
(
S
ij
S
ij
)
5/2
+ (S
d
ij
S
d
ij
)
5/4
(2.17)
and the velocity gradient tensor
S
d
ij
=
1
2
( g
2
ij
+ g
2
ij
)
1
3
ij
g
2
kk
(2.18)
where g
2
ij
= g
ik
g
kj
and
ij
is the Kronecker symbol and the tensor
S
d
ij
=
S
ik
S
kj
+
ik
kj
1
3
i
j[
S
ik
S
kj
ik
kj
] (2.19)
and nally the vorticity tensor
ij
=
1
2
_
u
i
x
j
u
j
x
i
_
(2.20)
2.1.4 Combustion
To compute the chemical reaction the nite rate chemistry model was used in CFX
for the transient simulation. To initialize the ow and the reaction for the transient
solution the k model was used along with eddy dissipation chemistry model in the
time average simulation.
Chapter 2. Governing Equations and Theory 12
2.1.4.1 Finite Rate Chemistry
The last conservation equation needed in this study is the conservation of species and is
as follows [38, 41]:
t
(Y
k
) +
x
i
(u
i
Y
k
) =
x
i
_
D
k
Y
k
x
i
_
+
k
(2.21)
with species mass fraction Y
k
, diusion coecient D
k
and reaction rate
k
. The reaction
rate equation or the source term in the transport equation is
k
= (MW)
k
q
k
q
kj
=
dC
M
k
dt
j
= (v
k
v
k
)
_
k
f
N
k=1
(C
M
k
)
v
kj
k
b
N
k=1
(C
M
k
)
v
kj
_
(2.22)
the k
f
is the specic reaction rate constant or the Arrhenius law
k
f
= AT
exp(
E
a
R
u
T
) (2.23)
where
A = pre-exponent constant
= temperature exponent
E
a
= activation energy
(2.24)
2.1.4.2 Eddy Dissipation
Using the transport equation of species above the reaction rate for fuel is[38, 42]:
f
u =
k
C
EDC
min
_
Y
fu
,
Y
ox
s
_
_
1
_
(2.25)
where
= 4.6
_
v
k
2
_
1/2
=
Y
p
r/(1 +s)
Y
min
+
Y
p
r/(1 +s)
(2.26)
and
Y
min
= min
_
Y
f
u,
Y
ox
_
(2.27)
Chapter 3
Method of solution
This chapter will cover the technical setups that are used to perform the study. The
geometry was created with the aim to match the dimensions used in the paper by Salgues
et al [1] where they performed an empirical combustion tests using liquid oxygen and
methane. Geometry was created to match [1] and meshing included a RANS mesh and
a LES mesh. All simulations were performed using ANSYS CFX and both time average
and time dependent simulation were performed. k turbulence equation and LES
were applied and Eddy Dissipation model and Finite Rate Chemistry model for the
combustion.
3.1 The Geometry
To perform this study a rocket style combustor was created using Computer Aided
Design (CAD) tool. The design and dimension were chosen to match geometry from
empirical data for validation purposes. The geometry is a simple cylinder section with
two inlets at one end of the cylinder and outlet at the other end as can be seen in gure
3.1. The cylinder section is 292.1 mm (11.5 inches) long and has a diameter of 50.8 mm
(2 inches). There are two inlets, one is for the oxygen and the other for the methane.
The oxygen inlet is 3.429 mm in diameter, disk shaped, with its center at the center axis
of the main cylinder. The methane inlet is an annulus shaped with an outer diameter
of 5.182 mm and inner diameter of 4.191 mm. The center location of the methane inlet
is the same as the oxygen inlet. There is a wall annulus between the oxygen inlet and
the fuel inlet as indicated in gure 3.2 with an inner diameter of 3.429 mm and outer
diameter of 4.191 mm. The outlet, nozzle throat, is an opening of 9.652 mm in diameter.
Figure 3.3 shows the original schematic from [1].
