Академический Документы
Профессиональный Документы
Культура Документы
I Facolt di Ingegneria
Corso di Laurea in Ingegneria Aerospaziale
Tesi di Laurea
Relatori:
prof. Michele Ferlauto
prof. Dario Pastrone
Candidato:
Sofia Sartori
May 2010
Summary
Plastic bags, vinyl discs, nylon... in the last 50 years they have become part of our
every day life. The growth of these products have meant the growth of the petrochemical industry too (cap 1). One of its main product is ethylene and it is usually
produced through steam cracking, a pyrolysis process were really high temperature
and innitesimal residence time lead to higher yield.
At the University of Washington, in Seattle, a new way to improve olen's production has been realized: the Shock Wave Reactor.
Shock wave reactor is presented in chapter 2 as a new method of pyrolysis, specifically aimed at olen production (ethylene in our case), which circumvents tube
temperature limitations by using shock wave (rather than convective heat transfer)
to control the temperature history of a feedstock.
II
Acknowledgments
I would like to thank Professor Tom Mattick and Professor Carl Knowlen to let
me being part of this challenging project. Besides thanks also to the University of
Washington and to Professor Adam Bruckner as represent of the Aeronautics and
Aerospace department to welcome me in their community .
A specially thanks also to Professor Ferlauto Michele and Pastrone Dario who believed in my work during all my stay in the U.S.A. and allowed me to follow this
new project.
III
Contents
Summary
II
Acknowledgments
III
1 Petrochemical industry
1.1
Technique of production
2
. . . . . . . . . . . . . . . . . . . . . . . . .
1.1.1
Cracking conditions . . . . . . . . . . . . . . . . . . . . . . . .
1.1.2
. . . . . . . . . . . . . . . . .
2.1
10
2.2
Nowadays project . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
Scramjet engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
2.3.1
19
3.3
Inputs
24
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
3.2.1
. . . . . . . . . . . .
27
3.2.2
Method of characteristics . . . . . . . . . . . . . . . . . . . . .
31
Nozzle analysis
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
39
3.3.2
42
3.3.3
Thermal analysis
. . . . . . . . . . . . . . . . . . . . . . . . .
46
3.3.4
Ansys analysis . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
3.3.5
52
4 Injectors design
4.1
38
3.3.1
59
60
4.1.1
. . . . . . . . . . . . . . . . . . .
60
4.1.2
Possible congurations. . . . . . . . . . . . . . . . . . . . . . .
61
IV
4.1.3
4.2
64
. . . . . . . . . . . . . . . . .
73
4.2.1
Fin analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . .
77
4.2.2
78
83
5.1
83
5.2
Number of injectors . . . . . . . . . . . . . . . . . . . . . . . . . . . .
86
5.2.1
90
5.2.2
Further optimisations . . . . . . . . . . . . . . . . . . . . . . .
94
6 CFD analysis
101
6.1
6.2
6.3
Bibliography
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Abstract
Chapter 1
Petrochemical industry
At presents days petrochemical is a very important sector of industry world and in
particular of the American one. The two main classes of petrochemical raw materials are olen (including ethylene and propylene) and aromatics . Both are produced
in very large quantities : World production of ethylene is around 110 million tonnes
per annum, of propylene 65 million tonnes, and of aromatic raw materials 70 million
tonnes. The largest petrochemical industries are in the USA and Western Europe,
though the major growth in new production capacity is in the Middle East and Asia.
1 Petrochemical industry
Ethylene (IUPAC name: ethene) is the chemical compound with the formula C2 H4 .
It is the simplest alkene. Because it contains a carbon-carbon double bond, ethylene is called an unsaturated hydrocarbon or an olen. It is extremely important in
industry and it is the most produced organic compound in the world.
Ethylene is used to make glycol, vinyl's, styrene's, and polyethylene. The applications of ethylene are numerous and ethylene derivatives are traded around the world.
Ethylene is primarily a petrochemical derived monomer used as a feedstock in the
production of plastics, bres and other organic chemicals that are ultimately consumed in the packaging, transportation and construction industries and in a multitude of industrial and consumer markets (packaging uses make up more than half
(55%) of ethylene derivative consumption worldwide).
Ethylene is used for dierent purposes: 80% is employed to create ethylene oxide,
ethylene dichloride, and polyethylene; in smaller quantities it is used as an anaesthetic agent, to hasten fruit ripening, as well as a welding gas.
Polyethylenes of
various density and melt ow account for more than 50% of world ethylene demand.
The primary use of polyethylene is in lm applications for packaging, carrier bags
and trash liners. Other applications include injection moulding, pipe extrusion, wire
and cable sheathing and insulation, as well as extrusion coating of paper and cardboard.
(500680 C, depending on the feedstock). The stream then enters a red tubular
reactor (radiant tube or radiant coil) where, under controlled residence time, tem
perature prole, and partial pressure, is heated from 500650 C to 750875 C for
0.1
0.5 s. During this short reaction time hydrocarbons that are in the feedstock
are cracked into smaller molecules : ethylene, other olen, and diolens are the major products. Since conversion of saturated hydrocarbons to olen in a radiant tube
1 Petrochemical industry
is highly endothermic, high energy input rates are needed. Reaction products leav
ing the radiant tube at 800850 C are cooled to 550660 C within 0.02 0.1 s to
prevent degradation of highly reactive products by secondary reactions. Resulting
product mixtures, which can vary widely, depending on feedstock and severity of
the cracking operation, are then separated into desired products by using a complex
sequence of separation and chemical-treatment steps.
used for all feed-stocks from ethane to gas oil, with a limitation in the end point
of the feedstock of 600 C. Higher boiling materials can not be vaporized under
the operating condition of a cracking furnace. Pyrolysis of hydrocarbons has been
studied for years, much eort has been devoted to mathematical models of pyrolysis
1 Petrochemical industry
reactions for use in designing furnaces and predicting the products obtained from
various feedstock under dierent furnace conditions. Three major types of model
are used: empirical or regression, molecular and mechanistic models .
The most important reactions in the cracking are the break of C-C bonding and the
dehydrogenation :
1 Petrochemical industry
Olen are thermodynamically unstable at all temperature but, compared with parafns with same number of C atoms,their stability increase with temperature . For
example over 1066 K ethylene is more stable than ethane. Generally speaking parafns and naphthas are more stable under 500 K while olen and aromatics are more
stable at higher temperature.
However over 820 K all hydrocarbons are less stable than their components : carbons
and hydrogen. From this consideration is understandable that cracking thermodynamic equilibrium is not reached but it is required that reactions go just in one way.
During this process feedstock is rapidly vaporized and heated until cracking temperature in a steam solution. After keeping the feedstock in cracking temperature
range for a few seconds, reaction is suddenly stop by cooling. Reaction time is very
important because inuence secondary reactions.
Usually long residence time are avoided because they let to thermodynamic equilibrium, that's mean decomposition of cracking primary products until carbon and
hydrogen.
Summarizing olen production is promoted by: high temperature, low pressure and
short residence time.
The most important parameter in this process is oil eective partial pressure in
cracking zone that is calculated as follow.
If cracking area absolute pressure is P atm and there is no steam as diluent, P is
also oil partial pressure . If ,instead , there are x water-vapour moles , oil partial
pressure is :
Usually in industrial cracking oil partial pressure around 0.1, 0.4 and 0.6 are used .
Feedstock is diluted into steam for three main reasons :
- hydrocarbon partial pressure is lower , therefore process selectivity is improved;
- para-aromatic hydrocarbons partial pressure in cracking zone is really low and this
could avoid coke envelopment and other deposit from carbon ;
- water vapour get down metal catalytic eect that can lead to coke formation .
Steam/hydrocarbon ratio is set depending on dierent factors as : feedstock quality
, cracking strictness , products distribution , set up features, total pressure at reactor exit .
Considering the features of industrial process it appears that to maximize the rate
of the cracking reaction :
- temperature can be increased so that gas particles move more quickly and collide
more often
- increase pressure which forces gas particles closer together and collide more often
- no catalyst is needed to increase the rate of this reaction since steam provides the
required activation energy.
1 Petrochemical industry
at a pressure lower than the atmospheric one, ethylene and acetylene are developed.
In ethane pyrolysis equilibriums are :
they are going to be promote by drop of pressure and inert gas introduction.
At low conversions, high temperatures, low pressure, ethane decomposition speed is
of the rst order and activation energy is about 72 kcal/mol. Increasing temperature
cracking speed and ethane conversion are increased. However, time has to be short
to avoid secondary reactions and development of free carbon (coke).
Because activation energy decrease with carbon atoms increasing in the molecule
where cracking is happening, also cracking temperature is going to decrease. Ethane
cracking products are shown in g. 1.4 .
Figure 1.4. Main product yield towards residence time at cracking conditions
(815 C, 1 atm)
1 Petrochemical industry
The main eects of the pressure are two : while pressure is raising dehydrogenation
is obstructed (reversible reaction, that implied moles number increasing) ; whereas
because C-C bond breaking is not a reversible reaction pressure hasn't any eects on
it. Instead it is going to promote condensation and polymerization olen reactions.
Therefore when an high yield of liquid products is desired, pressure is increased.
When high yield of inferior olen (ethelyne and propylene) is required low pressure
is sets.
Pressure can also aects where C-C bonding is going to break : at high
pressure hydrocarbons break in the middle of the chain and at low pressure they are
going to break at the extremity.
Chapter 2
The shock wave reactor
Steam cracking process used in industries have been improved during the last 50
years increasing feedstock and steam temperature and decreasing residence time in
order to avoid both heat transfer losses and secondary product developing. Moreover it has been long known that olen yield from hydrocarbon pyrolysis increases
with use of high processing temperatures and short reaction durations (g. 2.1).
the properties of the tube materials. Despite impressive yield gains resulting from
use of improved alloys and optimal tube geometries, much improved yields could
result from application of a short high-temperature " pulse " to the feed.
When an aeronautic engineer thinks to rise in temperature, pressure losses and short
time that easily let to think about shock wave (g. 2.2).
Moreover be-
cause heating is instantaneous and concerns the all volume fouling are minimized.
The last point of interest is that having a continuous ow device let possible to scale
the facility for bigger ow rate.
temperature TP , and was expanded to supersonic speed into a mixing channel. The
feedstock was preheated to a sub pyrolysis temperature Tf and was expanded supersonically into a channel parallel to the carrier uid (g. 2.4).
10
The purpose of accelerating carrier uid as it was heated was to increase uid's
enthalpy. Heated carrier uid gas was conveyed in a channel from heater section to a
nozzle section. This section acted like a venturi and expanded carrier uid adiabatically to higher velocity and lower temperature. Temperature must be kept below
pyrolysis temperature after deceleration of steam.
got pressurized. Ethane was also preheated. Preheating feedstock provides a higher
11
stagnation enthalpy downstream in mixing section than would be without this preheating.
Hydrocarbon used as feedstock was injected with a nozzle in order to accelerate it,
producing an higher velocity. Velocity of hydrocarbon, after injection, is lower than
mach speed of the carrier uid. Dierence between feedstock velocities and carrier
uid stream in mixing section leads to turbulent mixing of feedstock and carrier
uid. As mixing proceeds, temperature of stream approached an uniform temperature, intermediate between initial stream temperature at entrance of mixing section
and feedstock temperature at injection point. Stream of feedstock and carrier uid
fully mixes in mixing section and leaves this section as a supersonic mixture with a
common mach speed. A shock wave happens when velocity decreases. This means
that kinetic energy is converted in perceptible energy.
Reaction proceed
adiabatically, resulting in temperature decrease dependent on initial carrier temperature dilution. When desired conversion is reached reaction is rapidly quenched to
freeze product distribution using a steam cracker heat exchangers. In fact, cracked
gases coming from pyrolysis section are at 800-1300 C and they should be cooled
instantaneously to preserve their composition.
12
that preheat steam until 1400 K four H2 - O2 injectors are placed to start combustion process that allows to reach higher temperature (2100 K). Injectors provide the
two gas with a factor of dilution H2 : O2 of 2.5 : 1 to avoid oxygen surplus that
could lead to explosion when in contact with methane. These are followed by three
section of 4" tube where combustion, supersonic expansion and mixing are going to
happen (g. 2.5).
