E43 WIHHIHHIHWHIHIIHIHIHWIIIWINWIIWIWI

     

'THS

WV mm
it. man e
)W University

This is to certify that the
thesis entitled

DESIGN, FABRICATION AND TEST OF A GAS TURBINE
ENGINE AND WAVE ROTOR TEST BED

presented by
JOHN CHARLES QUACKENBUSH, II

has been accepted towards fulﬁllment
of the requirements for the

MS. degree in MECHANICAL ENGINEERING

 

 

MIL/Zr

Major Professor’s Signature
8 _/ I 3/08

Date

 

 

MSU is an afﬁrmative-action, equal-opportunity employer

 

 

I-l-I-I-C-O-O-I-O-0-.-'-I-D-C-O-O-O-O-O-O-I-O-I-O-O-n_l‘I-l-I-I-I-I.

o~<‘--A-l-l-u-n—a—.— —

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K‘IProleccaPreleIRCIDataDue indd

DESIGN, FABRICATION AND TEST OF A GAS TURBINE ENGINE AND WAVE
ROTOR TEST BED

By

John Charles Quackenbush, II

A THESIS

Submitted to
Michigan State University
in partial ﬁilﬁllment of the requirements
for the degree of

MASTER OF SCIENCE
Mechanical Engineering

2008

ABSTRACT

DESIGN, FABRICATION AND TEST OF A GAS TURBINE ENGINE AND WAVE
ROTOR TEST BED

By

John Charles Quackenbush, II

This thesis examines two core technologies developing within the energy
production and transfer discipline: smaller scaled gas turbine engines and performance
enhancing wave rotors. A combustion chamber was designed to satisfy the requirements
to operate both devices. The chamber was built to be integral within the test apparatus,
which was outﬁtted with a precise data acquisition system to critically monitor and
evaluate these technologies in real-time.

Both the wave rotor and gas turbine engine prototype were tested. The testing
was done separately to ﬁrst fully understand these technologies in order to later enhance
the gas turbine engine with the integration of the wave rotor as an engine topping
performance enhancer.

This thesis will discuss the analysis and decisions made to design and fabricate
this test apparatus. Data were collected for both devices allowing for solid evaluation

and recommendations on how to further proceed.

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................... v
LIST OF FIGURES ............................................................................................................ vi
NOMENCLATURE ........................................................................................................... ix
INTRODUCTION ............................................................................................................... 1
TEST HARDWARE ............................................................................................................ 3
TURBOCHARGER ......................................................................................................... 3
WAVE ROTOR ............................................................................................................... 7
COMBUSTION CHAMBER DESIGN ............................................................................ 10
PREVIOUS COMBUSTOR .......................................................................................... 10
DUMP COMBUSTOR .................................................................................................. 13
INNER CONE COMBUSTOR ..................................................................................... 15
COMBUSTION BYPASS ............................................................................................. l6
EQUAL ANGLE - SINGLE SPLIT .......................................................................... 19
OFF ANGLE - SINGLE SPLIT ................................................................................ 21
EQUAL ANGLE - DOUBLE SPLIT ........................................................................ 24
DIFFUSION LENGTH ................................................................................................. 25
LEADING CONE FLAME HOLDER .......................................................................... 28
GUIDED DIFFUSION .................................................................................................. 31
CENTRALIZED FLAME HOLDER ............................................................................ 37
IGNITION SYSTEMS ...................................................................................................... 44
INTERMITTENT IGNITION ....................................................................................... 44
CONTINUOUS IGNITION .......................................................................................... 46
STARTING SYSTEMS .................................................................................................... 48
GATE VALVE .............................................................................................................. 48
TURBINE NOZZLE ..................................................................................................... 51
COMPRESSOR NOZZLE ............................................................................................ 54
TESTING CONFIGURATIONS AND RESULTS .......................................................... 58
DATA ACQUISITION SYSTEM ................................................................................. 58
COMBUSTOR TESTING ............................................................................................. 62
GAS TURBINE ENGINE TESTING ........................................................................... 67
WAVE ROTOR TESTING ........................................................................................... 76
CONCLUSIONS ............................................................................................................... 79
RECOMMENDATIONS ................................................................................................... 80

iii

WORKS CITED ................................................................................................................ 84

BIBLIOGRAPHY ............................................................................................................. 85

iv

LIST OF TABLES

Table 1. Gas turbine engine 5 kW operating point ............................................................. 5

Table 2. Wave rotor ﬁrst test and validation operating point ............................................. 9

LIST OF FIGURES

Figure 1. Project goals and direction .................................................................................. 2
Figure 2. Garrett GT1241 turbocharger, basic operating principle .................................... 4
Figure 3. Compressor map, Garrett GT1241 manufacturer data [Garrett, 2008] ............... 6
Figure 4. Turbine map, Garrett GT1241 manufacturer data [Garrett, 2008] ...................... 7
Figure 5. Wave rotor, schematic view showing ﬂow directions ........................................ 8
Figure 6. Straight channel wave rotor ................................................................................ 9
Figure 7. Wave rotor, basic operating principle ............................................................... 10
Figure 8. Previous combustor design, inner annulus type ................................................ 11
Figure 9. Dump combustor, velocity contour ................................................................... 15
Figure 10. Inner cone combustor, velocity contour ........................................................... 16
Figure 11. Equal angle — single split, velocity contour ..................................................... 20
Figure 12. Equal angle — single split, hardware design ..................................................... 21
Figure 13. Off angle -— single split, velocity contour ......................................................... 23
Figure 14. Off angle — single split, hardware design ......................................................... 24
Figure 15. Equal angle — single split, velocity contour ..................................................... 25
Figure 16. Conical diffuser, velocity contour .................................................................... 27
Figure 17. Conical diffuser, inlet and exit velocity proﬁles .............................................. 27
Figure 18. Leading cone ﬂame holder, velocity contour ................................................... 29
Figure 19. Leading cone ﬂame holder, inlet and outlet velocity proﬁles .......................... 30
Figure 20. Lead cone ﬂame holder showing recirculation zones ...................................... 31
Figure 21. Straight duct section as fabricated .................................................................... 32
Figure 22. No guide vanes, velocity contour ..................................................................... 34

vi

Figure 23.
Figure 24.
Figure 25.
Figure 26.
Figure 27.
Figure 28.
Figure 29.
Figure 30.
Figure 31.
Figure 32.
Figure 33.
Figure 34.
Figure 35.
Figure 36.
Figure 37.
Figure 38.
Figure 39.
Figure 40.
Figure 41.
Figure 42.
Figure 43.

Figure 44.

Figure 45

Single guide vane, velocity contour ................................................................. 35
Double guide vane, velocity contour ............................................................... 36
Double guide vane, inlet and outlet velocity proﬁles ...................................... 37
Centralized ﬂame holder, velocity proﬁle ....................................................... 38
Centralized ﬂame holder, ﬁ'ont view ............................................................... 4O
Centralized ﬂame holder showing mixing holes .............................................. 41
Centralized ﬂame holder, bottom view ............................................................ 42
Centralized ﬂame holder, full view .................................................................. 43
Spark ignition with spark plugs on outer wall ................................................. 45
Spark ignition with spark plug inside ﬂame holder ......................................... 46
Ignited pilot ﬂame with MAPP gas and Oxygen gas kit .................................. 48
Broken compressor-turbine shaﬁ ..................................................................... 49
Swinging check valve used for engine starting ................................................ 50
Turbine nozzle used for engine starting ........................................................... 52
Compressor start nozzle, velocity contour ....................................................... 56
Compressor start nozzle, velocity vectors ........................................................ 57
Compressor start nozzle, operating characteristics .......................................... 58
Data acquisition devices: volumetric ﬂow, pressure and temperature ............. 60
Turbine inlet temperature thermocouple .......................................................... 61
Data acquisition board and control computer .................................................. 62
Combustor only test with bypassing air present .............................................. 64
Combustor only test without bypassing air present ......................................... 66
Centralized ﬂame holder operating at .071 kg/s, velocity contour ................. 66

vii

Figure 46. Open-loop turbocharger test without combustion, mass ﬂow and pressure ....69

Figure 47. Open-loop turbocharger test without combustion, mass ﬂow and pressure ....71

Figure 48. Open-loop turbocharger test with combustion, mass ﬂow and I.T.T ............... 73
Figure 49. Open-loop turbocharger test with combustion, pressure and I.T.T .................. 73
Figure 50. Open-loop turbocharger test with combustion, mass ﬂow and I.T.T ............... 75
Figure 51. Open-loop turbocharger test with combustion, pressure and I.T.T .................. 75
Figure 52. Wave rotor test apparatus ................................................................................. 76
Figure 53. Open-loop wave rotor test with combustion, mass ﬂow and I.T.T .................. 78
Figure 54. Open-loop wave rotor test with combustion, pressure and I.T.T. .................... 78

Figure 55. ASME bell mouth sealed for use with gas turbine engine, cross section view81

Figure 56. Hot surface i gniter commonly used in residential furnaces ............................. 82

viii

< r-i W

:I

NOMENCLATURE

Area

Diameter or width

Inlet turbine temperature

Mass ﬂow rate

Pressure

Gas constant for air (287 kJ/Kg—K)
Temperature

Velocity

Constant (3.1416)

Density

Volumetric ﬂow rate

ix

INTRODUCTION

This thesis is an advancement of previous work as well as a venture into newer
realms of research. The goals of this thesis are to examine two core technologies
developing within the turbomachinery discipline: smaller scaled gas turbine engines and
performance enhancing wave rotor devices.

In previous work a complex Brayton cycle analysis was conducted for a speciﬁc
turbine—compressor hardware couple for the purpose of developing a common gas
turbine engine capable of 5 kW electrical power generation [Kusner 2006]. This thesis
expands upon this work with the design of an appropriate combustion chamber to drive
the turbine-compressor couple to produce a desired power output. Once the combustor
was complete the engine was evaluated by direct testing within a precisely monitored test
apparatus utilizing a real-time data acquisition system.