13
Chapter 3. Method of solution 14
292.1 mm
Inlet Outlet
50.8 mm
Figure 3.1: Geometry created using Siemens NX showing main dimensions
Fuel inlet
Oxygen inlet
Figure 3.2: Closeup on the inlet section of the CAD, showing the oxygen and fuel
inlets
Due to combustion instability it was critical to model an inlet cylinder section with a
nite length to stabilize the simulation. In this study 6 dierent geometries were created,
all with dierent inlet length. Since the validation paper did not specify or show the
proper dimensions of the inlet, a trial and error approach was performed to match the
validation results. In gure 3.4(a) the geometry was modeled without a cylinder inlet
section. This proved to have very unstable characteristics and no feasible simulation
achieved. Figure 3.4(b)-3.4(f) shows the geometry with a cylinder section of 20 mm, 30
mm, 10 mm, 15 mm and 17.5 mm respectfully. The walls were not given any thickness
as it was not part of this study to include any structural nor heat transfer study.
Chapter 3. Method of solution 15
Figure 3.3: Original schematic of the validation combustor [1]
(a) Version 1, 0 mm (b) Version 2, 20 mm (c) Version 3, 30 mm
(d) Version 4, 10 mm (e) Version 5, 15 mm (f) Version 6, 17.5 mm
Figure 3.4: Comparison of dierent combustor inlet length
3.2 The Mesh
Even though the geometry is simple and relatively easy to mesh using structural hexa
mesh it was decided that the geometry would be meshed using unstructured tetra mesh.
Unstructured meshes are popular in industry because they are very exible, can t
around very complex geometry and do not need much interaction from the user. The
drawbacks are that in most cases the solver runs slower on unstructured meshes or it
takes longer to reach convergence [add reference]. To capture all complex ow conditions,
such as wakes, the user has to pack more elements in the wake region with unstructured
meshes. The mesh was constructed by creating surface mesh on each component, i.e.
inlets, walls and outlet and then volume mesh generated from the surface mesh.
In this study there were two meshes created, one for the time-average solution or RANS
simulation and the other for the Large Eddy Simulation. Both meshes are pretty ne
Chapter 3. Method of solution 16
Figure 3.5: Mesh created using Pointwise, LES mesh size
(over million cells) but only the LES mesh is used for the LES simulations and the mesh
sizes are set per Jiangs [3] typical mesh settings, see 3.2.
The RANS mesh has 1,048,720 cells or elements and 189,206 points or nodes. Com-
parison of the dierent mesh sizes are listed in 3.1. The LES mesh has 6,622,543 cells
and 1,149,690 points and can be seen in gure 3.5. The LES mesh uses wall model to
resolve the subgrid scale eddies and that way the cost of the mesh was reduced signi-
cantly as can be seen in 3.2. The meshes have cell volume ranging from 7.28914 10
13
to 7.74837 10
9
. The mesh characteristics can be seen in table 3.1, which shows the
mesh characteristics of the mesh used in this study as well as mesh characteristics from
Boudier et. al. LES mesh sensitivity study in complex geometry combustors [2].
Coarse mesh Fine mesh My RANS mesh My LES mesh
Total num. of points 230,118 7,661,005 189,206 1,149,690
Total num. of cells 1,242,086 43,949,682 1,048,720 6,622,543
Max. cell volume 3.12671 10
8
4.05748 10
9
7.74837 10
9
2.96836 10
11
Min. cell volume 1.81795 10
11
1.1828 10
12
1.07624 10
17
2.61382 10
18
Table 3.1: Comparison of mesh sizes [2]
DNS Wall-Resolved LES LES with Wall Model
Streamwise x
+
10-15 50-150 100-600
Spanwise z
+
5 10-40 100-300
WEall-normal (y
+
) 1 1 30-150
Number of points in 0 < y
+
< 10 3-5 3-5 -
Table 3.2: Typical LES mesh sizes [3]
Chapter 3. Method of solution 17
3.3 Boundary Condition and Procedure
This section will describe the boundary condition used at each mesh boundary, ow
condition and initial condition used in this study as well as the simulation procedure.