Combustion section is 30" long and is followed by a nozzle that has to accelerate
main ow to Mach Number near 2.2 . The nozzle has been designed trying to come
to an agreement between a very short shape, to avoid heat transfer losses, and a
longer one in order to avoid oblique shocks at the exit that are going to cause pressure losses .
Sonic injection of preheated methane follows the nozzle and last part of the 4" tube
include mixing section (30" long).
Through this tube we are going to have very high steam temperature (2100 K) so,
since in the previous work, channel temperature prole was dominated by convection
loss rather than endothermic chemistry; to avoid that, a new system of insulation,
realized with thin-walled liners, has been settled along both combustion and mixing
sections.
An outer liner (external walls) has been realized in stainless steel 304 (Tmelt about
1670 K) with thickness of 0.05", whereas two inner liners made by titanium (Tmelt
about 1950 K), thick 0.016", have been placed between main ow and outer ones.
Inner liners are also vented in order to minimize radial force and they are separated
by ceramic fabric spacers as shown in g. 2.6 to avoid contact due to thermal expansion during the experiment .
Thanks to inner liners temperature decrease and, since radiation grow with fourth
power of temperature, after, inner liners radiation contribute becomes negligible
13
P = AT 4
In this equation
14
Thanks to this new idea temperature history would looks like as follow g. 2.8
and substantial pressure losses could be avoid .
Figure 2.8. Temperature prole in 4 " section (Combustion and mixing section)
At the end of 4" tube shock wave section takes place in the 10" tube to decelerate
main ow and reach pyrolysis conditions.
going to start in order to recover pyrolysis temperature and initialize the reaction
(g. 2.9).
Due to weak oblique shocks and boundary layer interactions a gradual deceleration
is achieved and main ow could, nally, enter in the last section of this facility .
Just after injection section we have the test section pressure tap where thermocouples
and gas samples are placed to control reaction and detect shock location (g. 2.11).
15
Right at the end of reactor section two spray bars have the task to set in the
right place shocks wave and freeze reaction . Bars inject colder water spray to cool
gas until 800 K before they reach soot sensors and the dump tank (g.2.11) and,
adding mass to main ow they chock upstream ow causing shock, whose position
is controlled by the amount of water that is injected through the mass ow.
The last section is the diagnostic one were thermocouples and pressure sensors are
placed to verify nal conditions of ow.
16
by huge rockets that are very heavy, no-reusable, with low manoeuvrability and a
vertical take o is required.
these purposes because they are limited by the turbine blade velocity and also at
the plane velocity increase, more heat is released so there is a risk of melting for
the engine. To allow ight at higher speed avoiding problem of compressor-turbine
velocity and temperature load engineers have developed a new engine:
jet.
the ram-
compressed by the high speed of the vehicle, a combustion chamber where fuel is
combusted, and a nozzle through which the exhaust jet leaves at higher speed than
the inlet air.
Ramjet can y until Mach 6 but at these speed when the air ow
slow down to subsonic to enter in the combustion chamber temperatures rise until
3500 K leading to chemical dissociation. So, when the combustion starts, instead
of producing water (that will be followed by a very high increment in pressure and
17
a thrust) the reaction produce free radicals at lower pressure and with lower thrust
causing a deceleration of the plane. To avoid this deceleration, they have designed
the intake to slow down the air ow at the inlet less than before in order to have a
supersonic, and not subsonic, combustion. The history of these new engine started
in the World War II, when there was a lot of eort put in research towards high
speed jets and rocket aircraft as military weapon. In particularly the U.S. Navy's
needed an anti air weapon that could defeat aircraft threats using small, high-speed,
radar-guided anti-surface missiles. This request led to rocket- and ramjet-powered
ight vehicles.
When the war was over this new technology have been adopted by civil industries
however, the main goal of civilian air transport has been reducing operating cost,
rather than increasing ight speeds. Because supersonic ight using conventional jet
engines requires signicant amounts of fuel, airlines have favoured subsonic jumbo
jets rather than supersonic transports.
In the United States, from 1986-1993, a reasonably serious attempt to develop a
single stage to orbit reusable space plane using scramjet engines was made, but
the Rockwell X-30 (NASP) program was discontinued before it yielded any working
hardware.
Hypersonic ight concepts haven't gone away, however, and low-level investigations
have continued over the past few decades.
At the beginning the attention was focused on ramjets but scramjet oer several
improvements like :
- more eciently operational o-design;
- less dissociation of combustion products and lower static pressure, that causes
less heat transfer as well as structural load and ensure less weight. This thanks
to the supersonic combustion that produce lower static temperature in the
combustion chamber .
Nowadays Scramjet technology is seen as one with high potential, there are many
applications that could be developed. Right now they are thinking about applying
this to space launches but it is still an open topic with advantages and disadvantages
too.The scramjet is seen as an attractive alternative to rocket propulsion; where a
rocket carries its oxygen and fuel on board, the scramjet captures its oxygen from
the atmosphere thus making the vehicle lighter. Potential applications for scramjet
include satellite launch vehicles and hypersonic cruise missiles.
This engine oers dierence improvements because of its innovative conguration
that uses the speed of the aircraft to compress air, reducing drastically number of
moving parts . This is done with a particular inlet conguration and, after that, supersonic combustion take place releasing really hot and supersonic exhausting gases.
Not having high-speed turbine, as in a turbofan or turbojet engine, and compressor we can save a lot of weight, Higher specic thrust since it has not temperature
18
they are still looking for optimizing every single stage since, scramjet performance
are strictly related to the eciency of every single components.
19
Figure 2.13. Comparison between shock wave reactor and Scramjet structure.
duct that sits between intake and combustor. It allows a supersonic ow to adjust
to a static back-pressure higher than the inlet static pressure. When the combustion process begins to separate the boundary layer, a precombustion shock forms in
the isolator. The isolator also enables the combustor to achieve the required heat
release and handle the induced rise in combustor pressure without creating a condition called "inlet unstart", in which shock waves prevent airow from entering the
isolator.
which arise from injecting fuel into a supersonic airow, from propagating upstream
and aecting the intake's capture characteristics.
Long isolators are benecial for containing shock-trains, but when combined with
intake performance, the higher internal friction results in increased compression and
reduced eciency, with the overall eect being a reduction of engine performance.
However, the sudden pressure change across internal shock waves in the isolator
causes the ow at the wall to reverse locally as the ow near the surface separates
from the wall, which acts towards reducing the overall skin friction in the isolator.
In inlet design multiple weak shocks are preferred, instead of a strong one, because
20
with them static temperature will not rise too much and will not cause dissociation.
Also in our case a set of oblique shocks is preferred to a normal one because it led
to less total pressure losses.
In Shock Wave reactor a straight-constant area is present too. Here, however the
sequence of oblique shocks take place after the injection section and not before.
Moreover our "isolator" has the double purpose of enhance the mixing of CH4 with
the main ow and allow a recovery in static temperature to pyrolysis range.
Shock recompression within a duct is connect with interaction of shock waves with
subsonic region near the side walls, in fact is well known that the presence of boundary layers causes the recovery of pressure in a supersonic duct ow occur via a
spread-out recompression zone, often with multiple shocks, instead of sharp, normal
shock that spans the entire ow channel.
the boundary layers thickness increases at the incipient shock station is shown in
g. 2.14.
separation bubble in eect dening a converging/diverging channel for the core ow
[6.7]. As the boundary layers thickness increases, this eect leads to re-acceleration
of the ow to supersonic speed following an initial shock, and the formation of a successively weakr shocks, each exhibiting this re-acceleration features via the presence
of separation. Foe even thicker boundary layers, the lambda shocks degenerate to
a sequence of oblique shocks. Supersonic pressure recovery within a duct is fundamentally related to the oblique waves formed near the wall; the normal components
are merely Mach reections about the centreline.
After the isolator, in a scramjet, you have the injection section that is usually really
short.
In our experiments we have this section too and during the design phase we consider scramjet typical injecting techniques (side wall injection, ramp, n, centreline
21
injector, aeroramp..) g. 2.15, however our system require much ore space.
a combustion, doing the same thing without oxygen make the compose experience
chemical bounds division with creation of simpler ones. During pyrolysis, heat supply is used to divide chemical bounds .
Therefore we can see that the main dierence between these two process is that the
rst is used to produce energy while the second one use energy to achieve particular
products. The other big dierence is that the rst don't need to be stop before the
equilibrium is reached, on the contrary more the reaction is complete better it is
since there is more thermal energy to convert in kinetics one while, on the other
side, pyrolysis has to be frozen at a particular time to obtain the right type and
amount of composes.
22
Last section is the diverging nozzle. In scramjet we have an inlet at the entrance and
a nozzle at the exit but they are completely dierent from our converging-diverging
nozzle both for purposes and operating conditions.
The inlet have to receive supersonic ow and slow it down while the rst part of our
nozzle receives subsonic ow and has to accelerate it until sonic conditions.
Scramjet nozzle has to accelerate as much as possible the ow in the shortest length
while, our diverging section hasn't just to accelerate it using the shortest shape, but
also has to avoid oblique shocks after the exit since this is still part of the experiment
and big pressure losses are not acceptable.
23
Chapter 3
Supersonic nozzle design
In this chapter are explained the main techniques used in rocket nozzle design and
then our solution for the shock wave reactor is shown. Then several analysis (structural, thermal and uidodynamic) have been carried on.
3.1 Inputs
The realization of a new shock wave reactor experiment have been carried out from
August 2009, dividing work in dierent team in order to accelerate the process.
My task has been design converging-diverging supersonic nozzle and methane sonic
injectors.
This has been a new topic from the previous experiment where ethane was injected
subsonically . As input for our work we have had some constrains, concerning nozzle
design, and some parameters that have to be settled in injectors design too.
Initial conguration that we have to deal with is shown in g. 3.1 .
Firstly we
24
Meanwhile, concerning injectors design, main issue was to estimate penetration high
for a given injection pressure and diameter considering our domain (geometry of the
duct) and main ow characteristics . Once found a good modellization we had to
look for the best conguration in order to provide optimum mixing both in area
covered, shortest time required and less pressure and velocity losses.
To start a preliminary study we need to know which pressure our facility can provide
for methane injectors and which mach number we have at the throat.
3.2 Models
?
25
Most common shape for rocket nozzle suitable also for our project are :
nozzle with a 15-deg divergent half-angle and bell nozzle.
conical
shape realized through method of characteristic that led to adapted conditions with
the smoothest contour shape but, because of longer length this technique is not
usually applied in rocket nozzle design.
26
of characteristic).
Then looking at both contours we will extrapolate a new shape that is going to be
a compromise between these two optimums .
Design diverging section of a bell nozzle requires following data : throat diameter
Dt , axial length of nozzle from throat to exit plane (Ln ) expansion area ratio
of exit area to throat area), initial wall angle of parabola
angle
(ratio
e .
Wall angles
and
have been chosen looking at Fig. 3.5 where they are plotted
This last value indicate how long is going to be our bell nozzle compared with a
15 degree conical one that has same expansion area ratio and an initial wall angle
27
of 15 degree.
is equal to 2.58 (with exit diameter of 3.5"), looking at Rao plot g. 3.5
n = 34
and
e = 22 .
z = aR + bR + c R =
2
b2 4a(c z)
2a
1
tan2 e
1
b2 4a(c zn ) =
tan2 n
ze zn = LP
b2 4a(c ze ) =
(3.1)
(3.2)
(3.3)
(3.4)
R(ze ) =
b +
b2 4a(c ze )
2a
28
b +
2a
1
tan2 e
(3.5)
R(zn ) = R =
b +
b2 4a(c zn )
2a
kn =
( ke b)2
R =
4a2
( kn b)2
2
(R ) =
4a2
b +
1
tan2 n
2a
1
and
tan2 n
ke =
3.3,
3.4,
(3.6)
1
we get
tan2 e
(3.7)
(3.8)
3.7 and
3.8)with 5 unknown
n = 19
and
e = 12 .
Referring to Rao approximation it comes that our preliminary nozzle is very short
since converging section is going to be 1.5 and diverging section is going to be other
1.5", for 3 in the hole that is a really good result for heat losses.
Unfortunately, when injectors design has been started, it comes out that to achieve
29
3.6 .