This project was also to simultaneously develop and test a speciﬁc wave rotor
technology. As later discussed, the wave rotor is a non-steady pressure exchange device
that can Operate within a common engine cycle. Its purpose is to increase the pressure of
the main engine air intake before combustion. As a result the wave rotor has the ability
to increase efﬁciency and enhance the overall performance of many engine types: gas
turbine engine or others.

Like the gas turbine engine efforts discussed above the wave rotor technology will
be evaluated through direct testing. A test rig analogous to the gas turbine engine rig
must be developed. To properly evaluate the wave rotor, the high temperature points that
it would naturally operate in as a typical engine cycle enhancer must be achieved. It was

determined that a parallel design path should be taken. The combustor that is needed for

the gas turbine engine should also be designed to be capable Of generating the operating
conditions needed to evaluate the wave rotor at real operating points. Like the
combustor, the data acquisition system will also be designed to satisfy the needs of the
gas turbine engine as well as the wave rotor testing.

Figure 1 shows a ﬂow chart of what has already been accomplished with the gas
turbine engine and wave rotor developments as well as the design, building, and testing
objectives of this thesis. This ﬁgure concludes with the pinnacle goal of these efforts after
rigorous analysis and testing being the coupling of the two technologies. The gas turbine

engine will be enhanced with the wave rotor proving both technologies.

 

 

GAS TURBINE ENGINE
(GTE)
PROJECT DEFINITION

WAVE ROTOR (we)
PROJECT DEFINITION

 

 

 

 

 

 

CYCLE ANALYSIS CYCLE ANALYSIS

I I

COMPRESSOR-TURBINE WAVE ROTOR
SELECTION SELECTION

IIIIIIIIII III-IOIIIIIIIIIIIIIIIIII AIIIIIIIIIIIIII
COMBUSTOR DESIGN
SUITABLEFOR

GTE & WR

I

TEST RIG CAPABLE OF
TESTING GTE & WR

I

DAQ SYSTEM COMMON‘

/ FOR GTE a WR \

GTE TESTING WR TESTING

 

 

 

 

 

 

 

 

 

 

 

   

 

 

 

OBJECTIVE OF THIS
PROJECT

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

GTE ENHANCED
BY WR

   

 

Figure 1. Project goals and direction

TEST HARDWARE

TURBOCHARGER

The design intent for the turbocharger technology is to couple it with an
appropriate combustion chamber in a closed-loop orientation such that it can Operate as a
typical gas turbine engine. Gas turbine engines are common as off-the-grid electrical
power sources, generally at larger power outputs. This engine aims to generate power at
a smaller scale, 5 kW, which would prove to be more useful for personal residential uses.
The niche in the electrical generator market is currently dominated by reciprocating
internal combustion engines commonly running on gasoline or diesel ﬁiel. This gas
turbine engine will be run with propane as fuel.

A speciﬁc turbocharger and the precise Brayton cycle in which it needs to Operate
within were previously determined by the predecessor of this project [Kusner, 2006]. He
determined to use a turbocharger from Garrett which is a subsidiary of Honeywell.
Speciﬁcally the turbocharger is 3 Garrett GT1241 model, which is the smallest in the line
that Garrett offers. This turbocharger, when used in automotive applications is
recommend for internal combustion engines with 0.5 to 1.2 liters of displacement and

capable of 50 to 130 horsepower [Garrett, 2008].

Combustion
Air In

Combustion
Air Out

 

Figure 2. Garrett GT1241 turbocharger, basic operating principle

The Brayton cycle analysis previously executed has already set the Operating
point needed to run the gas turbine engine to produce the goal of 5 kW Of power. The
operating point has been determined to be 0.071 kg/s (roughly 9.4 Ibs/ min) of mass ﬂow
rate through the system. The turbocharger will ingest air at standard atmospheric
pressure and temperature. The pressure ratios of the compressor and turbine have both
been set to be 2.5 [Kusner, 2006]. This means that the 101.3 kPa atmospheric air
entering the compressor will be increased to 253.3 kPa. After being ignited in the
combustion chamber, this hot gas will expand across the turbine and be brought back to
near atmospheric pressure before being exhausted (Figure 2). In order for the compressor
and turbine to match in pressure and ﬂow there must be an energy input to the system to
overcome the mechanical losses. This energy will be added in the form Of heat addition

to the air after it has been compressed but before entering the turbine. By the cycle

analysis, this heat addition has been calculated so that the 0.071 kg/s Of air ﬂow must be
elevated to 1200 Kelvin before entering the turbine. When looking at the compressor
map for the Garrett GT1241 (Figure 3), it can be seen that the Operating point Of 9.4
lbs/min and a pressure ratio Of 2.5 falls nicely at the highest efﬁciency point for this
chosen turbocharger, 76% efﬁcient. Note that at this operating point the compressor-
turbine couple will be spinning at roughly 215,000 RPM. Table 1 shows a summary Of
the Operating condition that must be designed and tested [Kusner, 2006]. Figure 4 is a

turbine map analogous to the compressor map.

Table 1. Gas turbine engine 5 kW operating point

 

Turbocharger - GTE Operating Point
Mass Flow[ Combustor Temperature] Combustor Pressure

0.071 kg/s | 1200 K | 253 kPa

 

 

 

 

 

GT1241, 41 mm,50 Trim, 0.33 AIR

 

 

 

 

 

 

 

 

 

 

 

3DT—_TT_—_T_——T——_7
I I | I
l I l J
|
I
2.5 . —I
I
I
I
I
2.0 - —J
Pressure J
Ratio ' ‘ l
I é 200000 rpm J
' I
1'5 T_ ‘ “I; '"180000rpm —__j
’ ’ 1.60000 rpm J I
J ’ 140000 rpm J J
I ‘ 120000 rpim I I
I I I I
1.0 I I I A
0 5 1 0 1 5 20
Corrected Air Flow (lb/min)

Figure 3. Compressor map, Garrett GT1241 manufacturer data [Garrett, 2008]

GT1241, 72 Trim, 0.43 A/R

 

 

10 __—T—_7_-—T_—_T—__§_—_I————T—_——I
.,____j____|___3____I_____‘___I_.___i___l
I I g I I
______ _I___.___._I____§____I________I
8 : I , I E I , I
_——+——4——~9——4u—A—a—#——+——4
I 5 [Hi I I
CorrectedGasG *TT—T‘TT/é—T_——;—“‘T”‘_j_“—J
TurbineFlow .IL____; _ "___3. ______ §_L_ __;___
(lb/min) —I T I— I
4 _——+——j———f——T——ﬁ———r——v——j
________ I ___3____'___.___|_______.__.I
I E I I
2 L _____ n__-;__; _______ I _______ I
I i I , | , I
______ —I_T—i——_ Maximum Efﬁciency 65% I
O ' I ‘ I ' I ' 4.

1.00 1.50 2.00 2.50 3.00

Pressure Ratio

Figure 4. Turbine map, Garrett GT1241 manufacturer data [Garrett, 2008]

WA VE ROTOR

A wave rotor uses shock waves to pressurize ﬂuids by transferring energy from a
high-pressure ﬂow to a low-pressure ﬂow. It consists Of many curved or straight channels
arranged around the axis Of a rotating drum or disc.

Speciﬁcally for this project, an automotive Comprex 7 type waver rotor was used
for testing. Roughly 150,000 Of Mazda’s 626 powered by a diesel engine were
supercharged with a Comprex pressure wave supercharger [Dempsey, 2006]. Mazda
used this particular wave rotor design as a performance enhancement to this automotive
internal combustion engine. This project intends to build a test rig capable of testing this
wave rotor at operating points that would simulate its application as a gas turbine
enhancer.

A wave rotor works with the propagation Of shock and expansion waves. Figure

5 is a schematic view of how a wave rotor could be connected with a typical gas turbine

engine. The incoming air is ﬁrst compressed by the engine’s main compression stages
(state 0 to 1). This compressed air then enters the wave rotor channels through
speciﬁcally designed porting. Within these channels energy transfer takes place. The
cool, compressed air is exposed to hot combustion gases from the combustion chamber
(state 3). The pressure and temperature difference between the two ﬂuids induces a
shock wave that is used to compress the cool air. The pressurized air is fed to the
combustion chamber (stage 2). On the other hand, the hot exhaust gases are expanded,
leave the wave rotor (stage 4) and are further expanded across the engines turbine stages
[Akbari, 2006].

The wave rotor spins at a speciﬁc speed about its central axis. This speed and the
porting geometry that controls the ﬂows in and out of the wave rotor are precisely
determined such that the compression and expansion processes take place and enter and

leave the rotor at the correct ports.

 

Combustor

 

 

 

1 Wave Rotor 4

 

 

 

—h Compressor Turbine I—>

/ \

Figure 5. Wave rotor, schematic view showing ﬂow directions

 

 

 

 

 

Figure 6 is a top View of this speciﬁc wave rotor. This wave rotor is of the

straight channel type.

 

Figure 6. Straight channel wave rotor

The main objective for this test rig will be to evaluate Operating conditions by
directly simulating them. Table 2 shows the ﬁrst Operating point to be tested and
validated to. This operating point serves as a design constraint on the test rig, meaning
that the rig must be capable of this operating condition. It is worth noting that the wave
rotor will be spinning at 13,000 rpm at this Operating point. The electric drive motor to

accomplish this can be seen in Figure 7.