3.3.1 Walls
All walls were treated as smooth with no slip condition. No slip condition tells the solver
that the velocity at the wall the nodes are zero. In sub micro and nano scale or in slip
ow conditions, see 2.1, the velocity is not zero at the wall and the particles will slip.
Smooth wall condition is used to model smooth surface which will not cause disturbance
in the boundary layer like what would happen with rough surface.
3.3.2 Outlet
The ow regime in the outlet was treated subsonic but there are options to treat the
outlet ow regimes as supersonic. The outlet was set to have relative pressure of zero
and the outlet was treated as pure outow, that is, the ow can only ow out but not
in.
3.3.3 Inlet Methane
The ow regime in the methane inlet was treated as subsonic and this was also the
case for the oxygen inlet. Mass fraction for the inlet was set to 1 for CH4 to treat the
combustion model with purse methane. The methane was given a mass ow rate of
0.03923kg/s and temperature to 300 K.
3.3.4 Inlet Oxygen
To simulate pure oxygen the mass fraction was set to 1 for the O2 only. Most combustion
simulations have atmospheric air which would only contain 21% of the mass. The oxygen
was given a mass ow rate of 0.1177kg/s. The temperature was kept constant at 300 K.
3.3.5 Procedure
In the validation phase the RANS mesh was used only for quicker solution turnaround
time. Geometry 3.4(a) through 3.4(f) were tested and results compared with validation
Chapter 3. Method of solution 18
paper [1]. The focus was to compare the velocity of RANS simulation with the validation
and change the geometry till a matching velocity was found. Boudier et. al [43] indicated
in their study that velocity, along with other variables, were not so sensitive to mesh
resolution. Once a close match was found with the RANS mesh the LES simulation
with LES mesh was ran on the same geometry and results compared. The last step
was to compare combustion instability and use a new pulsating technique to reduce the
instability and increase performance.
Appendix A
Appendix Title Here
Write your Appendix content here.
19
Bibliography
[1] Delphine Salgues, Geraldine Mouis, Seong-Young Lee, D. M. Kalitan, Sibtosh Pal,
and Robert Santoro. Shear and swirl coaxial injector studies of lox/gch4 rocket
combustion using non-intrusive laser diagnostics. American Institude of Aeronautics
and Astronautics, (1):114, January 2006.
[2] G. Boudier, L. Y. M. Gicquel, and T. J. Poinsot. Eects of mesh resolution on large
eddy simulation of reacting ows in complex geometry combustors. Combustion and
Flame, 155(1-2):196214, October 2008.
[3] Xi Jiang and Choi-Hong Lai. Numerical Techniques for Direct and Large-Eddy
Simulations. Taylor and Francis Group, LLC, 2009.
[4] Roger E. Anderson, Scott MacAdam, Fermin Viteri, Daniel O. Davies, James P.
Downs, and Andrew Paliszewski. Adapting gas turbines to zero emission oxy-fuel
power plants. American Society of Mechanical Engineers, June 2008.
[5] Paul Barrett. Gas turbine range extender. In Low Carbon Vehicle Technology
Project at Low Carbon Vehicle Event 2011. CENEX, CENEX, September 2011.
[6] Naimish B. Harpal, Eric Besnard, and Reza Toossi. Modeling of lox/methane
impingement, mixing and combustion. American Institute of Aeronautics and As-
tronautics, (AIAA 2010-7135), July 2010.
[7] Eric A. Hurlbert. Liquid oxygen/methane propulsion for exploration systems space-
craft. In Helen W. Lane and Kamlesh P. Lulla, editors, 2007 Biennial Research
Report. NASA, NASA, December 2007.