30
continuity
ux + vy + (ux + vy ) = 0
(3.9)
momentum along x
uux + vuy +
px
=0
(3.10)
uvx + vvy +
py
=0
(3.11)
momentum along y
energy
uSx + vSy = 0
Considering:
d
2 da
dS
=
1 a
1
2 da
dS
dp
=
p
1 a
1
Replacing those in continuity and conservation equations we get:
1
1
aux +
avy = 0
2
2
2
a2
uux + vuy +
aax
Sx = 0
1
( 1)
uax + vay +
31
(3.12)
uvx + vvy +
2
a2
aay
Sy = 0
1
( 1)
uSx + vSy = 0
We obtain
ax
and
ay
nuity equation.
Then from Crocco's equation for steady state motion (
enthalpy (h
= 0)
= rot(q).
= 0)
we have :
q =
With
T
S
1
3.14, 3.15).
a2 Sy
=0
( 1)u
(3.13)
(3.14)
uSx + vSy = 0
(3.15)
Then we try to pass from a PDE system to an ODE one building a linear combination.
a2 Sy
)+3 (uSx +vSy ) = 0
( 1)u
(3.16)
1 , 2
and
solved:
a2
) + 3 (u u) = 0
( 1)u
M 2 1.
expression on 1
With
The
uv a2
;
1 2
u a2
and
1 = 2 = 3 = 0,
v
2 = ;
u
we get:
uv + a2
3 2
u a2
left-hand (C+ ) and right-hand (C ) characteristics and its are real only if the ow
is supersonic (g. 3.7).
32
We were able to reduce the equations to ordinary dierential ones that are valid along
characteristics curves, which represent the path of propagation of disturbances in
such ows. A characteristic is dened as a curve along which the governing partial
dierential equations reduce to an "interior operator", that is a total dierential
equation known as compatibility equation.
The compatibility equations describe the variation of ow properties along the characteristic curves. In the table below
33
The intersection of the line segment, wall contour, and the C+ characteristic
denes the nozzles contour. Line segments always emanate from the end of the previous line segment. For the rst line segment, its origination point is the last point
on the known expansion arc. Since we start from the symmetric axis of the nozzle
this technique result simplied. Because the particular position of these points only
the right-running characteristic is employed .
However to start the calculation here, we need values of Mach number along the
symmetric axis so, since we know the throat area (At ) and the exit one, assuming a
34
preliminary linear development of the contour nozzle wall along the diverging secA
tion, knowing the ratio i , with this equation 3.17 we use Newton Rhapson method
At
and we dened Mach number development along that axis .
At
=
Ae
+1
1
2
1 2 1
1
+
M
)]
e
Me2 + 1
2
(3.17)
This method is able to provide a very smooth contour nozzle contour shape knowing
which are ow condition up steam the nozzle.
We decide to implement a Matlab program written by B.J. Olson that solve this
problem for a quasi-2 dimensional ow based on Anderson equation.
This program estimate the shape that is going to give the highest thrust, that mean
that the exit area is not going to be an input (as for as) but it's going to be an
output. Still, we decide to use this program because we are going to scale it anyway
and the exit area coming from this is not going to be much more bigger than our
exit area (less than half of an inch).
Estimating Mach numbers along nozzle centreline, with a linear approximation, this
program found the exit area that led to an adapted nozzle and then it applied the
method of characteristic to solve the nozzle inner ow eld and give us the best contour shape. This method becomes more accurate as the number of characteristics
used in the calculation increases and it proceed tracing characteristic curves in the
direction of the main ow (forward method g
3.10).
The program used requires as input entrance and exit conditions : temperature,
pressure, ratio between heat capacities ( ) and specic gas constant (R) .
Before the nozzle a combustion take place so we decide to use the program CEA to
estimate our main ow properties. CEA (Chemical Equilibrium with Applications)
35
J
.
KgK
Then, the last input required, is throat radius that we can calculate since the mass
Kg
ow is xed (0.35
) and we want the throat to be chocked.
s
First of all we used mass ow equation through a generic section to estimate the
m
= uA
Assuming an isentropic expansion of the ow through the nozzle we used gas perfect equation and replaced static magnitude with stagnation ones expressed as Mach
functions, so the previous equation became :
m
=
p0
RT0 (1 +
M
+1
1
M 2 ) 2(1)
2
(3.18)
As we said to control mass ow, a chocked throat is required, so replacing the values
in equation ( 3.18) with our condition, throat diameter (dt ) has been calculated from
the area :
Mt = 1 Mach number at the throat
= 1.19
36
Figure 3.12. Contour shape of divergent section of a supersonic nozzle design with
characteristic method.
37
The only problem in our case is that we don't have restrictions at nozzle exit just
for pressure and temperature but also our exit area has been already chosen. But
since this shape is going to be use just as approximation and it is going to be scaled
we decide to keep the result from this program .
Looking at both contours we extrapolate our shape as a compromise between this
two optimum.
Length : for the nozzle divergent section length we have, 1.5, from bell nozzle and,
6.5, for method of characteristic; as compromise we decide to realize our nozzle
with a length of 4.5 .
Contour : we took the shape from the method of characteristic and we import it in
an excel plot scaled with our new length, then we try to improve the shape designing
dierent shape (about ten), each with little dierent changes in the initial and nal
slope (some are shown in Fig.
38
3.18, using
an iterative procedure, it is possible to estimate the much number at the inlet (Minlet
= 0.2 ).
Ai
is known too and imposing the
At
equivalence of the mass ow through the throat with the one through other section
Then, because the geometry is known, the ratio
3.19.
m
t=m
i
we have
p
t0
RTt0 (1 +
Mt
pi0
+1 At =
1
RTi0 (1 +
Mt2 ) 2(1)
2
Mi
+1
1
Mi2 ) 2(1)
2
Ai
(3.19)
Where "i" indicate dierent stations along nozzle length and "t" the value at the
throat .
Because through the nozzle an isentropic evolution is supposed in this equation total pressure and temperature are not going to change through dierent sections so
they can be deleted as also
Mi =
(1 +
1
Mi2 ) 2(1)
2
Ai /At
(3.20)
Then, with an iterative process, mach number can be evaluated along all the nozzle
and, because the assumption of an isentropic behaviour through it, also all thermodynamic magnitudes can be evaluated with the following equations .
pi+1 = pi (
1 + 1
Mi2
2
1
)
1
2
(1 + 2 Mi+1
)
Ti+1 = Ti (
1 + 1
Mi2
2
)
2
(1 + 1
M
)
i+1
2
39
(3.21)
(3.22)
i+1 = i (
1
1 + 1
Mi2
2
1
)
1
2
(1 + 2 Mi+1
)
(3.23)
In the pictures below their trend along the nozzle have been plotted Fig. 3.14
Fig. 3.15 Fig. 3.16 Fig. 3.17.
Mach
2.2
2
1.8
1.6
Mach
1.4
1.2
1
0.8
0.6
0.4
0.2
3
Length (inches)
40
Pressure (KPa)
250
200
p (KPa)
150
100
50
3
Length (inches)
temperature(K)
2100
2000
1900
T (K)
1800
1700
1600
1500
1400
3
Length (inches)
41
Density
0.35
0.3
rho (Kg/m )
0.25
0.2
0.15
0.1
0.05
3
Length (inches)
So their thickness is denitely going to aect the exit Mach number since
= 0.37x
Ux
) 1
5
(3.24)
( )1
Development that comes from a shape of the velocities prole proportional to
42
x5
is nominal boundary
X=P
P dx
0
P = M 4 (1 +
M 2 3.343
)
(1 + 0.128M 2 )0.822
5
Since we have an axi-symmetric geometry instead of using just P (as for at plate),
we will use P r where r is the radial distance from the axis and is parameter set
43
equal to
5
when ReX is about
4
106 .
(
M2
UX
a0
RX =
=
XM 1 +
0
5
)3
(3.25)
= 0.37XRX2
(3.26)
Where X is an equivalent at plane length, P is a function of Mach number and sux
T , a reasonable
= 0.5.
and
atures, where
To be more complete we decide to evaluate our boundary layer with both the model,
and in Fig. 3.19 results are shown . First model
Boundary layer
1.5
0.5
0.5
1.5
3
Length (inches)
44
At the end of the nozzle we are going to have a boundary layer thickness of
about 0.189", that means that the 12.6 % of the exit area is going to occupied by
the boundary layer.
Moreover also a structural analysis has been done to consider all kind of failure in
our project.
In order to do that forces on the walls due to the main ow in the
nozzle have been calculated integrating the pressure on the inner surface of the nozzle . These forces are plotted has vector on the nozzle walls in the follow picture
Fig. 3.16).
In this picture the biggest arrow represent a force per unit area of
257P a/m2
while
2.5
R (inches)
1.5
0.5
3
z (inches)
28P a2 ,
that the highest force acting on it is going to be about 2.5 MPa, value that is far
from the limit stress of the material that we are considering for our piece (titanium
and stainless steel). Looking at results we can say that we are not going to have
structural problems due to forces acting on nozzle walls because they are really little.
45
a given region over time. We model the transient heat conduction through nozzle
walls discretizing the following equation ( 3.27) with nite dierences method, using
approximation to rst order both in space and in time.
1
T
kr
r r
r
Assuming homogeneous material
T
T
+
k
= cP
z
z
t
and k constant after a few
1 T
2T
2T
1 T
+ 2 + 2 =
r r
r
z
t
(3.27)
passages we get :
(3.28)
k
.
CP
In the equation 3.28 the derivatives of rst and second order are discretized as follow.
Where
Ti,jn+1 Ti,jn
T
=
t
t
T
Ti+1,j Ti,j
=
r
r
2T
Ti1,j 2Ti,j + Ti+1,j
=
2
r
(r)2
Ti,j1 2Ti,j + Ti,j+1
2T
=
2
z
(z)2
Replacing these in eq. 3.28 we can calculate the value of temperature at dierent
time step.
In order to estimate the heat to the wall from the main ow we have to consider
all the three dierent ways of heat transfer : conduction, convection and radiation.
During our analysis we will assume that we have adiabatic walls (no heat transfer
46
qr
kT
=
dAr
r
Where
dAr = 2rdl
(3.29)
so:
qr = 2krdl
T
r
(3.30)
(3.31)
0.026 0.2 cP
hg =
D0.2 P r0.6
)(
p0
c
)0.8 (
D
rc
)0.1 ](
A
A
)0.9
(3.32)
Where subscript 0 represent stagnation conditions and * stand fore throat dimensions:
- Pr is Prandtl number
p0 A gc
- c =
, the characteristic velocity
(m)
- rc is throat radius of curvature in a plane which contains the nozzle axis.
i = [
[
][2] TTwh
1
0g
1+
)
1
Mi2
2
1
2
]0.80.2 (
1+
1
Mi2
2
)0.2
- Twh = Tw1
- Mi local Mach number
-
relation
Third, using Stefan-Boltzmann law for a grey body, that calculate the total energy radiated per unit surface area of a black body in unit time we have:
qrad =
qr
= Tw4 Ts4
dAr
47
(3.33)
Where
temperature .
In the picture below a transient simulation (running for 15 seconds) with titanium walls, 0.25" thick, is shown g. 3.21.
As expected peak temperature has been reached at the throat and it's about 986
So in our case happens that we have the same heat ux for zirconia and titanium
but, since they have dierent thermal conductivity, low for the rst and high for
the second, they will reach dierent temperatures. In fact for zirconia we are going
to have a little amount of heat from the ow (low k ) so wall temperature will be
close to the main ow one but then, because the temperature gradient is really high,
temperature through the wall is going to decrease quickly to lower values.
For titanium is going to happens the opposite : there is going to be a big amount
48
Figure 3.22. Variation of temperature distribution in nozzle walls during experiment time.
of heat from the ow (high k ) so on the wall temperature is going to be lower but
then we will have a low temperature gradient so through the wall temperature will
decrease slowly.
Then because the dierence between wall temperature (Tw ) and gas temperature
(T ) is going to be lower for ceramics, that means that the heat ux (and so heat
losses) are going to be lower too ( 3.31).
However looking both at costs and machining issues we decide that titanium thin
wall would be better instead of ceramic ones.
However running simulations with stainless steel with thickness wall equal or more
than 1", it comes out that also stainless steel will not reach the melting point and
it shows a good behaviour for heat losses too so we decide to cut our nozzle from a
piece of stainless steel 304 with 4.5" as outside diameter (minimum wall thickness
0.75 ").