Table 2. Wave rotor ﬁrst test and validation operating point

 

Wave Rotor Operatinc- Point
Mass Flow Combustor Temperature Combustor Pressure
0.056 kg/s 850 K 250 kPa

 

 

 

 

 

 

 

a .
v- ... l. _

Combustion ~
Airln J j. . _ Ambient

\

Air In

, COmpressed
Air Out

 

a ‘1, >._- .L R

Figure 7. Wave rotor, basic operating principle

COMBUSTION CHAMBER DESIGN

PREVIOUS COMBUS TOR

The previous combustor was an inner annulus design [Kusner, 2006]. This
annular design was made of an inner and outer pair if pipes both equally concentric about
the ﬂow direction. The outer pipe was the same 2-inch diameter, schedule 40, steel pipe
Of which the rest Of the rig was made. The irmer can (Figure 8) was made of a l-inch

diameter, schedule 40, steel pipe. A large majority of the ﬂow was designed to pass

10

around this inner pipe. This bypassing air is not needed for the combustion process and
can serve as cooling air for the much hotter combustion inside. Some of the air enters the
inner can of the combustor. Shortly downstream of where the air enters, the fuel is
injected. The air and the fuel mix inside the inner can with the help of mixing holes
drilled perpendicularly through to the outer ﬂow. This air and fuel mixture will then be
ignited at the end of the inner can by two automotive spark plugs. The spark plugs were
placed at the end with the intent that the ﬂame would attach itself to the end of the inner
pipe. At the end of the inner can the air that bypassed will then be able to mix back with

the hot exhaust gas that was combusted.

MN ,y .
-M*§.w -' .
i' Flow Diraction

‘¢ '4‘ War
.21 .hb.‘

r .&‘.4.

 

Figure 8. Previous combustor design, inner annulus type

This design was tested by Michael Kusner [Kusner, 2006]. His testing showed
that there were issues maintaining stable combustion. Some combustion was witnessed,
but it was only in short bursts that seemed to correlate with the spark ignitions of the
automotive sparkplugs used. This intermittent ignition made it impossible for the test rig
to reach any reasonable Operating point. Besides the intermittent ignition the other

assignable cause was determined to be the geometry of the combustion chamber. It is

desired to run the test rig at a mass ﬂow rate of .071 kg/s. It was previously determined
that 0.030 kg/s of this mass ﬂow is needed for combustion and the remaining mass ﬂow
bypasses the inner can [Kusner, 2006]. Using continuity, the bulk velocity of the air can

be determined for the bypass and inner can ﬂows.

_ 7r (D12 - I322)

 

 

 

 

 

 

 

 

Abypass - 4
2 2
= 7: (0.0525 0.0334 ) = 0.00129 m2
4
it D 2
Acombustion = 43
2
= 7‘ 0'0266 = 0.00056 m2
m = p A V
mbvmss
Vb, =
pass p Abypass
= '0‘" kgls = 14.52 m/s
2.18864 kg/m3 0.00129 m2
V _ mcorrbustion
conduction '-
' p Aconbustion
= '030 k9" = 24.48m/s
2.18864 kg / m3 0.00056 m2 (1)

These velocities are much tOO fast to sustain stable combustion Of propane. The

ﬂame speed Of propane at this operating point has been determined to be roughly 0.5 m/s

12

[Kusner, 2006]. As calculated, the velocities Of the air through the combustor are much
larger than the ﬂame speed Of propane, meaning that the air is capable Of pushing the
ﬂame downstream faster than the ﬂame can travel back upstream. This allows the ﬂame
to detach itself from the inner can, leave the combustor and travel downstream. This is
called a ﬂame out. High velocities are likely the cause of the ﬂame propagation and
explain why intermittent ignition was witnessed during testing. It was determined that a
new combustion chamber design was needed. Any new design needs to take these high
speed air velocities into account to prevent and minimize the likelihood of ﬂaming out

the combustor.

DUMP COMB US TOR

The ﬂame speed of propane has previously been calculated to be around 0.5 'm/s
[Kusner, 2006]. In order to have a stable ﬂame it is important that any combustion
chamber design have some region where the air is decelerated to this velocity. This
decelerated region would be where the ﬂame could stabilize and support sustainable
combustion. The ﬁrst design alternative that was examined to accomplish this
deceleration was a simple dump combustor design. The advantage Of a dump combustor
is that it is capable of quickly expanding and decelerating the incoming air to a much
larger diameter over a short axial length. Unfortunately, this sudden expansion causes a
large pressure drop in the combustor. To evaluate this as a possible option, a speciﬁc
geometry was analyzed using Fluent (Figure 9), a computational ﬂuid dynamics soﬁware.

The air will be entering the combustor through a standard 2-inch-diameter inlet
which corresponds to the diameter the air is extracted from the compressor. Using the

continuity principle it was found that roughly a 12.77-inch-diameter cross section was

13

needed to decelerate the air to a velocity below 0.5 m/s at the desired mass ﬂow rate of

0.071 kg/s.

m =pAV
m =0.071kg/s
p-_-2.169kg/m3

Vair = Vﬂamespeed = 0-5 m /s

A = m = 0.0649 m2 = 100.60 in2

P vﬂamespeed
4 ﬂ = 12.77 in
7r (2)

 

D:

 

This quick calculation essentially determined the overall geometry Of the dump
combustor. The axial length of the combustor was limited to 24 inches in length for
practical purposes and also to be easily adapted in to the current space available. The
analysis showed that at the desired mass ﬂow rate the velocity of the air in the 2-inch-
diameter inlet is roughly 17 m/s which is well above the desired 0.5 kg/s target. The
sudden expansion to the 12-inch-diameter does cause the air to decelerate. Near the outer
walls Of the large diameter region the air decelerates down to the desired velocity. The
large majority of the air remains at high velocity and quickly passes and exits through the
center Of the combustor. This design would be undesirable. Most of the air would not be
able to support the combustion process due to these high velocities. The ﬂame front
would be limited to the outer region Of the combustor where there is not enough air to

mix with the propane to provide the heat input that is needed to reach the Operating point.

14

1706*01
I 61er+01

153901

._ :r‘Itgurs OI 'vPIOC It‘. I‘."E'..:JI)|IIJd5-? [ITII‘sI

 

FLLIENI ‘i _' £-'-:

Figure 9. Dump combustor, velocity contour

INNER CONE COMBUS TOR

An initial design was to use an inner cone as a type of bafﬂe to create low
velocity regions in the ﬂow that could provide stabilization for the ﬂame. The main
advantage to having an inner cone bafﬂe design would be its simplicity. The outer
diameter can remain consistent with the rest of the 2-inch-diameter piping. One draw
back is that the ﬂow must be decelerated to reach the desired velocities by introducing a
pressure loss caused by the bafﬂes. Again a representative Fluent model was executed.
As before, the velocity of the air entering the combustor is roughly 17 m/s. Figure 10
indicates some low speed regions were developed down stream of the obstructions placed
in the ﬂow. These same obstructions also cause several areas where the ﬂow was
accelerated because it was being forced tO ﬂow around the bafﬂes. Also it was found that
at a 2—inch-diameter cross section it takes a considerable amount of blockage to create a

slow 0.5 m/s region. This blockage causes an unacceptable amount Of pressure loss. The

15

analysis showed that only 9 percent Of the total pressure at the inlet is maintained at the
outlet. Therefore, the inner cone design was deemed not to be the best Option. The air
must be expanded and decelerated in a way that conserves pressure yet still reaches the

desired velocities.

71r~01
SSEI CI
433~QI
Crie+01

1839+C‘1

1 JL
1

l

1

‘I

‘I

‘l

1 I

I’D

‘I‘fntts'urs ‘If “IPICCIII,” lvfag'uluclr: unis)

 

Figure 10. Inner cone combustor, velocity contour

COMBUSTION BYPASS

Due to previous analysis it was determined that the air must be decelerated by
diffusion rather than a pressure loss device. This way a larger portion of the high
velocity air can be transferred to static pressure rather than simply lost. Due to the large
mass ﬂow rate it would take a cross section equivalent to a 12.77-inch diameter to
decelerate the entire ﬂow Of air to the 0.5 m/s ﬂame speed of the fuel. The combustion
process only needs a flow rate of 0.03 kg/s to reach the energy input for the desired
operation point. This means that more than half the air flow does not need to be

decelerated for combustion and can go through the bypass instead. If the ﬂow is split

16

into the air needed for combustion and the air that can bypass, the amount of air that
needs to be decelerated is also reduced. A smaller cross section is needed to reach the
ﬂame speed if only the air required for combustion is decelerated. This smaller diameter
will provide a more compact and reasonably sized combustor and will have less loss in
pressure because a large portion of the air does not need to be altered and can bypass the
combustor as is.

This thought leads to the question Of how best to split the air ﬂow to the correct
proportions with minimal pressure loss introduced. Several orientations were examined
using a planar analysis. In reality, the geometries will be made Of 2-inch-diameter
piping. The analysis simulates the geometry as a per unit thickness in and out of the
analysis plane. In (order to relate the analysis to the true geometry a correlation factor is
needed to relate the velocities calculated in Fluent to what they would be for the true
round cross sections.

Mass ﬂow is simply deﬁned by a certain density of ﬂuid passing through a given
cross sectional area at an average velocity. For a planar analysis the ﬂow is two
dimensional meaning there is no cross section perpendicular to the ﬂow direction. The
ﬂow is only analyzed in a single plane. Fluent calculates the mass ﬂow in the same
manner as above but does it as a mass ﬂow per unit of cross section. This can be better
understood by thinking that for a planar case the mass ﬂow is calculated using just the
width (D) of the cross section the ﬂuid is ﬂow through. The depth in and out of the plane

is assumed to by unity (1).

l7

Aﬂuent = D

m = P D Vﬂuent (3)

The pipe that is desired to be represented is round and has a true non-zero cross
section. In reality the pipe has a diameter (D) and a dimensionally accurate mass ﬂow

can be calculated.

JI'D2
Aactual = T

«D2

 

m = P Aactual Vactual = P vactual

(4)

The mass ﬂow is what is being targeted in these analyses. Therefore the mass

ﬂow in the analysis can be set equal to the mass ﬂow of the true geometry. By doing this
a relationship is formed between the actual velocities with the velocities Fluent will

calculate.