[8] Huu P. Trinh. Liquid methane/oxygen injector study for potential future mars
ascent. American Institute of Aeronautics and Astronautics, (AIAA 2000-3119),
July 2000.
[9] Robert Zubrint, Steve Price, Larry Masont, and Larry Clark. Report on the con-
struction and operation of a mars in-situ propellant production unit. American
Institute of Aeronautics and Astronautics, (AIAA-94-2844), June 1994.
20
Bibliography 21
[10] Central Intelligence Agency. Natural gas - proved reserves, the world
factbook. http://www.cia.gov/library/publications/the-world-factbook/
rankorder/2179rank.html, 2011.
[11] R.J. Jensen, H.C. Dodson, and S.E. Clain. Lox/hydrocarbon combustion instabil-
ity investigation. NASA Contractor Report, July 1989.
[12] Songjiang Feng, Wansheng Nie, Bo He, and Yufeng Cheng. Three-dimensional
numerical simulations of low frequency combustion instability in a lox/methane
rocket engine. American Institute of Aeronautics and Astronautics, (AIAA 2010-
8776), August 2010.
[13] F. Zhuang, W. Nie, W. Zhao, and W. Liu. Liquid rocket combustion instability
analysis methodology - methods and representative examples. American Institute
of Aeronautics and Astronautics, (AIAA 98-3690), July 1998.
[14] F. Zhuang, W. Nie, W. Zhao, and W. Liu. Liquid rocket combustion instability
analysis methodology - methods and representative examples. American Institute
of Aeronautics and Astronautics, (AIAA 98-3690), July 1998.
[15] N. Ierardo and A. Congiunti. Mixing and combustion in supercritical o2/ch 4 liquid
rocket injectors. American Institute of Aeronautics and Astronautics, (AIAA 2004-
1163), January 2004.
[16] A.Minotti and C.Bruno. Comparison between real and ideal sub and supercriti-
cal combustion simulations of lo2-ch4 lre ames at 15mpa. American Institute of
Aeronautics and Astronautics, (AIAA 2008-951), January 2008.
[17] L. Cutrone, F. Battista, G. Ranuzzi, S. Bonifacio, and J. Steelant. A cfd method for
simulation of mixing and combustion in high-pressure lox/methane rocket engines.
American Institute of Aeronautics and Astronautics, (AIAA 2008-949), January
2008.
[18] Nicolas Guezennec, Matthieu Masquelety, and Suresh Menonz. Large eddy simula-
tion of ame-turbulence interactions in a lox-ch4 shear coaxial injector. American
Institute of Aeronautics and Astronautics, (AIAA 2012-1267), January 2012.
[19] A.Minotti and C.Bruno. Flame temperatures in non-premixed ames. American
Institute of Aeronautics and Astronautics, (AIAA 2008-998), January 2008.
[20] Naimish B. Harpal, Eric Besnard, and Reza Toossi. Modeling of lox/methane
impingement, mixing and combustion. American Institute of Aeronautics and As-
tronautics, (AIAA 2010-7135), July 2010.
Bibliography 22
[21] K. Tajiri and S. Menon. Les of combustion dynamics in a pulse combustor. Amer-
ican Institute of Aeronautics and Astronautics, January 2001.
[22] R. Erickson and B.T. Zinnt. One-dimensional numerical simulation of a pulse com-
bustor using the linear eddy mixing model. American Institute of Aeronautics and
Astronautics, July 2002.
[23] Subhashis Datta, Achintya Mukhopadhyay, and Dipankar Sanyal. Eects of ow
pulsations on the dynamics of a thermal pulse combustor. American Institute of
Aeronautics and Astronautics, August 2009.
[24] Mark D. Moore. Personal air vehicles: A rural/ regional and intra-urban on-demand
transportation system. American Institute of Aeronautics and Astronautics, July
2003.
[25] Ben T. Zinn. Pulse combustion: recent applications and research issues. Symposium
(International) on Combustion, 24(1):12971305, 1992.