With this conguration stagnation temperature loss from steam is minimal and is
about 80 K .
49
Then we made a thermal analysis of that model. We sat material properties : density, thermal conductivity and specic heat of stainless steel 304 and then we give
boundary conditions (BC).
We run two dierent simulations: one to validate the Matlab code and one try to
see what could happens in the worst experiment conditions.
So we use the following boundary conditions :
- insulated wall from the outside,
- preheated wall at 400 K ,
- bulk temperature 2000 K ,
W
.
m2 K
These two last assumptions are conservative since we assume constant bulk tem-
perature and lm coecient when in the real its change along the nozzle.
They
variation for the temperature is from 2100 K to 1650 K and for the lm coecient
50
from 600
1000 K Fig. 3.24) . We have a dierent between Matlab and Ansys of about dierence of 1.4% due to the fact that we have assumed constant quantities that in the
Matlab code are not .
51
These analysis have been fundamental to decide the most appropriate material
to use for our nozzle, combining high temperature, cost and machine requirements.
governing equations of continuity, momentum, and energy and species transport are
solved simultaneously (that means that coupled solution is enabled ).
Since, has we have seen before, our ow is no more laminar turbulence model has
been enabled. Turbulence is a really complicated phenomena, it could be thought
of as instability of laminar ow that occurs at high Reynolds numbers (Re). Such
instabilities origin form interactions between non-linear inertial terms and viscous
terms in N-S equation. These interactions are rotational, fully time-dependent and
52
linear
eddy viscosity models, non-linear eddy viscosity models and Reynolds stress model
(RSM). The linear eddy viscosity ones are turbulence models in which the Reynolds
stresses, as obtained from a Reynolds averaging of the Navier-Stokes equations, are
modelled by a linear constitutive relationship.
There are several subcategories for the linear eddy-viscosity models, depending on
the number of (transport) equations solved for to compute the eddy viscosity coefcient: algebraic model, one equation models and two equation models.
Algebraic turbulence models or zero-equation turbulence models are models that
do not require the solution of any additional equations, and are calculated directly
from the ow variables. As a consequence, zero equation models may not be able
to properly account for history eects on the turbulence, such as convection and
diusion of turbulent energy (too simple for use in general situations, but can be
quite useful for simpler ow geometries or in start-up situations).
One equation turbulence models solve one turbulent transport equation, usually the
turbulent kinetic energy. They are: Prandtl's one-equation model, Baldwin-Barth
model and Spalart-Allmaras model.
Two equation turbulence models are one of the most common type of turbulence
models. Models like the k-epsilon model and the k-omega model have become industry standard models and are commonly used for most types of engineering problems.
Two equation turbulence models are also very much still an active area of research
and new rened two-equation models are still being developed.
By denition, two equation models include two extra transport equations to represent the turbulent properties of the ow.
account for history eects like convection and diusion of turbulent energy.
53
Most often one of the transported variables is the turbulent kinetic energy,
k.
The
The second variable can be thought of as the variable that determines the scale of
the turbulence (length-scale or time-scale), whereas the rst variable,
k,
determines
k-
class of models and has become the workhorse of practical engineering ow calculations in the time since it was proposed by Launder and Spalding [6.9]. Robustness,
economy, and reasonable accuracy for a wide range of turbulent ows explain its
popularity in industrial ow and heat transfer simulations.
It is a semi-empirical
model, and the derivation of the model equations relies on phenomenological considerations and empiricism. This semi-empirical model is based on model transport
equations for the turbulence kinetic energy (k ) and its dissipation rate ().
model transport equation for
transport equation for
The
Tmin,1 T Tmax,1 : CP = A1 + A2 T + A3 T 2 + ..
Tmin,2 T Tmax,2 : CP = B1 + B2 T + B3 T 2 + ..
Where the two dierent temperature range are 300-1000 and 1000-5000, while the
coecients Ai and Bi are the default values.
The density, since we are going to have a supersonic ow, is set to be evaluated with
54
pop + p
R
T
Mw
Where p is the local relative (or gauge) pressure predicted by FLUENT, pop is the
operating pressure and Mw is the molecular weight.
And viscosity, that change with temperature too, will follow Sutherland law. Sutherland's viscosity law resulted from a kinetic theory by Sutherland (1893) using an
idealized intermolecular-force potential. The formula is specied using two or three
coecients, we choose two coecient equation show below .
3
C1 T 2
=
T + C2
As boundary condition we set : "pressure inlet" type for inlet surface and for the
exit one "pressure outlet" type.
Pressure inlet b.c. :
- stagnation pressure 250000 Pa,
- static pressure 244180 Pa (evaluated from the denition of total pressure since
the Mach number at the enter of the nozzle is known),
- total temperature 2100 K .
Pressure outlet b.c. :
- static pressure 28414 Pa,
- total back ow temperature 1500 K.
For the outlet boundary conditions values come from Matlab quasi 1D program previously described.
To solve the equations, since we were just modelling the nozzle ow eld for a
rst check, we decide to use an explicit scheme that provide a less expensive solution
both in terms of time and computational costs.
Between the discretization schemes available for the pressure, momentum, and energy equations, rst order upwind method had been chosen. Same scheme has been
set for convection-diusion equations discretization.
Then, to start the solution calculation, the last parameter to set is Courant number.
For ANSYS FLUENT's density-based solver, the main control over the time-stepping
scheme is the Courant number (CFL), the time step is proportional to the CFL as
follow.
2CF LV
t = max
Af
f f
55
Af
max
f
shows that the maximum allowable CFL for the multi-stage scheme used in the
density-based explicit formulation will depend on the number of stages used and
how often the dissipation and viscous terms are updated . But in general, you can
assume that the multi-stage scheme is stable for Courant numbers up to 2.5. This
stability limit is often lower in practice because of non-linearities in the governing
equations.
The stability limits of the density-based implicit and explicit formulations are signicantly dierent. The explicit formulation has a more limited range and requires
lower CFL settings than does the density-based implicit formulation.
Since we are using an explicit formulation we set the Courant number equal to 0.95
.
With these settings the simulation have been ran until the convergence, in terms of
6
energy and continuity residuals ( 10 ), have been achieved .
We decide to run a simulation also with the model of the nozzle that we get from
bell-approximation method in order to verify that with our optimized contour shape,
oblique shocks at the exit are going to be weaker and we are going to have less total
pressure losses than if we use bell contour one (results are shown in g. 3.25 and
g. 3.26 )
Looking at results we noticed that, as predicted, pretty strong oblique shocks were
coming from the diverging section of the bell model while, in the optimized shape
one, much more weaker shocks develop . This happens because our shape has not
the ideal length, a shorter shape led inevitably to a curvature at the exit that's a
bit still too sharp.
56
Figure 3.25. Comparison of Mach prole in bell model (on the top) and in the
optimized shape (in the bottom).
57
Figure 3.26. Comparison of total pressure prole in bell model (on the top) and
in the optimized shape (in the bottom).
58
Chapter 4
Injectors design
One of the most "critical" section in this project has been the injection one. In fact,
after heating to really high temperature (2100 K) and accelerating to supersonic velocity, we have to inject a non-neglectable amount of methane in the main mass ow.
The diculty in this design is that we have to achieve particular ow conditions to
make the pyrolysis happens but we have also some uncomfortable restrictions. Our
goals are:
- inject as far as the duct centreline ;
- cover the maximum traversal area with the jet "cloud" ;
- achieve complete thermal and molecular mixing in the shortest length (because
of the heat losses) within millisecond residence times;
- supersonic Mach at the end of injection section (
1.5);
59
4 Injectors design
losses, because they can compromise our ability to have still a supersonic ow. In
fact in our pipe the only driving force is the pressure dierence between upstream
5
(2.5 10 Pa) and downstream (dump tank 20KPa), so if upstream we will have
too many losses pressure will decrease too much we will have the normal shock too
early in the duct.
The use of
60
4 Injectors design
Experiments and numerical research [6.10] show that increased compressibility reduce mixing layer growth rates and reduce mixing, in fact it reorganizes the turbulence eld and modies the development of turbulent structures. Several phenomena
result in the reduction of mixing with increasing ow velocity, including the velocity
dierential between injected oe and main one . The primary physical mechanism
by which mixing occurs in jets injected into supersonic ow are compressible shearmixing layers, eddies in the shear layers entrain injectant into the main ow and
increase concentration gradients. Experiments have shown also that the shocks that
would result from wall and strut compression appear to enhance the growth of the
two-dimensional eddy structure (rollers) of a mixing layer.
In conclusion mixing in supersonic main ow take more space than in subsonic conditions, the structure that develops from the injector is really complicated and cause
a lot of pressure loss but on the other hand, the complex structure of shock waves,
contribute in mixing enhancement, achieved with a characteristic vortical ow structure arising from centre axis. This is explained by two kinds of ow properties, a
pushing away force caused by high momentum air stream behind shock plane, and
Baroclinic Torque generated from spatial mis-alignment between density gradient of
mixing layer and pressure one of shock plane. The former spread mixing layer to
radial direction with ow direction compression and the latter rolls up the spreading
portion of mixing layer.
[6.12]).
v
u
u
J =t
(U 2 )jet
(U 2 )freestream
(4.1)
61
4 Injectors design
where strong pressure and density gradients exist, baroclinic torques can introduce additional vorticity.
- Circular, elliptic and diamond shape for injection port. Looking at the performance it seems the the second and the third shape bring some improvements
like spreading the injected ow more and more quickly in the lateral direction
and for the third also achieving more penetration and mixing faster. Unfortunately they all present more dicult in machining and, in particular, it's very
dicult to choke their geometries and contoll the mass ow through them .
- wept and unswept ramps: injecting from the base. With small injection angles
they can produce near streamwise injection. Ramps cause streamwise vorticity
and enhanced mixing. Swept ramps create more vorticity than unswept ramps.
- In-stream injectors (Struts): mainly desirable due to streamwise injection capability which reduces stagnation pressure and thrust loss.
- Pulsed injectors: mechanically complex for SWR (moving rotary valves with
similar motions to those in a rotary engine) but mixing is improved due to
additional shear layer area.
- Aero-ramp injector: arrays of injector nozzles at various inclinations to approximate the physical swept ramp design.
Summarizing we can identify three main group of injection scheme : wall injection
(downstream-facing step in combustion chamber) which involves diusive process,
central injection (injection from strut centrally located in the chamber) that combine both diusive and penetrating system and port injection (injection from holes
in combustion chamber walls) that uses a penetrating mechanism.
About the rst conguration that presents two main problems for us: it leads to
longer sections that means more heat losses but, more than that, is that considering
a step conguration means a reduction in nozzle exit area (that is means reduction
in Mach number at the exit) and bigger area to cover with the jet ( no good because
we have a limited injecting pressure, so limit in penetration power).
Moreover also the second scheme couldn't work for our project due to the very high
temperature in the main ow that will melt the centre strut.
We decide then, to focus on the penetrating mechanism. At the beginning we decide to start from the most simple design (transverse injection) to understand the
complex dynamics of an injected ow and discover the relative weaknesses.
In the following picture g. 4.1 a scheme of transverse injection oweld has been
shown.
Looking at the g. 4.1 we can see how the principal feature in the secondary jet is
the barrel shock structure which terminates in a Mach disk and is followed by a less
clearly dened expansion zone.
62
4 Injectors design
Figure 4.1. Schematic of typical ow pattern at the injection port of an underexpand jet.
Our rst issue was nd a model able to predict which was the penetration high of
a transverse injection with our characteristics in terms of : main ow velocity (Uf ),
injection ow velocity (Uj ), injection pressure (pj ) and mass ow (m
j ). This rst
analysis was necessary to understand with which velocity was better inject methane.
Looking at literature [6.11] has been shown that penetration high is really inuenced
by
(eq. 4.1) and, in particular, we have seen that ow injected with supersonic
the mixing will take more time to develop completely (with low
injec-
tion ow structure become steady more quickly). Moreover have been proved that
injecting under-expanded jet give advantages both for penetration and mixing enhancement; so we decide to inject methane in sonic conditions for the reason written
and also because chocking methane exit area we can control the mass ow.