7rD2
mactual = mactual = P _ Vactual = P Dvﬂuent

4
REDUCING TO

4

Vactual = vfluent """
7rD (5)

Now it can be seen that the actual velocity can be calculated from the Fluent
velocity simply as a function of the width D Of the plane Of ﬂow. For a true round cross

section the diameter is analogous and equal tO the width of the plane in the Fluent

geometry.

l8

EQUAL ANGLE - SINGLE SPLIT

The air is extracted from the compressor in a 2-inch-diameter pipe. The simplest
design would be to split the ﬂow in two directions with a standard “y” shaped pipe
design. For stabilized turbulent ﬂow in a pipe, it could be assumed that if the ﬂow is split
into two equal diameter pipes at equal angles from the upstream pipe, the ﬂow would be
separated quite closely in half. For the operating point desired, 0.03 kg/s must be
separated ﬁom the full 0.071 kg/s which is not half. The angles could be adjusted such
that the correct proportions are naturally metered. This would be achieved by having the
bypass pipe more in line with the upstream pipe than the combustion air pipe, seeing how
more than half should be bypassed. However, there would be no margin or correction that
could be used if it was later determined that the air is not being split properly. Or a
different operating point may later be desired. Therefore, it was decided that the best
option would be to have the angles split equally at 60 degrees from the centerline and
then slightly throttle the combustion air side to adjust for the precise ratio desired. This
throttling comes at the cost of pressure loss, so an analysis was run to determine how

much throttle would be needed to proportion the ﬂows correctly.

19

. 13 T5901
I 5 416501
‘ 07mm

'Tcn‘mrs of "Jelemty Magnitude II‘IIsj

 

Figure 11. Equal angle — single split, velocity contour

The Fluent model was meshed using a planar type of analysis (Figure 11). The
pipes were all 2—inches wide to be analogous with the true 2-inch-diameter pipes. A thin-
walled obstruction was placed on the right side Of one of the downstream pipes. This
wall simulated the throttling effect of a butterﬂy valve that was intended to be used. It
was found by Fluent simulation that the desired throttling of 0.030 kg/s through the
combustion and .041 kg/s through the bypass was achieved when the butterﬂy valve was
placed at roughly an angle of 9 degrees Off from the full Open position. Even though this
was a planar analysis and the true geometry has round cross sections, it proved that a
large pressure drop was not needed to achieve the desired ﬂows. The analysis indicated
that 68 percent of the total pressure was maintained as an average at the outlets relative to
the inlet. Figure 12 shows a possible design that would utilize this type of ﬂow

separation.

20

 

Figure 12. Equal angle — single split, hardware design

OFF ANGLE - SINGLE SPLIT

Due to the large cross sectional area needed in the combustion chamber to
decelerate the ﬂow it was thought for manufacturing purposes that the chamber should be
axial and concentric to the inlet and outlet directions of the overall test rig. Speciﬁcally,
this means that the chamber should be concentric to the ﬂow before it splits into the
combustion and bypass air. It should also be concentric to the ﬂow after the two ﬂows
mix back together. By doing this, the overall size of the combustion chamber and bypass
assembly can be minimized. The large chamber will be placed in the center and the

smaller bypassing pipes can be easily routed around the outside of the chamber. Another

21

reason for putting the chamber in the center was that the combustion process was
expected to be at 1200 K. This temperature is close to the melting point of most steels
used for this fabrication. By using the bypass air as a coolant the combustion chamber
can be ﬁred at temperatures above the melting point Of steel because the bypass air can
provide cooling to the outer walls of the chamber. Therefore, having the chamber
inboard and the bypass ﬂow around the outside would be preferred to cater to this design
intent.

A conﬁguration where the combustion air is maintained in a straight line and the
bypass air is extracted from the main ﬂow was considered. Less then half of the total
mass ﬂow needs to enter the combustion chamber. With the combustion pipe being axial
with the upstream inlet a larger portion of the air will naturally enter the combustion
chamber then will turn at an angle towards the bypass. This means that more throttling of
the valve then the previous equal angle — single split design will be needed. A similar
analysis as before was run to determine how much more throttling would be needed to

achieve the 0.030 kg/s of mass ﬂow through the combustion chamber (Figure 13).

22

5 158701
4 849-01
4 49001
4 lie-01
3 80c- 01

:1 469-01

31119-01

» 769-01

C Sill-Eula 3f \J’Eiscny Magmtude (In/5‘1

 

Figure 13. Off angle — single split, velocity contour

The analysis showed that the throttle valve must be placed at roughly 13.7 degrees
off of fully Open to achieve the desired throttling. This was 4.7 degrees more throttling
than the equal angle — single split design. However, by analysis 81 percent Of the total
pressure was maintained across this geometry then the previous. It was worth noting that
the axial placement of the combustion line required more throttling but yet less pressure
was lost on average across the outlet ports. Figure 14 shows a preliminary design that

would use this orientation.

23

 

Figure 14. Off angle — single split, hardware design

EQUAL ANGLE - DOUBLE SPLIT

For manufacture and air cooling reasons later discussed it was determined that the
incoming air will need to be split three ways as Opposed to two. The combustion air is
still axial to the incoming air; however the bypass air is split at a 45 degree angle from
the center line with in the same plane (Figure 15). Having the bypass air split away from
the main inlet line at a steeper 45 degree angle causes less air to naturally bypass and
more air to ﬂow to the combustion. Again an analysis determined that the combustion
pipe must be throttled to roughly 21.8 degrees tO achieve the 0.030 kg/s flow rate through

the combustor and 65 percent of the total pressure was maintained from inlet to outlets.

24

As expected the other .041 kg/s Of air divides evenly through the two symmetry bypass

pipes.

C :ntrmrs )f Melanin Magmtlli‘ls~ I.T‘l/s}

 

Figure 15. Equal angle — single split, velocity contour

DIFFUSION LENGTH

Now that incoming air has been split and metered into combustion and cooling
bypass air, it was time to reexamine the combustion chamber design. This time the
amount Of ﬂow in the combustor has been reduced from the total mass ﬂow Of 0.071 kg/s
to only 0.030 kg/s. This means that the amount of mass ﬂow that needs to be decelerated
is less. The ﬂame speed Of propane has been calculated to be near 0.5 m/s, so any
combustion chamber design should be able to create low speed regions capable of
combustion [Kusner, 2006]. To assure this, a quick calculation can be done to set the

cross sectional area of the combustor necessary to decelerate the ﬂow to 0.5 m/s.

25

m =pAV
m =0.03O kg/s
p=2.189kg/m3

vair :vflamespeed '3 0-5 m IS

A = m = 0.0274 m2 = 42.47 in2

P Vﬂamespeed
: LG = 8.29 in

7r (6)

 

D

It was found that a 8.29-inch-diameter cross section correlates with the area needed tO
decelerate all of the air to be at or below 0.5 m/s. It would take an extremely long length
of ducting at this larger area for all of the high velocity area to decelerate and fully
develop at this slower speed.

In order for the entire combustor to be around 3 to 4 feet in length the diffuser
section must be limited to roughly a foot in axial length. A ﬁrst look was to use Fluent to
determine the velocity proﬁle Of a simple conical diffuser assuming that the diffuser is
restricted to 11 inches of axial length. It was determined that the geometry would be a 2-
inch-diameter cross section diffused to an 8-inch-diameter cross section, which is the
next nominal dimension larger than the idealized 8.29-inch-diameter needed to decelerate
the air to at or below 0.5 m/s. This geometry was meshed and run in Fluent with the 0.03

kg/s mass ﬂow expected in the combustor.

26

I; 7Ee+00
s 42e+00
6 09900
5 75.:ij
5 416+ 03
5 079+ 03
A 736* CO
A 409.90
4 Oée+CO

.3 729,00

0 ['05-’03
’E'I‘nlrs ‘If Velrut.‘ Magnitude Imisl j -_, - (a.
ILLE‘II .5 2‘ Irv-I 1p ‘59‘J‘rJEII—"i :.nj

Figure 16. Conical diffuser, velocity contour

7" Tinlet
oullel
7 036+ 30
3.000.000...

Lawn)

‘I CICIeHZO

4 L139- CO
VEIOCIt y

Magnitude
(m/s)

.3 I14 3 TR

Posmon I’m)

Ji—IOCCv Lingrxtuue

Figure 17. Conical diffuser, inlet and exit velocity proﬁles

The analysis was run as an axisymmetric model and was found to be turbulent

flow. The inlet velocity through the 2-inch-diameter pipe was 6.76 m/s for the 0.03 kg/s

 

mass ﬂow. As can be seen in Figure 16 there was a signiﬁcant region below 0.5 m/s.
Figure 17 shows the velocity proﬁles across the inlet and outlet cross sections as a
ﬁinction of their radial distance outward from the central axis. The outlet shows that at a
radial distance Of 0.06 meters (2.4 inches) the ﬂow of the air is at or below the 0.5 m/s
target velocity. However as expected, the velocity down the center Of the diffuser was
found to be around 1.5 m/s which is 3 times the velocity desired. This analysis proved
that slower velocities could be achieved with a shorter length Of diffuser than the
idealized 7 degree angle diffuser. The shorter length causes there to be a higher velocity
jet down the central axis that does not decelerate as well. This axial length is acceptable,
but more analysis must be done to determine what type of ﬂame holder must be used to

help the ﬂame attach and stabilize in the regions with velocities above the ﬂame speed.

LEADING CONE FLAME HOLDER

A seemingly logical place to locate a ﬂame holder would be as far upstream as
possible. This would allow for the best use of the available space. Placing the ﬂame
holder near the diffuser inlet should also help with difﬁision. In the previous analysis
there was a high velocity jet that did not fully diffuse. The ﬂame holder will also serve as
a splitter that will force the high velocity jet radially outward. Giving the incoming jet a
radial component should cause the air to make better use Of the expanding area and
decelerate further.