[26] H. M. Heravi, J. R. Dawson, P. J. Bowen, and N. Syred. Primary pollutant pre-
diction from integrated thermouidkinetic pulse combustor models. JOURNAL OF
PROPULSION AND POWER, 21(6):10911097, November 2005.
[27] U.S. Department of Energy National Energy Technology Laboratory. Pulse com-
bustor design a doe assessment. Technical Report 1190, U.S. Department of Energy,
July 2003.
[28] Subhashis Datta, Achintya Mukhopadhyay, and Dipankar Sanyal. Eects of ow
pulsations on the dynamics of a thermal pulse combustor. American Institute of
Aeronautics and Astronautics, (AIAA 2009-5500), August 2009.
[29] Tao Geng, Fei Zheng, Andrey V. Kuznetsov, William L. Roberts, and Daniel E.
Paxson. Comparison between numerically simulated and experimentally measured
oweld quantities behind a pulsejet. American Institute of Aeronautics and As-
tronautics, (AIAA 2008-5046), July 2008.
[30] Michael Kowalkowski, Yu Matsutomi, and Stephen Heister. Flame sensing in pulsed
combustion using ion probes, diodes and visual indications. American Institute of
Aeronautics and Astronautics, (AIAA 2009-4945), August 2009.
[31] T. Bai, B. T. Zinn, J. I. Jagoda, B. R. Daniel, and X. Chen. Experimental study
of the eect of fuel and air feed rate modulations upon pulse combustion in a rijke
type pulse combustor. American Institute of Aeronautics and Astronautics, (AlAA
94-0216), January 1994.
Bibliography 23
[32] Taichang Zhang, Xuejun Fan, Jing Wang, Jianguo Li, and Gong Yu. Pulsed com-
bustion of hydrocarbon fuels in a supersonic model combustor. American Institute
of Aeronautics and Astronautics, (AIAA 2011-5547), August 2011.
[33] Michael Zeutzius, Toshiaki Setoguchi, Kunio Terao, and Hideo Miyanishi. Propul-
sion system for hypersonic space planes. Journal of Propulsion and Power, 16(2):
243250, March 2000.
[34] K. Tajiri and S. Menon. Les of combustion dynamics in a pulse combustor. Amer-
ican Institute of Aeronautics and Astronautics, (AlAA 2001-0194), January 2001.
[35] Andrew C. St. George and Ephraim J. Gutmark. Trends in pulsating turbine
performance: Pulse-detonation driven axial ow turbine. American Institute of
Aeronautics and Astronautics, (AIAA 2012-0769), January 2012.
[36] Uche Oteh. Mechanics of Fluids. AuthorHouse, 2008.
[37] George Karniadakis, Ali Beskok, and Narayan Aluru. Microows and Nanoows
Fundamentals and Simulation. Springer Science and Business Media, Inc., 2005.
[38] H K Versteeg and W Malalasekera. Introduction of Computational Fluid Dynamics
The Finite Volume Method. Pearson Education Limited, second edition, 2007.
[39] David C. Wilcox. Turbulence Modeling for CFD. DCW Industries, Inc., second
edition, 1994.
[40] F. Nicoud and F. Ducros. Subgrid-scale stress modelling based on the square of the
velocity gradient tensor. Flow, Turbulence and Combustion, 62(3):183200, April
1999.
[41] Thierry Poinsot and Denis Veynante. Theoretical And Numerical Combustion. R.
T. Edwards, Inc., second edition, 2005.
[42] Guan Heng Yeoh and Kwok Kit Yuen. Computational Fluid Dynamics in Fire
Engineering Theory, Modeling and Practice. Butterworth-Heinemann, 2009.
[43] G. Boudier, L. Y. M. Gicquel, and T. J. Poinsot. Eects of mesh resolution on large
eddy simulation of reacting ows in complex geometry combustors. Combustion and
Flame, 155(1-2):196214, October 2008.