In the past several models to predict penetration high of a transverse injection
have been made. Since they led to results that were dierent from each other in a
non-negletable way we decided to consider all of them and make our nal decision
comparing the all.
63
4 Injectors design
h
h
p j j
=
J,Mf ,Mj , ,
d
d
p 2 2
)
(4.2)
h
is the penetration distance scaled by the jet-exit diameter. It is important
d
to establish the interdependence of certain variables in eq.( 4.2). For compressible
Where
J=
j pj Mj2
f pf Mf2
p2
pf
p 2 = pf
Which for
Mf 2
and
(4.3)
2f Mf2 (f 1)
f + 1
(4.4)
2f Mf2
p 2 pf
f + 1
This enables us to relate
to
p2
(4.5)
h
h
2f pj 2
p j j
=
J
Mj ,Mf ,Mj , ,
d
d
f + 1 p2
p2 2
)
(4.6)
pj
, the momentum ratio is very strong function of Mj .
p2
pj
For xed J , one has no choice but to decrease
while increasing Mj , and vicep2
pj
versa. This make it dicult to distinguish between the individual eects of
and
p2
Mj . Eq. 4.6 shows that the cross-ow Mach number has very weak inuence on J ,
It is evident that, for given
therefore Mf can be varied by a large extent while keeping all other variables virtually xed. The same applies to density ratio, since
It's important to remind that all the model proposed have been
64
4 Injectors design
extrapolated by experiments that were showing similar conditions with our project
in terms of : free stream Mach number, injection pressure, injection velocity and
momentum ratio.
1) Model by F.W. Spaid [6.13].
Here the eective shape of the injectant is represented as a quarter of a sphere of
radius
Figure 4.2. Spaid's model of the eective shape of the injectant causing body
interaction in the free stream.
Spaid gave the following equation to predict the penetration high that the secondary
ow will reach .
h
=
c
{(
1
M
)(
Poj j 2
P CP
)[
2
2
j 1 j + 1
)(j +1)/(j 1) (
P
1
Poj
Where
the stagnation pressure behind a normal shock, found from theory of supersonic ow
of an ideal gas:
CP
2
=
2
M
[(
+ 1 2
M
2
65
+ 1
2 +1
2 M
1
1
4 Injectors design
y
de J
de
= 1.23
)0.344
(4.8)
de J
( )
1
pj 2
y
4
=
M
j
Dj
peb
1
Where
(4.9)
[6.16].
Looking to previous work [6.15] a solid body drag model was introduced and was
used to calculate the trajectory of either a matched pressure or over-expanded jet
and the portion of jet trajectory downstream of the Mach disk in an under-expanded
jet. The development of this model rests on the similarity that exists between a jet
discharging into a quiescent medium and jet discharging into crossow., in fact the
66
4 Injectors design
Figure 4.3. Billing's new model of the eective shape of the injectant causing body
interaction in the free stream.
At the end, looking to experimental data they achieve a new expression for maximum ordinate Mach disk that result as a correction of the previous one given by
Billig and Schetz (eq. 4.10).
( )1
1.8[1 0.7exp(0.07 pebj )] ( Dj )
2
y
1/4 pj
+
=
M
1
j
a
Dj
peb
{1 + 0.16[1 exp( M
)]2 } 2 Dj
Mj
p
(4.10)
Comparisons between this equation, Spaid's model and experimental data have been
shown in gures below (g. 4.4).
In these pictures eq.44 stands for eq. 4.10 .
To dene the trajectory of the Mach disk centreline in both models
67
and
4,
the
4 Injectors design
Figure 4.4. Comparison of calculated jet cross-sectional areas and measured jet
concentration contours (on the left). Comparison of calculated values of the
maximum ordinate of the Mach disk (on the right).
trajectory shape is assumed to be parabolic in the region after the injection using
the following equation.
y
=
Dj
Where
[(
x
Dj
)(
y1
Dj
)(
x1
y1
)]
y1
(4.11)
is the nor-
x1
Mf
x1
= 1.25[1 exp
]
y1
Mj
5)Model by L.S. Cohen and L.J. Coulter [6.17].
Ground on experimental data they suggest the following behaviour :
poj
H = 0.6455D
peb
Where
poj
)1
2
(4.12)
[6.18].
Because the transverse jet interaction ow eld contains shock waves, ow recirculation, and separation zones, which are often three-dimensional, completely analytical
solution require the use of full elliptic form of the partial dierential equations of
68
4 Injectors design
motion. In this paper Roger wants to analyse this phenomena to obtain the ow
properties that can be input into a detailed computation of a supersonic combustion
ow. Considering all the data available he assumed that the ow in the vicinity of a
two dimensional injector can be represented by the 2 D model presented if g. 4.5.
The ow is conned in a duct with walls that diverge at small angles
The secondary ow is injected at an angle
and
B .
wave is assumed to be a linear combination of the separation and bow shocks. This
means that, at large distances from the injector, the turning imparted to the ow
is equal to that of a wedge with the same shock angle as the upstream separation
region.
Unlike other models a distinction between the location of the local peak
concentration
zone
zk
(represented by the top of the Mach disk) and the of the mixing
zk
f Mf2 pf
J
= 1.5
d
j Mj2 peb
For
)1
2
(4.13)
with dierent ow conditions and at the end he said that, since there are no reason
to justify the use of one over the other, the edge of the mixing region is taken as an
average value given by equations 4.14 and 4.15: . The rst is obtained for Mf =2.72.
z
Mf
= 1.68
d
Mj
)1
69
1
2
( )0.0866
x
d
(4.14)
4 Injectors design
Where
)0.143
(4.15)
equal to the one just at the exit of the jet port, instead of injecting under-expanded
jet, would give an enhancement in penetration because this second jet (as well as
over-expanded) generates shock waves in the jet ow leading to creation of Mach
disk, which renders the jet ow subsonic.
cross ow, the static pressure around the jet varies greatly along its circumference
as well as along its trajectory, so it is dicult to dene pressure matching.
The
computation of Heister and Karagozian [6.20] showed that equivalent pressure (the
static pressure averaged around jet perimeter) is approximately 0.5 p2 for the range
of
Mf
considered in that paper, where p2 is the pressure after the bow shock .
Mf
modestly improves
penetration, while the jet Mach number and density ratio have no noticeable eect.
Highly under-expanded jets achieve about the same penetration as pressure-matched
jets.
Papamoschou said also that penetration height increases with
in a fashion similar
to that found in subsonic ows, so, looking to the equation founded experimentally
by Pratte and Baines [6.21] that model the development of the upper edge of a
subsonic transverse jet (Eq.
(eq.
4.17).
y
x
= 2.63
rd
rd
)0.28
h
= 4.34J 0.36
d
(4.16)
(4.17)
Where h is dened here as the maximum height of the jet trajectory. Moreover he
x
= 6.
shows that jet trajectory is seen to level o at
d
However looking at g. 4.6 and comparing Papamoschou "equivalent subsonic"
70
4 Injectors design
model with eperiment results we can see that it usually overstime the real penetration high and, precisely, the error is about
+43%
when
J =2
and
+9.7%
when
J = 3.5.
71
4 Injectors design
Table 4.1. Below the value of dierent parameters that describe our condition
p
(kPa)
(K)
R
J
)
KgK
Kg
)
s
Jet
1034.214
850
1.13
519
0.1
Main Stream
250
2.12
2100
1.19
470
0.35
With these values it comes out that when we inject we will have a J
1.96.
To compare to each other these dierent models we need to know the ow eld (pressure, temperature, density and velocity) along the duct but since dierent shocks
are going to develop and reect front the jet port, this would be very dicult, so
we decide to plot the penetration high as we were injecting at the beginning of the
nozzle so we can have an idea of the jet high at the exit that (g. 4.8 ).
pj dj 1500.24
Spaid
Gruber
Billig and Schetz
BilligOrthLasky
Rogers zdelta
1.5
Rogers zk
Papamoschou
0.5
0.5
1.5
3
J =1.9644
72
4 Injectors design
Since we know, with a good condence, which is the error from Papamoschou's
approximation for two dierent value of
behaviour for these two values of momentum ratio and, looking at the error between
them and Papamoschou we will be able to dene the best jet-model.
Spaid
Gruber
Billig and Schetz
BilligOrthLasky
Rogers zdelta
pj dj 3500.24
Rogers zk
1.5
Papamoschou
1
0.5
0.5
1.5
3
J =3.0006
Figure 4.8. Jet penetration development according to the 7 models described for
J=3.
Looking at the plot we can see that when J=3 there are three models at about
9.7%
and
zk )
when J3 we can see that, between these three approximations, the one that is also
about
43%
73
number of injectors
4 Injectors design
R of
(Aji ).
injected ow, we can calculate with eq. 3.18 the mass ow through that area
Kg
Then we subtract this value from the total mass ow of methane (0.1
) and we
s
divide that for the total number of injectors, n minus one. After, inverting eq. 3.18,
we calculate the diameter of the other ports. We place these calculations in a loop
that will stop when the diameter calculated at the end will be equal to the one insert
for rst.
Using this program we found that if we want 4 equal injectors each port would be of
0.223" . However, using Spaid's model we can see that with our injecting pressure,
transverse jets would not achieve the centreline (g. 4.9).
pj 150 dj 0.25
1.5
0.5
0.5
1.5
3
J =1.9644
dj
and
m
j,
we will achieve
good idea because that means increase the strength of the bow shock in front of it
and reduce the number of injectors will let to a less ecient covering frontal area
(see subsection 5.5.1).
Considering that we will need to enhance the penetration, we decide to use a n.
Looking at literature we found several papers about the use of ns in injection in
supersonic ows. It comes out that the n served two main purposes : enhance mixing, improve penetration high and minimize the stagnation pressure loss associated
74
4 Injectors design
with fuel injection process. In particular driving dierent experiments we notice that
the reduction in shock-caused pressure loss would be around 35%. In fact typical
transverse injection causes deection in supersonic airow by setting up an oblique
shock. Potential amount of stagnation pressure loss due to oblique shock can vary
widely between a small fraction of a percent and tens of percent so a n that would
create a weaker oblique shock could led to big improvements in the injecting phase.
When a n is placed before a port the penetration into the free stream is guided
by the n wake, the jet is pushed to the top of the n relatively unobstructed, then
the penetration beyond the n heigh appears to follow a similar trajectory as in the
baseline case (transverse injection without n) [6.22]. Moreover the ow injected is
delivered to the core of the supersonic ow rather than staying in the boundary layer
as in the baseline case. From experimental study results clearly that the location
of the secondary ow delivery into the supersonic ow can be controlled by the n
heigh. In particular comes out that at 25 jet diameter downstream the penetration
is 1.5 times the high of the n [6.23] g. 4.10.
Figure 4.10. Comparison between instantaneous images of baseline case fuel injection (top gure) and n-guided one at various axial distances downstream.
The n employed in these study was a triangular blade with a rectangular base
characterized by the following ratios :
length/d =
10 (l/d) g.
high/d =
3 (h/d),
width/d =1 (w/d)
and
To nd the best geometry, in terms of length-high to width ratio we refers to another
study [6.23] where an analysis over the improvment derived from dierent proportions has been made. Three dierent pylons with optimal heights, widths and pylon
75
4 Injectors design
Figure 4.11. Pylon geometry shown with injection port and dened axis system.
distances from injection coming from a previous computational research [6.24], have
been studied.
XP and W . Two pylons in this experiment were designed using the optimal
case ( XP /d 2 and W/d 1). One is named medium pylon (h/d=4 , w/d=1.12
, l/d=6.9) and one is named tall (h/d=6 , w/d=1.12 , l/d=10.4). The third pylon
uses the second best case ( XP /d 3 and W/d 1.5) and is named wide (h/d=4,
w/d=1.6 , l/d=6.9). Inclination angle ( ) of each pylon is derived from l , h, and w .
were
Observing the self normalized value for penetration (yj /hn ) and for spread in the
with direction results shows that best congurations are medium and wide one (2025% better).
The wide pylon is the most eective lifter, the tall pylon loses its
values. Tall pylon is no good because of pressure loss and strength of the
shock that is going to produce. All pylons provided more penetration and less lateral
spread than a at (no n) injection. Looking at results the wide and the medium, in
second place, come to be the best performer overall. The medium lifted the fuel jet
away from the boundary layer quickly and established a good penetration height.