The same diffuser geometry was analyzed with the addition of the ﬂame holder
described. The mass ﬂow through the diffuser was again set to 0.03 kg’s (Figure18).

The results showed a great reduction in the high velocity jet. As predicted, the incoming

28

air quickly impinged on the conical shape of the ﬂame holder and deﬂected radially
outward. This resulted in a much larger region of velocity below the ﬂame speed.

7' 016+00
l3 668*00
6 319*00
5 959+OO
5 6‘9+OO
5 ZEEtOO
A 91e+00
4 5136900
4 219+:le

Z BONGO
2 459+OO

C. (Intcurs OI ‘I’EIGCII, Magmtude I‘m/s)

 

‘A‘I'
FLUENT >3 2 {an :1: says;

Figure 18. Leading cone ﬂame holder, velocity contour

Figure 19 is a plot analogous to Figure 17 Of the incoming and exit velocity
proﬁles for the difﬁiser. The velocity proﬁle across the entire outlet cross section is at or

below the desired 0.5 m/s goal.

29

' el
-' outlet
IOIIIOOIOI

6 009+00 ‘
s tun—«00

4 00940

VeIOCIty 7
MagnItude 0.3.403
(m/s)

‘ 039+UO

I 008+0‘3
Cl C'CetCCI '

TI 1‘14 I] 136

Posmon (m)

III.
FLUC‘IT .3 I 15:0 3: segrega

 

Figure 19. Leading cone ﬂame holder, inlet and outlet velocity proﬁles

Having a leading cone type ﬂame holder seemed promising. A closer looked
proved Otherwise. Showing the velocity magnitudes in vector form (Figure 20) indicates
having the ﬂame holder close to the inlet causes a large recirculation zone downstream.
The air is rapidly forced outward and soon circulates back inward to ﬁll the void caused
by the ﬂame holder. This inward swirling is noticeable throughout the entire length of
the diffuser. In fact a large portion of the diffuser’s inner ﬂow has a velocity magnitude
upstream. This upstream ﬂow causes large amounts Of shear against the rest Of the
downstream ﬂow. This shearing created a large and unwanted pressure loss, only 11
percent Of the total pressure was found to be maintained. The ﬂame holder design proved
to be capable of decelerating the ﬂow but there needs to be more guidance with how the

ﬂow is directed and expanded.

30

5 508*00
5 139200
A TTEHJO
A AOeHSD
4 036‘00

‘v'eIOCIIy ‘v’eclors COIS’E‘] B“, \Jeisclt, Iv‘agmtude ("I453

 

Figure 20. Lead cone ﬂame holder showing recirculation zones

GUIDED DIFFUSION

At this point in the design process several attempts were made tO make
consistently shaped cones and round ducts. The large diameter round ducts were easier
to make (Figure 21) with some minor but acceptable ﬂaws. The tapered cones required a
2-inch-diameter on the smaller end of the cone. The tool being used to form the sheet
metal was a simple hand cranked sheet metal roller. This roller could not provide the
force needed to roll the sheet metal in such a tight 2-inch-diameter cross section nor, were
the rollers small enough. Sending the parts out to be constructed professionally by a
vendor with a power roller was not within the budget Of this project. Therefore, an

alternative Option to using conical shapes was used.

31

 

Figure 21. Straight duct section as fabricated

A solution was to use a square cross section. This way all of the sheet metal
pieces would be ﬂat pieces that could be cut with any saw capable and then seam welded
together.

Now that it was known that square cross sections will be used, the analysis
process changed slightly. The previous combustion chamber designs were axisymmetric
and were analyzed as such in Fluent. The best way to represent square cross sections in
Fluent without going to a full three dimensional analysis is with a planar analysis just like
in the previous combustion bypass analysis. However this time, the planar mesh needs to
be analogous to a square cross section. Again a correlation factor needs to be determined
to relate the velocities calculated in Fluent with what the velocities would be in the true
geometry. This factor was calculated in exactly the same manner as it was for a round
cross section except the calculation for the true area is for a simple square and not a

circular cross section.

32

A....... = D2

mactual = P Aactual vactual = P D2 vactual
Afluent = D

mﬂuent = P Aﬂuent vﬂuent = P D Vfluent

mfluent = mactual = P D vﬂuent = P D2 Vactual
REDUCING T0

vfluent

V =
actual D (7)

As can be seen the true geometry velocity is analogous with the velocities Fluent
calculates but on a per width basis. Fluent represents the velocity as the true velocity
divided by the width of the plane the ﬂow is passing through. It should be noted that the
velocities and distances are true engineering units; in this case standard SI units are being
used throughout all analysis.

As a bench mark, an analysis was run showing just the main diffuser along with
the nozzle at the end of the combustor (Figure 22). This geometry was set up and run
without any ﬂame holder geometry to ﬁrst establish how the ﬂow reacts to the new

planar analysis, which represents the new square cross sections.

33

 

Figure 22. No guide vanes, velocity contour

The true geometry in mind is now an 8-inch by 8-inch square cross section as
Opposed to an 8-inch round cross section. In order to properly understand these plots the
8-inch width must be converted to SI units and used to divide out the Fluent velocities.
Equation (7) shows what the Fluent equivalent velocity is for the 0.5 m/s ﬂame speed
goaL

D=8in = 0.2032 m

YM-OSm/S

Vactual

vﬂuent = Vactua. : (0.5 m /s) (0.2032 m) = 0.102 m/s (8)

A velocity of 0.102 m/s shown in the plot above would be at or below 0.5 m/s in
the true geometry. Just like in the round cross section analysis, there was a signiﬁcant
region below the desired ﬂame speed temperature. Again there was a high velocity jet

region along the central axis that does not decelerate as well as the ﬂow further outboard.

34

As expected the square cross section design will need the addition of ﬂame holders
and/or guide vanes to encourage diffusion in a way that best conserves pressure.

Due tO the recirculation zone issues with the abrupt ﬂame holder in the
axisymmetric analysis, it was thought best to look further into guide vanes to enhance
diffusion. The guide vanes should gradually diverge the air entering the diffuser. By
doing so, the ﬂow should have a more even velocity distribution across the width of the

diffuser, which will reduce shearing and the likelihood Of recirculation zones.

 

Figure 23. Single guide vane, velocity contour

In the results above two simple guide vanes were placed at the inlet of the diffuser
to represent a 4-sided, pyramidal shape in true geometry (Figure 23). This guide vane
forced some of the ﬂow outboard as expected, which resulted in much lower velocities in
the main body Of the combustor.

The idea to use guide vanes was taken one step further with another smaller

divider added along the central axis (Figure 24). This divider served to diffuse and

35

separate the center jet even further, resulting in much better diffusion than the single
guide vane. In fact, after viewing the velocity profile (Figure 25) across the widest point
of the combustor it was noticed that the velocity was below 0.5 m/s (analogous with
0.102 m/s as plotted in Fluent) for the entire cross section. This means that all of the air
has been properly diffused, and it suggests that the propane ﬂame would be capable Of
sustaining in this environment. The total pressure that was maintained across the
combustor was quite similar for the single and double guide vane analysis, 34 and 33

percent respectively.

- CITI‘IIITQ ‘Il Irv—lent, TVJF’L}IJIIIII1?IW‘$VI

I’L‘JEII' t} 2 {Ed "i s»._

 

Figure 24. Double guide vane, velocity contour

36

' In el
outlet

Velocity
Magnitude
(m/s)

0

3
Positlon (m)

‘v ~=l3r Itv Magrntude

 

Figure 25. Double guide vane, inlet and outlet velocity proﬁles

CENTRALIZED FLAME HOLDER

Now that the velocity Of the combustor was within appropriate limits it was time
to bring some reality to exactly where the ﬂame should be initiated within the chamber.
Even though the previous guide vanes proved capable of decelerating the air they do not
provide a location for the ﬂame to initiate and attach. There needs to be a location for the
fuel to be injected, then mixed with the air and ﬁnally ignited. A design that can achieve
this would be a large pyramidal shaped ﬂame holder along the central axis. Within this
ﬂame holder the propane will be injected. Small mixing could be placed through the
holder to allow some air to mix with the fuel. Just before the fuel air mixture leaves the
ﬂame holder, it should be ignited. The tapered leading edge Of the ﬂame holder acts
much like the guide vanes in the previous analysis to diffuse the incoming air outboard.
The inside of the ﬂame holder acts exactly as a fuel injector nozzle and also as the point

Of ignition.

37

The plot below shows the geometry of the ﬂame holder. The analysis was again
run with the expected 0.03 kg/s of air ﬂow through the combustor. The leading contour
of the ﬂame created a large low speed zone well below the 0.5 m/s ﬂame speed down the
central axis Of the combustor. In this region the ﬂame will be ignited, and with these
lower velocities it should be able to stabilize and attach down stream of the holder.
Having the ﬂame centrally located in the combustor allows for the air ﬂow around the
outside to act as a thermal layer and protect the outer wall of the combustor from the

much hotter temperatures further inboard.

. , “It few-HI. I.’IEITDIIIII1I'-IIT“'SI

 

Figure 26. Centralized ﬂame holder, velocity proﬁle

This analysis solidiﬁed the ﬁnal combustion chamber design (Figure 26). The
ﬁnal design will have a single flame holder along the center axis with a 4-inch cross
section at its widest point. The top Of the ﬂame holder will taper upwards to the top
where the fuel will be injected at the leading most point. The main body of the ﬂame

holder will be held in place by a l '/2-inch pipe perpendicular to the ﬂow direction.

38

Within this support pipe is where the pilot ﬂame will be introduced to the inside of the
ﬂame holder. The fuel is injected within the ﬂame holder and forced downward and must
pass by the pilot ﬂame to exit the holder. This will guarantee ignition. On the ﬂame
holder, l/8-inch-diameter holes where drilled into the side walls Of the leading guiding
surface. These holes allow air to enter the ﬂame holder over an evenly distributed area.
This air is intended to premix with the fuel for better ignition.