However, when we decide our n geometry, we took in consideration the medium
model and not the wide one because we have to consider that we are in a cylindrical
section (so we should have a n as thin as possible) and also that the strength of
the oblique shock is directly proportional to the inclination angle of the ramp, so at
the end we choose a ramp with these ratios : h/d=3 , w/d=1.12 and l/d=10. We
took a shorter n for two main reason : rst, because we want a smaller angle at
the wedge and second because, looking at the improvement brought by a n, this
high would be enough for reach a satisfactory penetration in our case .
76
4 Injectors design
Figure 4.12. Transverse Section through SWR duct showing methane distribution
at 12 injector port diameters downstream of injection. (approximately to scale).
Once we have chosen the geometry we have to choose the material to realize that
n. We decide to make a thermal analysis of the n with Ansys using ve dierent
material : titanium, zirconia, alumina and copper.
As for the nozzle analysis we made and discretize a solid model rst, with tetrahedral solid element and then, we drive a thermal transient analysis (for 15 seconds) .
As boundary conditions we set for
3/4
rest insulation. Then we set as bulk temperature 2100 K and uniform temperature
at the bottom of the n equal to 500 K . From these analysis comes out that a n
in all materials mentioned before, except zirconia, were going to melt (g. 4.13) .
Looking at the results appears that the only material able to resist in our experiment environment was a n made with zirconia, however this solution will present
some issues as realize the attach between zirconia n wall and stainless steel tube
walls, big eect of blockage caused by the n encumbrance ((n frontal area equal
to 30% of the transverse duct section) and the main problem would be when this
77
4 Injectors design
Figure 4.13. Thermal analysis for a n made in titanium. After 15 sec it has
reached the melting temperature (1941 K).
into supersonic cross ow has been studied for years, recently there have been many
research activities [6.25] and [6.26] to investigate mixing characteristics of multiple
transverse injection system or aerodinamic ramp injection systems. Experimental
and computational studies [6.27] have reported that an aerodinamic ramp injection
have better mixing characteristics than a single transverse one. However complication for an ecient injection system can arise due to many other parameters that
in this system need to be set as: position of each injectors, distribution of mass ow
rate and momentum ux, injection angle and combination of injection angles.
78
4 Injectors design
The ow eld of dual transverse injection into a supersonic cross ow is very complex due to various shock structures and vortical ows around injection ows, Fig.
4.14 shows schematic of that system.
Figure 4.14. Schematic view of mean ow eld of dual transverse injection.
We can also see that the strengths of the shock wave ahead of the rear injection is
weaker than near the front injection ows, which implied that the front injection
ows blocked the main ow ahead of injection ows and assisted the rear injection
ows to expand more strongly. There are many shock waves: three dimensional bow
shocks formed ahead of the front jet and ahead of the rear jet, a separation shock
generated by interaction between the front bow shock and boundary layer and Mach
discs. A distinctive feature of the dual injection is that the Mach disc of the rear
jet is located at higher position from the wall and is larger than the one of the front
jet (g. 4.15 ).
The front jet blocks the supersonic air ow blowing towards the rear jet leading
to a reduction of the momentum of the main ow . Thus the eective jet-to-cross
ow momentum ux ratio of the rear injection ow increased, which assists the rear
injection ow stronger expansion. This blocking of main ow would have signicant
inuences on the mixing process. As we can see there are also many vortical features: vortex pairs formed along the jet ows, horse shoe vortices, separation bubble
and recirculating wake ows. Streamwise vortices roll up injection ows and thus
accelerate mixing in supersonic ow elds.
In these comparison the mass ow injected is the same because the diameters of the
aeroramp are such as the sum of the mass ow through both of them is equal to the
mass ow through a single injector.
79
4 Injectors design
Figure 4.15. Comparison between single and dual transverse injection system for
penetration high (left) and Mach disc position (right).
Looking to g. 4.15 we can observe that the heigh of the Mach disc of the rear injection ow is strongly related with the distance between injectors, while the heigh
of the Mach disc of the front injection ow is mantained regardless of this distance.
The Mach disc heigh of the front injector in the dual injection system is smaller
than that of the single injection system because the diameter of injection hole of the
dual injection is smaller than that of the single injection system. It is shown that
the mixing characteristics of dual transverse injection system are very dierent from
those of a single injection system. This system is also called "aeroramp" because
the front port act as a ramp and project the jet coming from the rear port at higher
penetration, in fact the rear injection ow is strongly inuenced by blockage eects
due to momentum ux of the front injection ow and thus has higher mixing rate
and a higher penetration but have also more losses of stagnation pressure than a
single injection system.
tance between injectors increased until a critical distance but then decreased after
that critical distance. Thus there existed an optimal distance between injectors for
mixing characteristics, which coincided with the optimal distance for the blockage
eects. The optimal distance between injectors for mixing characteristics increased
as the magnitude of
increased.
A study to found the best distance between two port in a dual transverse injection
system has been made too [6.27]. From experiments we can see that the strength of
the shock wave near the front of the rear injection ows increased as the distance
80
4 Injectors design
between injectors decrease. In the paper [6.27] from Sang-Hyeon Lee, four dierent
injectors distances have been considered : 2, 6, 8 and 14 jet diameters. The experiment was drove for two dierent value of momentum ratio values (1 and 2) but we
are going to consider just the second one since it reects our case. In this study it
is shown that for a J=2 the best distance between front and rear injector is 8
while the one that give less stagnation pressure loss is 2
dj
dj ,
(Fig. 4.16 ).
It comes out that if you want to achieve a compromise between these two optimal
the best solution would be place the two injector 6
dj
Looking at results we notice that stream vorticity, or vortex pair, produces a large
convection ow in the plane perpendicular to the streamwise direction and promotes
mixing and a higher mixing rate largely depend on a higher streamwise vorticity.
Mixing rate are closely related to the magnitude of
nitude of
J.
J.
This is due
to the fact that a smaller magnitude of J has a weaker inertia of jet ow and thus
the airow could more easily disturb or stir the jet ow with a lower magnitude of
J.
As we have seen the dual injection has a higher penetration than that of the single
injection. The improvement in penetration is about 1.5 times the high with a single
transverse injection, while the lateral spread will slightly decrease. The penetration
distances are closely related to the magnitude of
tum ratio has an higher penetration than with
the fact that the jet ow with higher
81
4 Injectors design
In general, the dual injection systems have more losses of stagnation pressure than
the single injection system but these are not so great considering the mixing enhancement that this conguration brings.
82
Chapter 5
Our injecting conguration
Finally, after considering all the possible injection solutions, we decide to use the
aeroramp. However to decide the number of injectors and their position we should
make another analysis. In fact in order to consider all the aspect that will let to
the best mixing we should not just look at the penetration high but, also, at the
lateral spread and at the transversal area occupied by the jet. This second parameter is going to be very important to dene if we are going to have an ecient mixing.
z/y
z/y
is about 1.4. Then, looking at the same plot but for J=3, we achieve
is 1.29 so, making an average between the two, we can say that when J=2
the lateral sprea is approximately 1.3 times the penetration high of a transverse
injection.
83
Figure 5.1. Comparison between lateral spread and penetration high between
dierent congurations.
These data have been conrmed also from pictures and plot from others experiments [6.16], [6.28], [6.29] and [6.30].
However we have to remember than injecting with n and aeroramp improve high
penetration but also decrease lateral spread. So, if the plume for a single transverse
injection without n, about 10 diameter downstream, can look as a rectangle (with
the longer side on the bottom), when we use an aeroramp it will look more as a
rectangle but with the longer side on the lateral part (g. 5.2 ).
84
Looking at the picture we can see that even if the rectangular approximation seems
good this usually overestimate the eective area covered (Ac ) by the jet we decide to
take in consideration other approximations suggested by papers previously quoted.
In fact in literature dierent models have been suggested : quarter of a sphere, a
full sphere, an ellipse and rectangle. In the following pages we will introduce briey
these model and the equation used to estimate that area and then we will choose
the best approximation.
1) From Spaid.
He assume that the jet is going to have a shape of a quarter of a sphere, where the
radius is equal to the peak of concentration value (eq. 5.1).
21/2 h
Ac
=
Dj
Aj
1 pj j 4
2
2
Mf pf f CP
j 1 j + 1
](j +1)/(j 1) [
pf
1
pj
Ac
Aj
)1
2
h
Dj
)
(5.2)
Ac
Aj
)1
h z
=
2 2
(5.3)
Where
Ac
Aj
)1
2
= (h z)
(5.4)
best two models are the last two but since the elliptical model will underestimate
the eective area and the rectangular one will overestimate it we decide to take an
average between this two g. 5.3.
85
Spaid model&high1
A ellipt2
Spherical model3
Rectangular model4
Average model6
Covered area
3.5
Area (inches2)
2.5
1.5
0.5
3
4
Pressure150psia
86
avoid because it would cause a non-homogeneous mixing) with just one injector;
- total area covered, that represent the area covered by all the jet plume. We
look at this value because the mixing eciency won't be just aected by the
penetration high but also from how well we will be able to spread the methane
in the all duct;
- mass ow of main injector, that means the % of the total mass ow (0.1
Kg/s)
that is going to pass through the main port, for congurations where is foresee
a main injector, or through each port for the conguration where all jets have
same dimensions. This last parameter will be very important to have an idea
of the strength of the shock that will develop in front of the port.
After these consideration we decide to analyse six dierent congurations that we
can divide in three conguration with one main and some secondary jets (two, three
and four) and three with all the jets af the same size (three, four and ve in the
total).
Looking at the three congurations with all the port of the same dimension we can
see dierent trends in the plots below (g. 5.4 and g. 5.5) where the forur paramenters listed before are plotted compared with dierent values of jet port.
Figure 5.4. Characteristics of conguration with 5 (left) and 4 (right) equal ports
.
87
We observe that penetration high and the % of the total diameter covered increase
when number of injectors decrease, while, the area covered increase and the mass
ow through the port decrease when we have more jets.
scheme with three and with ve injectors presents two over four best situations but,
still, having all jets of the same size doesn't seem to be an optimal solution because,e
even if we take the scheme that present the biggest improvement of one of the four
parameters plotted, this would not be enough for our purposes. For example, considering three injectors we will have the highest penetration but that will be just
40% of the high so we won't even reach the centreline. While the scheme with ve
jets bring optimal area covering (84%) but leave a really big gap in the middle of
the section. Even looking at the conguration in the middle (with four injection)
that will present average performance this won't work for our requirements.
Since having all the injectors of the same size won't help we decide to consider congurations with one bigger injector (main) and other smaller injectors (secondary).
In the following pictures plot for dierent number of ports are presented ( 5.6 and
5.7).
The parameters are plotted compared with dierent value of the main jet port.
At rst we can notice that the conguration with one main and 3 secondary jets is
not going to work since, the maximal area covered for any jet diameter that we can
choose, would never be more than the 70% and also the penetration high would be
as maximum 40% that's not enough for us.
88
Figure 5.6. Characteristics of conguration with 5 (left) and 4 (right) equal ports
.
So we consider the last two scheme: 5 and 4 injectors in the all. We should remember that, actually, the main injector for us is going to be replace with an aeroramp
89
through which is going to pass the same mass ow just splitted in two jet. In fact
dividing the mass ow that is supposed to pass through the main in two and using
the eq. 3.18 we will found the size of front and rear aeroramp injectors. Moreover,
using an aeroramp, we will have an enhancement in the penetration high (it will be
around 1.5 times single transverse penetration ) so we can consider as good also the
conguration where an high of the 43% is reached with the main injector. Looking
at the point of optimus for both total area and total high covered we can see that for
both conguration is around 76% but, while for ve jet congurations that means
65% of the total mass ow through it (strong bow shock), for the four jets one that
means just 43% of the total mass ow. Moreover we prefer to have less secondary
injectors because we don't want the single jet cloud to interact to much with each
other.
So, looking at these considerations, we opted for a conguration with one main injector and three secondary ones. Precisely we choose a main injector with a diameter
of 0.335" that will become two diameter of 0.237" and the secondary jets will have
a diameter of 0.171".
injecting with an angle of 90 will denitely give the highest penetration but, on the
other side, will also give more stagnation pressure loss than injecting with a lower
angle (60 or 30 ), on the other hand lower angle won't reach the penetration that
we need. Knowing that we decide to use an aeroramp, a double transverse injection
that will give more penetration and less loss than a single transverse one g. 5.8 .
Still, even with a double transverse injection, pressure loss where not negletable
so we decide to incline front and rear aeroramp jet too.
shock to be as weak as possible and since, in a dual injection scheme, the jet that
characterize the strongest oblique shock is the rst injector and the one responsible
for the total penetration high is the second, we decide to incline the front jet of 30
from the horizontal and the second one of 60 . As we can see from the g. 5.9, even
inclining the jet we will have enough penetration, unlike the case when there were
a single inclined jet. In the picture below the penetration high is shown as methane
mass fraction concentration.
90
Figure 5.8. Mach contour for single transverse injection (top) and dual injection
(bottom) .
About the aeroramp distance, as said before, the optimal distance would be 6
jet diameters from each other but, since we are going to incline the jet, the blockage
eect is going to be smaller so,even if putting closer these two injectors the oblique
shock in the front would be stringer it's also true that the penetration would be
higher, so we decide to place them a bit closer: 5 dj far from each other.
The last choice that we have to make is about the angle and the position of the
secondary jets. At the beginning we tried to leave them as simple transverse injectors
91
Figure 5.9. Comparison between dierent jet angle for mass fraction of methane.
in the same plane of the main one, in order to achieve more penetration, mix quicker
and in a shorter length and thus have less heat loss. However, as we can see from the
picture 5.10 this conguration will lead to a normal shock before the injectors and
consequently subsonic ow after them. To avoid that we try to incline the secondary
jet too with an angle of 60 from horizontal, from the picture we can see that the
normal shock become an oblique-lambda one but, still too strong because after the
ow will be subsonic.
92
Figure 5.10. Mach contour comparison between transverse and oblique secondary
jet.
Because of that, since a more complete mixing can be achieved anyway even if
the secondary injectors don't reach the centreline, for example leaving a longer space
after them while, the subsonic ow can't be turned in supersonic any more because
the pressure in the receiver tank. So we decide to privilege the necessity to have less
stagnation pressure loss and after the injection still a supersonic ow. We incline
the secondary injectors of 30 from horizontal and place them a bit downstream
from the aeroramp. At the beginning we choose a distance between the centre of
93
the rear aeroramp and the centre of secondary injectors of about 5 inches.
This
distance have set at this value after an approximation estimate of the distance were
the oblique shock generate from the aeroramp will impinge on the upper wall, since
we want to place secondary jets after that so they can exploit the blockage eect
of the shock and having a bigger eective momentum ratio
J.
ulation (g. 5.8) we can see that the wave angle of the oblique shock in front the
aeroramp is about 40-45 degree. Using the fundamental relation for oblique shocks
(eq. 5.5), where
a wave angle of 45 deg correspond to a ramp angle of about 13 deg, this happens
because even if we inject with an angle of 30 deg the jet ow is instantly turned by
the supersonic main ow. So considering an oblique shock with this wave angle, and
since our section in 3" high, using simple trigonometric relations we have that the
oblique shock is going to impact on the upper wall after 4.6", so we decide to take
a little margin and place secondary injectors at 5".
M12 sin2 1
tan = 2cot
M12 ( + cos2) + 2
]
(5.5)
We will call the solution with secondary jets moved downstream "staggered injection".
However before
achieving our nal conguration, we have run several simulations in Fluent trying
to optimize several aspect of that design.
the several conguration that we test and the result that we achieved, the CFD
model, boundary conditions and type of discretization used will be explain in the
last chapter too.
We identify ve dierent aspect that could leave to improvement: distance between
end of the nozzle-front aeroramp jet, distance between front and rear aeroramp,
angle of aeroramp and secondary jets , distance between primary and secondary
injectors and shape of the section after the nozzle.
1)Distance between end of the nozzle-front aeroramp jet.
This distance was aected from three dierent needs:
- rst, we want to save as much length as possible because of the heat losses.
- Second, the distance from the end of the nozzle has to be at least 2" for
technical-machining reasons. In fact the nozzle is going to be machined in a
94
separate piece from the injector section and then they are going to be welded
together. So we need to leave a space that allow us to eventually detached the
nozzle and re-welded a new one without compromise the entire section.
- Third, this distance can't be too short because, due to the non-neglectable
thickness of boundary layers at the exit of the nozzle, if we place the injectors
too close the separation bubble that is going to form in front of the bow shock
will aect the boundary layer too, making it becomes more thick. We want
denitely to avoid that because a thicker boundary layer will reduce the eective exit area and therefore reduce the Mach at the exit of the diverging section.
As compromise between all these points we decide, as optimal distance, to take 2"
since, from simulations comes that the length was enough to don't aect to much
the exit ow eld and also, we noticed that moving the injector more downstream
was not bringing enough improvement in term of velocity and pressure losses to
justify heat penalties.
2)Distance between front and rear aeroramp.
After considerations considerations based on experiments reported in the paper
cited before, we have decided to place the dual injectors to a distance of 5dj from
each other; however, since in literature no references were founded about inclined
aeroramps, we decide to run two dierent simulation, with distances of 3dj and 7dj
to see if will be any improvement from the 5dj one.
Exactly as predict in paper [6.27] we can see (g. 5.11 and g. 5.12)that reducing the
that distance will increase the strength of the shock in front the aeroramp, reducing dramatically the main ow velocity while increasing that distance will decrease
the strength of the oblique shock and improve a bit the jet penetration. However
comparing the 7dj with the 5dj one we can see that in the second we will have less
stagnation pressure losses and also, since the rear jet is delivered at higher levels,
there is a bigger area at the bottom where the mixing doesn't occur, so since these
two are very critical points in our project, we decide to conrm 5dj as space between
the two aeroramp jets.
3)Angle of aeroramp and secondary jets.
About aeroramp conguration, to try to reduce stagnation pressure loss, we try to
incline less (50 deg instead of 60 deg) the rear jet because this should leave to a
weaker oblique shock. We didn't try to change the angle of the front jet because:
increasing it would create design problems (like machining injector with a so small
inclination) and increasing it would just let to stronger shock and more stagnation
pressure losses.
but unfortunately also the penetration high is lower and a hole appears a hole in the
centre of the duct looking to the species distribution. Moreover, inclining the second
95
Figure 5.11. Mach contour comparison between dierent distance from front and
rear jet in the aeroramp conguration.
injector too much, it will cause problem in the realization of the feeding system for
the aeroramp because 5dj correspond to just 1.2" of distance and because we have
injector pretty inclined we risk not to have enough space for tube and plugs, so we
decide to keep 60 deg as angle for the rear jet.
For the secondary injectors we try to give them a bigger angle to increase the penetration and reduce the length required by the mixing to develops completely. We
inject then, with an angle of 45 deg instead of 30 . Unfortunately even if we have
an improvement in the penetration high the increment in the slope of the injector
cause an increment in the thickness of the boundary layer before that. This disturb
go backward through the subsonic boundary layer, until the exit of the nozzle and
incrementing the layer also here provoke the starting of a more steep oblique shock
96
Figure 5.12. Stagnation pressure comparison between dierent distance from front
and rear jet in the aeroramp conguration.
that, will combinig with the bow shock of the front aeroramp lead to a normal shock
at the end of the nozzle. That cause a deceleration in the main ow and more stagnation pressure losses, so, even if that means more heat losses we denitely don't
want to move closer secondary injectors.
Moreover, considering the estimate of the area occupied by a jet plume downstream
the injection point (see section 5.0.3), we decide to place secondary jets one, on
the upper wall, opposite to the aeroramp and the other, on the lateral side at 25
deg from the horizontal middle axis . In the gure below is shown the secondary
jets position related to the cloud that we estimate they will develop around 25dj
downstream that, considering aeroramp jet diameter, means about 6" downstream.
4)Distance between primary and secondary injectors.
97
Figure 5.13. Scheme of the transverse injection 6" downstream the aeroramp
injection point.
Again we try to reduce this distance because of the heat losses. In our simulations we
don't run the case with a bigger distance that would probably lead to enhancement
in the mixing because, without liners, every inch is going to make the dierence
between the success or not of our experiment. We try to place secondary injector
just 4" instead of 5" downstream but, as you can see from g. 5.14, even this very
little change will make the rst bow shock much more stronger so, also in this case
we conrm our rst choice and we keep as distance 5".
5)Shape of the section after the nozzle.
During the injector project we decide to inject in a pipe with 3" as diameter instead
of 3.5" because, since our restriction in the injection pressure, we were not able
to reach the centreline in a bigger section.
and then there are other two congurations available for the
section after the nozzle: one that consists in a straight tube, 16" long, 3" internal
diameter with a 4" long expansion section at the end (conguration
2)
and another
that consists in a gradual expansion from 3" to 3.5" that start from the end of the
nozzle and end last 20" (conguration
For conguration
3).
two dierent shape for the expansion section are available: one
2a)
98
Figure 5.14. Mach contour comparison between dierent distances from primary
to secondary jets: 5" on the top and 4" at the bottom.
2b).
To choose which conguration was the best we mainly looked at four parameters:
total pressure losses, Mach Number at the end of the all section, eciency of mixing (in terms of methane molecular fraction, where the optimum is 0.23) and total
temperature at the exit.
In the table below has been summarized a comparison between parameter variation
in the three dierent congurations.
Looking for the result we can see that the rst conguration is the worst since
Figure 5.15. Comparison between variation of four main parameter in four dierent congurations.
99
all the parameter of eciency are the lower. That happens because with a bigger
exit area the Mach at end of the nozzle is going to be higher, that means lower
momentum ratio and lower penetration for aeroramp and also for the secondary injectors. This reduction in penetration will lead to a non fully mixing with a big gap
in the centre of the transverse section. Moreover faster main ow interacting with
aeroramp injected mass ow will bring stronger oblique shocks with higher pressure
losses.
About conguration
better result the mixing, because the increasing of the transverse section, is going to
characterized, even after several inches, by dense methane cloud not homogeneously
mixed with steam.
Since an ecient mixing is one of the fundamental requirements and because except
this conguration
tion number
2.
and
In this last case analysing the two dierent shape for the expansion
section it comes out that the smoother ones give some improvements, however the
cost and the time for machining that will be so high compared with a simple conical one that we prefer to avoid this option. In fact even with a cone we are going
to reach the goals that we have been required to achieve at the beginning of the
project.
So we opt for the rst conguration where the passage to the bigger section is
achieved by a conical section 4" long.
where the wall are inclined by a very little angle: 3.6 deg.
100
Chapter 6
CFD analysis
Finally a computational uid dynamics analysis was carried out to verify if the initial requirement of supersonic nozzle and injection section have been respected. A
code based on nite volumes method, Fluent, has been used.
101
6 CFD analysis
As before, to discretize our domain we are going to use tetrahedral elements but
since the complex shock structure that we expect in dierent areas of our section,
we create several regions with dierent grade of mesh renement. Our domain is
going to be 42" long and about 3.5" width. We decide to set a uniform body sizing
4
for all our model with a element size of 1.27 10 inches and then we rene it
adding area with higher density of elements where we expect big pressure gradient,
shock structure and boundary layer development.
Workbench-Ansys called " sphere of inuence" that gives you the option, instead of
specifying uniform mesh density for the entire geometric entity, to dene a coordinate
system and a radius to designate a sphere where elements will have a certain size.
So we use this function to intensify the mesh near all the injector, because here is
going to develop a sequence of shocks and Mach disk to allow the jet to expand. We
already know that here the ow eld is going to be very complex with big changes
in ow properties and because is fundamental solve properly this region to get the
right solution in the eld near here, we decide to put not just one sphere of inuence,
but four: two in the injector port and two in the area just before them (g. 6.2), all
5
with dierent element sizes. In particular we have as size one 1.27 10
inches, for
5
5
size two 2.28 10
inches, same for size three and for size four 7.62 10
inches.
We decide to do this gradual reng because using just one really ne mesh for all
the model would be just a waste of time with very big computational cost that will
make very un-ecient our analysis, moreover our machine has limits in memory and
using everywhere a very ne mesh we will denitely overbook it.
102
6 CFD analysis
Figure 6.2. Mesh of our uid model with enlargement of the critical zones .