As previously determined, the main body of the combustor will have an 8 inch
square cross section. The diffuser leading in to the combustion region will also be a
square cross section that increases from a 2-inch to 8-inch cross section just like the
analysis validated.

The bypass air was determined to be split from the combustion air in the equal
angle - double split analysis. This air will be introduced to the outer skin of the
combustion chamber. Speciﬁcally, the bypass duct is a 9-inch square cross section that
encompasses the 8-inch cross section of the combustion chamber. This creates a duct of
l-inch width around the perimeter of the combustion chamber wall. Directing the bypass
air through the duct allows the air direct contact to the outer wall Of the combustion
chamber to provide a source Of cooling. Cooling of the combustor is necessary to protect
fromthe high ﬂame temperature. Downstream of the ﬂame holder and expected ﬂame
front, the bypass air enters the inner combustion cavity through large 1 to 2 inch holes in
the wall between the bypass and combustor (Figure 29). At this point all the bypass air is
reunited with the hot combustion gas. The introduction of the bypass will now further
reduce the total air temperature to a desired 1200 Kelvin. At the bottom of the combustor

is a nozzle similar to the analysis, which returns the ﬂow back to a 2-inch diameter round

39

cross section. These hot combustion gases are then ready to be expanded across the

turbine. Figures 27 through 30 show the actual combustor as fabricated.

Bypass Air Bypass A”

Combustion
Air 7

Fuel
Injection

{Ti

' ' Flow
. ,. Direction

Pilot Flame
Entry Point

 

Figure 27. Centralized ﬂame holder, front view

40

Mixing
Holes

 

Figure 28. Centralized ﬂame holder showing mixing holes

41

f‘.‘ 3’; O .
Fuel
Injection _
0 kill 7 J . .

-I
.'.

v

"—5,,” p3“

- , I?

.‘ Ignition
" Point

.Ils‘

,3
is i
I I

I
. I
i.
I»
i

a:

Bypass Enters
Combustor

 

Figure 29. Centralized ﬂame holder, bottom view

42

Bypass/Combustion
Air Separation

 

Figure 30. Centralized ﬂame holder, full view

43

IGNITION SYSTEMS

IN TERMI TTEN T IGNITION

The ﬁrst combustor designs utilized commonly available automotive spark plugs.
Initially these spark plugs were located on the outer wall of the combustor near the fuel
injector (Figure 31). Like in an internal combustion engines, these spark plugs work
when coupled with a distributor that builds up an electric charge and then releases when
the circuit is closed. The discharge causes a hot electrical arc to span across the
electrodes of the spark plug several times a second. It is desired that the ﬂame in the
combustion chamber be continuous and therefore the periodic ignition of the spark plugs
is not ideal. The combustion chamber has been designed to have velocities at or below
the ﬂame speed of the propane fuel. This means that after the spark plugs ignite the
propane there should be regions that can sustain the ﬂame. At this point the ﬂame should
be self stabilizing and the spark plugs will only continue to ignite as a back up source of
re-ignition in the case of the combustor ﬂaming out.

This conﬁguration shown in Figure 31, with the spark plugs being located on the
outer walls of the combustor was tested. This orientation proved inadequate to ignite the
propane. The velocity Of the incoming air quickly carried the fuel downstream before it

had a chance to diffuse outwards and be ignited by the plugs.

44

 

Figure 31. Spark ignition with spark plugs on outer wall

The solution to this was to place a single spark plug directly downstream of the
fuel injection after the fuel has a chance to mix with some air (Figure 32). Even with the
spark plug running, it took several seconds for the propane to ignite and sometimes it
never did. The times it did ignite there was a loud bang followed by a large release of
pressure. This quick release robbed the system Of air leaving no chance of continued
combustion afterwards. For the spark plug to properly ignite the fuel, several things need
to happen. Not only does the fuel need to mix properly with incoming air but the ignition
must start from the small electrodes Of the spark plug. After these large explosions were
witnessed it was apparent that the spark plug was not a large enough source of ignition.
The ignition source should provide combustion instantaneously after fuel delivery. It is

dangerous to have unburned fuel accumulate in the chamber.

45

 

Figure 32. Spark ignition with spark plug inside ﬂame holder

CONTINUOUS IGNITION

It was decided to reject spark plugs as an ignition source all together. A pilot
ﬂame like those found on residential ﬁirnaces was considered. The main issue to
consider with using a small pilot ﬂame is that it is susceptible to being blown out by the
high velocity air. This means that the ﬂame itself must be robust and able to handle the
velocities calculated in the chamber design analysis. Many types of pilot ﬂames were
tested but the best source turned out to be MAPP gas mixed with pure oxygen. A small
bottle of MAPP gas and oxygen was able to be Obtained from a local hardware store as a
package set (Figure 33). This set was marketed with a small metal torch that regulates
the mixture of MAPP gas and oxygen. When a small amount of MAPP gas was mixed

with a large amount of oxygen, a very small blue ﬂame was achieved. Premixing the

46

MAPP gas with oxygen makes this pilot ﬂame robust. The MAPP and Oxygen mixture
was already properly set by regulating the two bottles. This allows the mixture to ignite
as soon as it is released from the nozzle tip and attach itself to the end Of the nozzle. The
pilot ﬂame attaching to its nozzle reduces the likelihood of blowing out.

The ﬂame torch was located downstream of the fuel injector in the same location
as the single spark plug was ﬁrst attempted. The pilot ﬂame was tested in the combustor
and proved to have no issues with igniting the propane. The fuel was instantly ignited
when introduced into the chamber and remained ignited. Now that stable combustion
was reached, the mass ﬂow was varied across a broad range to test for ﬂame stability.
The pilot ﬂame was even removed and stable combustion was still maintained. This
proves that the air in the combustion chamber was indeed diffused and decelerated
enough to allow for a stable ﬂame to exist without blowing out. It was decided that this

pilot ﬂame will be the ignition source for future testing.

47

 

Figure 33. Ignited pilot ﬂame with MAPP gas and Oxygen gas kit

STARTING SYSTEMS

GA TE VALVE

Combustion can not start the engine from rest by itself. The engine was started
with a high pressure air line. This air line was used to force air through the compressor
and then expand across the turbine. Under normal Operation, a compressor pulls forward
on the shaft. Forcing the air through the compressor in such a direct manner would cause
large stresses on it by pushing it aft instead. Severe loading on the compressor-turbine
couple was believed to be the result of previous testing, which caused the shaﬁ to break
at high velocity (Figure 34). Breaking the shaft caused the compressor to surge forward

into the inlet shroud, which ruined the blades.

48

 

Figure 34. Broken compressor-turbine shaft

At this point an identical compressor-turbine—bearing assembly was purchased.
To prevent this problem from happening again, a one—directional swinging check valve
was designed (Figure 35). This valve was placed in the engine ﬂow between the
compressor and turbine. The swinging gate only allows for the starting air to ﬂow
downwards towards the turbine. Now the turbine drives the compressor which in turn
pulls in ambient air. As the compressor is spooled up to speed, the starting air will be
reduced slowly and the swinging gate valve will progressively close. This will allow the
compressor fed air tO cycle through to the turbine, thus the engine begins to run under its

own DOWCI'.

49

     

-, J r : - . L '.
, ‘ . . . ._'- ' .r .
. »- - I
., a; . w
. . . _ ..

Figure 35. Swinging check valve used for engine starting

During testing it was found that the gate valve actually allowed the high pressure
starting air to ﬂow in all directions. This means that a signiﬁcant proportion of the
injection air was back-ﬂowing and pushing against the compressor. Enough of the ﬂow
went towards the turbine to spin the turbine-compressor couple in the correct direction.
However, the system was never able to accelerate up to the operating point because the
compressor was not capable of developing the pressure to push back against the 90 psi
start up air pressure. The startup air was capable of back-ﬂowing due to the design of the
air injector itself. The air was injected perpendicular to the intended ﬂow direction. The
idea was that this perpendicular ﬂow would actuate the swinging gate to open when the
start up air was being injected and direct the ﬂow downward towards the turbine. By
testing it was found that the injected air was quickly stagnated because Of the

recirculation zones caused by being injected perpendicular to the intended ﬂow. This

50

allowed the static pressure to build up in the system and the air was able to back ﬂow
around the cracks in the swinging gate towards the compressor and out the inlet.

The combustor was located directly below this air injector. The large increase in
area was also an issue with the back-ﬂowing. From the injector and upstream to the
compressor is a consistent 2-inch-diameter cross section. Downstream of the injector
towards the turbine the combustor drastically increases to an 8-inch square cross section.
Therefore, the start up air that properly ﬂows towards the turbine must ﬁrst be expanded
to this larger cross section and then contracted back to the 2-inch-diameter pipe before
entering the turbine. At the location of the air injection it is easier for the air to ﬂow out
towards the compressor than it is to exit out the turbine properly. It was concluded that
not only does the geometry of the start air injector need to be improved but also its

physical location in the engine.

TURBINE NOZZLE

In attempting to correct these issues a new air injector was designed. Due to the
expansion issues caused by the injection air ﬂowing through the combustor, it was
determined that any new design should be located downstream of the combustor just
before the turbine inlet. As a design constraint any starting system must be able to handle
the high temperatures caused by the combustion. Another issue was the high static
pressure build up that caused the air to back ﬂow in the check valve design. Two
modiﬁcations have been made to remedy this. First the air will be injected in the
downstream direction towards the turbine inlet, concentric and parallel to the intended
ﬂow of the engine. This prevents any recirculation zones and forces the air directly

towards the turbine with minimal static pressure build up that would encourage back

51

ﬂow. Also the air was injected through 3 3/8-inch diameter pipe. The old design injected
the air through a 1 inch diameter pipe. This smaller diameter forces the air to enter the
system at a higher velocity which increases the dynamic pressure, and as a result drops

the static pressure that previously led to back-ﬂow.