Using this kind of renement injectors inlet looks like in g. 6.3. Jet inlet discretization is very important because that will determine the accuracy with the
injected mass ow is going to be calculated and that is going to aect bow shock
strength.
Another really critical points boundary layers discretization. After running a couple
of simulations we analyse the behaviour of the ow near the wall and then we decide
the best model to divide that area was using a function called "ination". In nearwall regions, boundary layer eects give rise to velocity gradients which are greatest
normal to the face. Computationally-ecient meshes in these regions require that
the elements have high aspect ratios, if tetrahedra are used, then a prohibitively
ne surface mesh may be required to avoid generating highly distorted tetrahedral
elements at the face. CFX-Mesh overcomes this problem by using prisms to create a
mesh that is nely resolved normal to the wall, but coarse parallel to it. The mesher
can use the local face element normals to 'inate' 2D triangular face elements into 3D
prism elements at selected walls or boundaries. We can control the creation of these
elements and determine their size and distribution in near-wall regions (g. 6.4).
To make more ecient our computational calculation, since in the converging part
103
6 CFD analysis
of the nozzle the boundary layer is really thin compared to the rest of our model
where the ow is supersonic, we decide to discretize the boundary layer with two
dierent kind of ination. They will have both 7 layers but while in the converging
5
section they are going to be just 4,36 10 inches thick, in the rest of the domain
5
they will be 7.62 10 inches.
104
6 CFD analysis
Discretizing our uid model as previously described at the end we have 673554 elements that is a number really close to the maximum allowed from our computer.
To verify if our mesh was enough ne to properly solve that problem we use another
function in Fluent: "adapt". With this option you csn decide maximum and minimum number of cell that you want to be rened and you can choose also the coarsen
and the rene threshold in terms of the gradient of one of the ow magnitude. We
did that for pressure gradient and we nd that the only area were our mesh was not
adequate was the area just at the exit of aeroramp injector and is for that reason
that we decide to rene it in the way before explained.
When the
reactants mix at a molecular level they instantaneously form products. The model
assumes that the reaction rate may be directly related to the time required to mix
the reactants at the molecular level.
mainly by eddy properties. Therefore, the reaction rate is proportional to a mixing time dened by the turbulent kinetic energy,
rate
This concept of reaction control is applicable to a wide range of industrial combustion problems. This time, unlike the nozzle analysis, we enable the calculation
of multi-species transport that models mixing and transport of chemical species by
solving conservation equations describing convection, diusion and reaction for each
component specie. In the species model we enable also the calculation of thermal
diusion.
As "material" we choose a mixture of steam and methane, with the density that
follow the ideal gas law and the Specic heat (CP ) evaluated through a mixing law.
This law will compute the mixture's specic heat capacity as a mass fraction average
of the pure species heat capacities(eq: 6.1).
CP =
Yi CPi
(6.1)
i
Thermal diusivity and mass diusivity have been set constant while viscosity will
vary with temperature following Sutherland's law (section 3.3.5.).
105
6 CFD analysis
In the solution method we decide to use, also here an explicit formulation, and to
avoid instability we use a very low Courant number, 0.05, when we turn on each
injectors. In fact, because of the complexity of the ow eld we should approach
to the solution through dierent steps.
injectors o and we wait until residuals are pretty low and at, so we start turning
on one by one all injectors; at the beginning the Courant number is 0.05 and then
we gradually increase that while the solution is becoming more stable. All the other
parameter for spatial discretization are the one already described in section 3.3.5.
As boundary condition now we have to set, not just for main ow inlet and outlet, but also for all the injector inlet. As for the main oe inlet we choose the type
"pressure inlet", suggested when you have compressible ow at the entrance. Here
stagnation (1034214 Pa) and static (598240 Pa) are required. We impose also the
velocity vector of methane to be normal to the inlet. We decide to put the boundary
condition at the inlet of the injector pipe and not right at the exit because this will
be better also to verify the mass ow through the ports.
model parameters, after verifying that changing them no improvement have been
achieved, we leave the one set by default.
Here we specify also methane total temperature (850 K) and species mole fraction of
methane (1) and steam (0). At main ow inlet we set the opposite (methane 0 and
steam 1) while at the exit, since we didn't want to inuence too much the solution,
we didn't specify anything about the species mole fractions.
Finally in this last simulation we decide to modelize also to add the fact that the
wall are going to be preheated so we sat wall temperature at 600 and we gave an
average thickness of 0.75".
106
6 CFD analysis
look at: the value of residuals, their trend (if they are still decreasing of if they are
at) and also their local value and not just to the average one. Moreover is always
suggested to look at the net mass inbalance, in fact to consider that a problem has
been completely solved that value has to be around 2% of the total ow thorough
the system (for us that means at least 0.0009
Kg/s).
107
6 CFD analysis
that, because of the shock sequence, a gradual recover in static temperature (g
6.7).
Figure 6.6. Mach contour reection and lambda structure and section geometry.
Figure 6.7. Static temperature contours at the beginning of the mixing section.
After this sequence of shocks, in the mixing section, the static temperature is going
to be around 1550-1650 K that is good since pyrolysis is going to occur at 1600 K. In
108
6 CFD analysis
the section before instead, from that simulation comes that the static temperature
will rise a lot especially in some zones where the ow will slow down to subsonic due
to the presence of recirculation bubbles and detachment of boundary layer caused by
shook wave formation. When we look at the static temperature we should remember
that in these analysis steady ows were considered, due to excessive time that would
be required to model transient ows, so at the end we will denitely have less high
temperature critic areas.
The total temperature will be around 1650-1700 K and we are going to ow is going
to be pretty well thermally mixed (g. 6.8). We can also notice that, even if we will
have pretty high heat losses, since we will pass from 2100 K to 1700 K, we will still
be in the range useful for our reaction.
About the heat losses, to have a rst estimate of how much heat we are going to
loose in these two sections, we set run two dierent simulations : one with adiabatic
walls and one with wall heated at 600 K. Then we integrate the total temperature
through four dierent transversal planes along the tube and, comparing these two
sets of data we achieve a roughly estimate of the losses g. 6.9).
below we can see, as expect that the bigger amount of heat losses will be in the
diverging section of the nozzle where we will have about the 64% of the total losses.
109
6 CFD analysis
This part is particularly critical since here we have the higher temperature and
higher velocity . At the end it comes that we will loose around 55 K .
Figure 6.9. Total temperature losses along nozzle-injector and mixing sections.
Looking at stagnation pressure trend along the duct, we can see that we will pass
from 2.5 bar to 1.2-1.3 bar, so we will loose more that an atmosphere mainly due
to the injection of methane. Especially we can see that in the injection section the
zone where we have more losses in the region just after the second aeroramp. The
only way to minimize pressure loss in this area is injecting from the rear aeroramp
jet with a lower angle, accepting to loose some penetration high.
Figure 6.10. Stagnation pressure losses along nozzle-injector and mixing sections.
Finally we can see that, for jet penetration high and methane-steam mixing, our
goals have been achieved.
110
6 CFD analysis
aeroramp jet penetrate much more further o the centerline and also the secondary
jets, seem penetrate more than predict due to the blockage eect of the primary
injectors. Even with a low momentum ratio and low injecting pressure (compared
with values usually used in transverse injection to achieve same results) we will be
able to cover all the transversal section diameter. Looking also at the mixing at the
end of this section we will have a pretty homogeneus mixing where about the 80%
of the area is fully mixed (that means that in this area methane mass fraction is
about 0.23).
Figure 6.11. Methane mass fraction contour along the duct and at the exit plane.
Moreover, placing dierent planes in sequence after main injectors, we can observe
the development of the jet plume downstream ( 6.12), with the typical cloud structure that rises prograssively until 6" downstream (next to last plane), after no more
impruvment have been noticed.
We can compare the prediction made in section 5.2.2 with Fluent result at the same
distance (g.( 6.13). We can see that, for secondary injector, since in our prediction
we have take in account the increment in
timeted, while for the aeroramp seems that we underestimate a little the exective
high penetration.
Concluding we have several observation to make: about the supersonic nozzle design
it come out that the quasi 1D model gave an error of just 7% if compared with a 3D
axisymmetric one. In fact with our Matlab program we get an exit Mach number of
2.12 while, from the CFD simulation we get about 1.9. These error is understandable if we consider the dierent approximation made.
About the injection design we can say that the best technique founded to inject a
non-neglectable mass ow in a supersonic ow at very high temperature is the dual
injection one, also called aeroramp. This technique has shown much more eciency
111
6 CFD analysis
Figure 6.12. Jet plume development 8" downstream the rst injection .
Figure 6.13. Jet plume development 8" downstream the rst injection .
112
6 CFD analysis
than predict. In fact, reading previous experiment about this conguration it seems
that with a dual transverse injection we will have an improvement of 1.5 times in
the high penetration. When we calculate that with our values of
J,
and
both
for main stream and jet even considering that the penetration after the injection
was growing with a parabolic development it comes out that we underestimate the
eective penetration high. Looking at the result seems that aeroramp will improve
the high not just of 1.5 times but for 2 times the value of single transverse one.
Moreover it seems that inclining the jet to 30 deg, but injecting after a primary injection will give the same penetration of injecting directly with 90 deg. This shows
that realizing a staggered injection has been a good compromise that let us achieve
the same penetration high with launch less pressure losses (shock in front an oblique
jet is weaker than one in front of a transverse jet).
Looking to the results, since the penetration high has been initially underestimate,
a good optimization will probably be inclining more the rear aeroramp in order to
have less pressure losses. However we should be careful because with this conguration every single parameter is strictly connect with the others. In fact, changing the
angle, will probably give, on one hand, less pressure losses but on the other will also
give less "blockage eect" for the secondary jets with the consequence of a lower
penetration high due to the reduction in
main ow is faster so J seen from secondary jet will be lower as explain in eq. 4.1).
More also changing rear jet angle will change also the optimal distance between
aeroramp jet and between them and secondary ones.
About mixing this can probably be improved inclining rear aeroramp too. In fact
due to the higher penetration reached this had let to a bigger methane gap in the
bottom of the tube so, inclining more will give less penetration and a lower gap.
Again, doing that, we should be sure to not leave big gap in the middle due to
decrease of penetration high.
Results achieved in this project about sonic injection in supersonic mainow are very
important because directly connect to the eld of supersonic combustion engine. In
fact, as said before, our project was presenting several similarities with scramjet
and it also presents more critical conditions if we consider the bigger mass ow
injected (about ten time the amount injected in scramjets ) and higher temperature. So we have proved that aeroramp could be the best way to inject fuel because :
- thanks to the higher penetration no intrusive injector are needed to cover all
the transversal area;
- realisation time, costs and diculties are reduced dramatically due to the very
simple geometry;
- since they have a really high eciency lower injecting pressure is required that
means reduction in power spent in the fuel system;
113
6 CFD analysis
inject a mass ow equal to the 30% of the main mass ow in a length of 6" without
causing normal shock and the transition to subsonic of the main ow. If we consider
that in scramjet the mass ow is much more smaller is understandable that the
pressure losses will be reduced exponentially leading to higher exhaust gas velocity
and thrust. Moreover aeroramp can replace one of the most used system of injection
(central injection) where the fuel is delivered from a strut that is mounted across
the combustion chamber. That conguration require a cooling system and it also
has pretty high free stream pressure losses.
114
6 CFD analysis
Bibliography
[6.1] Dieter K. Huzel,David H. Huang,
[6.6]
[6.7]
[6.8]
Homan,
Mixing Eects of PylonAided Fuel Injection Located Upstream of a Flameholding Cavity in Supersonic
Flow. , 2005.
[6.12] G.N. Abramovich, The theory of turbulent jets., Cambridge, MIT Press, 1963.
[6.13] F.W. Spaid, A study of secondary injection of gases into a supersonic ow.,
[6.11] D.R. Montes, P.I. King, M.R. Gruber, C.D. Carter,
Gruber, A.S.
Nejad,
1973.
[6.18] R.C.
Rogers,
115
6 CFD analysis
Scramjet
Montes, P.I.
King, M.R.
Gruber, C.D.
Hsu,
July 2000.
Comparison of Physical and Aerodynamic Ramps as Fuel Injectors in Supersonic Flow, Journal of Propulsion and
[6.26] C. D.
Sang-Hyeon Lee,
116