 

Figure 36. Turbine nozzle used for engine starting

This turbine injector nozzle showed several improvements over the previous
design when tested (Figure 36). The turbine quickly turned over with less air ﬂow than
the previous design took for the system to start spinning. The compressor instantly
started pulling in air from ambient. The previous injector design caused air to actually
ﬂow against the compressor. Therefore, this design clearly showed that it was capable of
accelerating the air in the proper ﬂow direction.

There were other issues when using this design to start the engine. For the ﬁrst
test runs the compressor-turbine couple was sped up to 70,000 rpm with the starting air.
At this point the combustion process was initiated. It was expected that the introduction

of the combustion should correlate with an increase in rpm, yet the rpm remained

52

constant even in the presence of the combustion. This suggested that the added energy
provided by the combustion did not correlate with an increase in turbine speed. A
hypothesis is that the turbine inlet nozzle may be choked.

The air was injected into the system through the high pressure line just upstream
of the turbine inlet nozzle and was quickly expanded across the turbine. The energy that
was extracted spun up the compressor. The compressor started pulling air from ambient
and passed it through the combustor. When mixed with fuel and burned, the air was fed
downwards towards the turbine. When this compressor fed air reached the turbine, it
combined with the starting air which was also being fed just upstream of the turbine. This
was potentially why the system was not able to increase in speed with the added
combustion.

The turbine nozzle was believed to be choked because at the turbine inlet the mass
ﬂow of the starting air and also the air from the compressor must pass. This would
explain why the added energy from the combustion process could not accelerate the
turbine; the turbine inlet was already passing as much air as it possibly could.

Having two sources of mass ﬂow when starting also created the issue of
transitioning from the starting air to the compressor fed air when stable combustion was
reached. When the starting air is removed, the operating point of the engine would be
quickly reduced causing a loss of stable combustion.

The proposed solution to these issues was another engine starting system
redesign. The new design should only force one source of air mass ﬂow through the
entire system. This eliminates the issue of choking the turbine inlet nozzle because the

amount of air used for starting the engine is also the air being used for the combustion.

53

This means that at any given time the mass flow of air across the compressor would be
the same as the mass ﬂow of air across the turbine, conserving continuity. The new
design should also aim to improve the process of reducing the starting air from the

system when the combustion process is sufﬁcient to run the entire system.

COMPRESSOR NOZZLE

A better design then the turbine nozzle starter was to inject the starting air through
the inlet of the compressor but in a safer manner than ﬁrst attempted. This solves the
issue of having two mass ﬂow sources feeding into the system because whatever would
be forced through the compressor would be pushed through the turbine. This also allows
for the total ﬂow of air through all points of the system to be measured, and reducing the
need for two ﬂow measuring devices to one. Also the transition period from starting air
to solely compressor fed air can be addressed by the axial location of the starting air
injection nozzle.

The air injection nozzle will be lined axially with the compressor inlet, but will be
placed several inlet diameters upstream of the compressor inlet. This will allow the
compressor to start pulling in air under its own power as the starting air is slowly
reduced, thus easing the starting air removal period.

Two unknown variables were how far upstream the starting air injector should be
placed from the inlet of the compressor and at what diameter the starting air should be
released from the injector nozzle. The distance of the injector from the inlet of the
compressor will be a trade off between the amount of starting air that makes it into the
compressor and the amount of air that misses the inlet completely. It could be guaranteed

that all of the starting air enters the compressor by placing the injector nozzle close to the

54

compressor. The trade off would be that the compressor will not be able to pull in air
from ambient under its own ability and the injector would create a blockage to the
compressor when the system is running under its own power. The other design constraint
was the diameter of the jet from the starting air injector. The diameter of this jet will be
driven by the exit diameter of the air injector. If the diameter of the injector exit is large,
the exiting ﬂow will be slower and as a result have a much higher static pressure. If the
nozzle is designed with a smaller diameter, the ﬂow will exit the nozzle faster and have a
larger dynamic pressure component. It was desired that the air enters the compressor in a
way that is similar in state to the air that would be entering the compressor during stable
combustion when at the Operating point. An analysis of the starting air injector nozzle
was conducted to answer these unknowns.

The ﬁrst decision was to determine at what diameter to eject the air from the start
air nozzle. The inlet of the compressor is 54-min in diameter, which is roughly 2 inches.
When air is released from the high pressure line it will form a high velocity jet. The
cross section of this jet will increase in diameter as the jet progresses axially, implying
that the jet should be released at a diameter smaller than the inlet diameter of the
compressor. To ease manufacturing a l-inch-diameter schedule 40 pipe was chosen as
the best choice among readily available pipe sizes.

The last step was to determine at what axial distance from the compressor inlet
the l-inch diameter pipe should be placed. To achieve this, an axisymmetric analysis was
modeled to represent this geometry (Figure 37, 38). On the left a nominal mass ﬂow of
0.05, 0.075 and 0.1 kg/s were ejected from the high pressure nozzle and allowed to freely

expand at atmospheric condition. On the right side of the mesh the geometry of the

55

compressor inlet was modeled. The outer perimeter of the mesh was modeled with outlet
boundary condition, creating a situation where the air ejected from the high pressure
nozzle was free to entire the compressor inlet or exit the analysis zone all together. For
the different mass ﬂow inputs from the high pressure line the amount of air that entered

the compressor inlet and the amount that missed were recorded.

:- , I".’II3.;I"IIIIII'II1II'II’S)

 

Figure 37. Compressor start nozzle, velocity contour

The axial position of the high pressure line from the compressor inlet was varied
from 8 to 14 inches in 2-inch increments. It was found that the percentage of mass ﬂow
that leaves the high pressure nozzle relative to the amount that reaches the compressor
inlet does not vary greatly based on the amount of mass ﬂow (Figure 39). In contrast it
was found that the axial position of the high pressure line has a dramatic effect. At 11.5
inches the amount of mass ﬂow that leaves the high pressure nozzle was equal to the
amount of air that reaches the compressor inlet. At an 8-inch axial position 125 % of the

mass ﬂow provided by the high pressure line reaches the compressOr inlet. This was

56

possible because the high velocity jet was entraining the stagnant air from ambient and
accelerating it towards the compressor. This added mass ﬂow comes at a penalty. The
high velocity jet was decelerated when it sheared with the stagnant ambient air, which in
turn was being accelerated. However, this was an acceptable penalty. The high pressure
line can provide more than enough air and it is more important to have the nozzle as far
away from the compressor as possible. Therefore, it was determined that setting the 1
inch diameter nozzle 1 1.5 inches from the compressor inlet in a concentric manner is the
best balance.

1 lie-+03
I EVE-3+0:

1039

:TII"'(:'1 R, ‘.’~-I’I'IlV Iu‘ilgt‘ui ldi’ I VIII.)

 

Figure 38. Compressor start nozzle, velocity vectors

57

STARTING INJECTOR, AXIAL PLACEMENT

 

 

 

80 l + 0.05 kg/s Mass Flow Percentages
70 . , , ~ , e ,, ~ , I + 0.075 kg/s Mass Flow Percentages
‘ 0.1 kg/s Mass Flow Percentages
0.05 kg/s Total Pressure Percentages

50 * * ,_ _ I.“ * +0.075 kg/s Total Pressure Percentages

 
 
 
 
     

Percent (%), Injector Nozzle to
Engine Inlet Flow

 

 

 

40 W , r J 3:70.?1 kg/sjotal Pressure Percentages
30 —- _ A , - — — —~ ~~~~~~~~~~
20__----- z__-___-_--
10*~****e _-A_Azz_._tm_w
0 . ..
8 9 10 11 12 13 14

Axial Distance From Compressor Inlet (in)

Figure 39. Compressor start nozzle, operating characteristics

TESTING CONFIGURATIONS AND RESULTS

DA TA ACQUISITION SYSTEM

The objective of this thesis was to design a test rig capable of running meaningful
and repeatable test runs. There are three major conﬁgurations of this test rig; the ﬁrst
was to run the combustion system solely by itself, then run the combustor with the
turbocharger, and then run the combustor with the wave rotor.

A ﬁnal fourth goal was to run the combustion chamber coupled with the
turbocharger and the wave rotor to test and validate these technologies. A logical step
was to ﬁrst design the combustion chamber and validate that it was capable of the

operating ranges desired to test the turbocharger and wave rotor.

58

The mass ﬂow rate desired for both the turbocharger and wave rotor was
nominally 0.071 kg/s. The combustor along with any of the acquiring devices should be
capable of a broad range above and below this nominal operating point. Several mass
ﬂow measuring technology types were considered. Devices that are capable of
measuring mass ﬂow directly are quite expensive. This is because they rely on multiple
types of technology within the same device to measure and then calculate the mass flow
rate, usually by measuring temperatures and inertial effects. A direct mass ﬂow sensor
was deemed too expensive to remain within the scope of this project. As a substitute a
volumetric ﬂow sensor was purchased with intent of externally calculating mass ﬂow.
Volumetric ﬂow rate is density dependent and therefore not a consistent way to measure
compressible flow. Mass ﬂow is simply the volumetric rate of the flow multiplied by its
density. To obtain the density of the ﬂow ideal gas was assumed. This was a strong
assumption because the medium being dealt with was air which should remain well under
supersonic speeds, certainly at the points where ﬂow rates will be measured. The ideal
gas law states that density is simply a function of temperature and static pressure. This
means that the static pressure and the temperature of the ﬂow, coupled with the

volumetric flow rate, will yield the mass ﬂow of the air.
V P3
_ = m
R T (9)

Even though three devices will be needed with this method, a volumetric flow

Vp:

sensor, a pressure transducer and a thermocouple, the total cost to achieve the mass flow

rate was still less expensive then a direct mass flow device. As can be seen in Figure 40

59

the temperature and pressure devices must be sufﬁciently close to the volumetric ﬂow

meter because the volumetric flow changes throughout the system.

 

Figure 40. Data acquisition devices: volumetric flow, pressure and temperature

A simple electric pressure transducer capable of outputting a signal proportional
to the pressure reading was used to measure pressure. Special care was taken to make
sure that only the static component of the ﬂow’s pressure was measured. This was
achieved by drilling a small hole perpendicular to the ﬂow through the wall of the piping.
It was important to ensure there are no burrs sticking inwards into the ﬂow. These burrs
could cause some of the dynamic pressure of the ﬂow to build up with the static pressure
tap and give a false reading.

The temperatures were measured by ungrounded K-type thermocouples. The
ungrounded feature of the thermocouple was desired due to the large amount of other
sensors that may electrically interfere with the readings. It also eliminated the need to
isolate the thermocouples from the metal test rig. Not only was the temperature needed at

the point of mass ﬂow calculation but the maximum temperature of the combustion was a

60

crucial factor as well. It must be monitored not only for matching the thermodynamic
cycle desired but to monitor the safety of the rig itself. The maximum combustion
temperature was expected to be 1200 Kelvin which is relatively close to melting limits of
the materials used to manufacture the rig. The K type thermocouple was chosen because
it has a range very close to the 1200 Kelvin expected in combustion but is also sensitive
enough to measure the lower temperatures expected at the mass ﬂow calculation points

(Figure 41).

 

Figure 41. Turbine inlet temperature thermocouple

Close control and monitoring of all testing was required. To meet this all of the
data must be captured and viewed in real time during testing. Therefore all of these

devices are managed by the use of a data acquisition board and fed to a computer for the

61

operator’s viewing (Figure 42). The computer managed these data inputs by a data
acquisition soﬁware program that was speciﬁcally tailored for this testing.

The data acquisition board has embedded cold junctions so the thermocouples can
be wired in differential mode with no external reference temperature needed. This board

was also capable of handling the ﬂow sensors as well as the pressure transducers.

ml --
mullﬁu“ - v.1

01"” “’3 1i til

:Q’H- rll

 

Figure 42. Data acquisition board and control computer

COMB US TOR TESTING

Previous testing of the original combustion chamber proved to be intermittent and
therefore prohibitive of further testing. This lead to the logical step to test this new
combustion chamber design uncoupled from the turbocharger or wave rotor. This way
the air delivery to the combustor was not dependent on the turbine-compressor assembly

for the turbocharger case or on the shock wave propagation for the wave rotor. Doing

62

this, any issues that were present with the combustion process could be isolated from the
rest of the system and better examined. With no compressor the air must be continually
fed to the combustor. This air will be supplied by the high pressure line available in the
test cell. The volumetric ﬂow meter along with the thermocouple and pressure transducer
will provide the mass ﬂow rate. Another thermocouple downstream of the combustor
will output what would be the inlet turbine temperature if the turbine-compressor
assembly was in place.

The goal of this testing was to prove that the combustor was capable of sustaining
a ﬂame at the desired operating point of 0.071 kg/s for the turbocharger and waver rotor
devices. The temperature of the ﬂow must be elevated to the turbine inlet temperature of
1200 Kelvin. If this ﬂow rate and temperature can be achieved, the test would be a
success.

The results of the ﬁrst test can be seen in Figure 43. The ﬁrst step was to light the
MAPP gas pilot ﬂame. Once this was done, the air ﬂow was slowly brought up to the
ﬁnal mass ﬂow operating point. At this time the fuel was added. Instantly, a low roar
was distinguishable followed by a sudden rise in the inlet turbine temperature
thermocouple. After this ignition was detected, the fuel source was throttled until the
desired 1200 Kelvin was achieved. This operating point was held for roughly a minute
and then the fuel was removed from the system. The temperature quickly dropped and
the air ﬂow was returned back to zero. The test was a success. After looking at the data
it was apparent that the chamber was not only capable of sustaining a ﬂame, but a
reasonable tolerance was able to be held on the inlet turbine temperature at the ﬁnal

operating point.

63

Combustion Test Run With Bypassing Air

 

 

 

 

__ '—£1.'T.Tf ,__ _ _____ _ __
1400 ”_f Mass Flow . . 0.500
—.071 k ls “ 1t 0.450
Q1200 ~ - ~9: .. ~ - J ~ ~1
V , 1- 0.400
931000 ' '5"
E . 0.350 E,
8 -, 0.300 "
E 800 - ,, - -, - ~ - 3
.2 0.250 1.1.
7 7 , 7 i 7 7 J, Q
g 600 1 0.200 3
.e 2
a 400 .#V ,, - - , - - a. g 0.150 2
1'5 - 0.100
.E 200 mm.
, 0.050
0 0.000
0 20 4O 60 80 100 120 140 160

Time (see)

Figure 43. Combustor only test with bypassing air present

After executing the ﬁrst test run it was noticed that the chamber seemed to have
no trouble holding a ﬂame without propagation. In the ﬁrst test 0.040 of the 0.071 kg/s
mass ﬂow was bypassed around the outer shell of the combustor. Due to the positive
results of the previous test a logical second test would be to see if the bypass air was truly
necessary. The bypass piping was removed, which forced all air to enter the combustor.
This increased the likelihood of ﬂame propagation and ﬂame-outs; yet there is a large
beneﬁt to this test. If the combustor is found to be capable of handling all of the ﬂow,
then the bypass system can be removed permanently. As previously analyzed, the bypass
system comes at the price of having to throttle the ﬂow. This loss in pressure could be

eliminated if the bypass is not needed. The bypass air does provide an amount of cooling

64

to the outer skin of the combustion chamber. When running without the bypass, the
temperatures must be closely monitored to protect the test rig.

This test was run in the same manner as the previous, lighting the pilot ﬂame ﬁrst.
The air was slowly brought up to the operating ﬂow rate and then the ignition was
started. Like the ﬁrst test the ignition instantly responded showing a quick rise in
temperature. Again the fuel was throttled back until 1200 Kelvin was achieved. This
temperature and ﬂow was held for roughly 2 minutes. For shut down, the fuel was ﬁrst
removed. When a temperature drop was noticed, the air ﬂow was slowly reduced to zero.
The test results (Figure 44) showed another success. The temperature proved to be
stabilized around 1200 Kelvin with no large ﬂuctuations. This test suggests that the
ignition system and the geometry of the combustion chamber were capable of handling
the entire mass ﬂow rate without bypassing any air. The outer skin of the combustion
chamber did not appear to be damaged by any elevated temperatures nor did any
thermocouples indicate violent temperatures due to the lack of bypass air. At this point it
was determined that the bypass will not be needed. This means that these pressures can
be maintained. The combustor chamber analysis was rerun (Figure 45) with the full
0.071 kg/s mass ﬂow to document these higher velocities experienced with no air being
bypassed. As expected, increasing the ﬂow rate through the combustor corresponded to a

proportional increase in velocity.

65

Combustion iTefstfRun With No Bypassing Air Flow
+|.T.T
1400 7- Mass Flow , , ~ -, . 7 7 7 ~ 0.500
5.0717kg/s7 ‘

    

1200 - 7,

 

g A
., 0.400 ﬂ
l- U)
3 1000 . - 0350 3‘,
E d)
a, H
,2 0250 3
_E 600 0.200 ‘5
U)
E 400 . . . . . . . 2 0.150 g
.. .2
,1; 200 ,1" 0.100 <

' ~ 0.050

0'7 2-22.2,,,zz,z,,-z0_ooo

0 20 40 60 80 100 120 140 160 180 200 220 240 260

Time (see)

Figure 44. Combustor only test without bypassing air present

Contours of Renal, ru‘lagmtude (misl

 

Figure 45. Centralized ﬂame holder operating at .071 kg/s, velocity contour

66

GAS TURBINE ENGINE TESTING

Once the combustion chamber had proven sufﬁcient for the operating point and
heat input the next step was to test with the turbocharger intact. As discussed in the
starting-systems section, all of the systems tested were capable of delivering air to the
combustion chamber. However, none of the starting system designs were able to take the
engine up to full operating point and run sustainably without starting air assistance.
When attempting to start the engine, there was always a point when the compressor-
turbine couple would not accelerate any further regardless of the amount of energy that
was added by increasing the fuel ﬂow. At this limiting point the compressor would
rapidly make a popping noise. This noise would be followed by a slight drop in ﬂow rate
through the system. The suspicion was that the surge margin of the compressor was
being reached and the noises being heard was the phenomenon known as compressor
surge.

Surging is the limit of a compressor to move and compress air. As air ﬂows
through a compressor, its pressure rises. This means that the air downstream of the
compressor has a higher total pressure than the upstream air. There is a limit to how
much ﬂow a compressor can handle aerodynamically. As more air passes through the
compressor, the aerodynamic losses are increased due to the higher velocities.
Eventually a point is reached when these losses are so great that the compressor cannot
compress any more ﬂow. When this occurs, the high velocity air will stagnate
downstream of the compressor. This stagnation causes the static pressure to rise
downstream of the fast spinning compressor blades. This air can now escape out the

front of the compressor because it is not being pushed downstream by newer compressed

67

air. When this air escapes forward, it slams the compressor rotor forward violently
against the volute.

To explore if this surging was taking place, the engine could be run in an open-
loop conﬁguration. By open loop it is meant that the compressor and turbine will not see
the same air ﬂow. The combustor will be routed with the turbine side of the
turbocharger. Air from the high pressure line will be fed through the combustor then
mixed with fuel before being ﬁred. Finally, the hot gas will be expanded across the
turbine and leave the system as exhaust. This process will in turn spin up the compressor.
Normally the compressor would be feeding air into the combustor. For an open-loop test
this air will be dumped to the atmosphere. For the engine to be able to run in a closed-
loop system the mass ﬂow of the compressor must match the mass ﬂow of the turbine.
This open-loop test will be able to measure the amount of ﬂow through the combustor-
turbine side which is being driven by the high pressure line manually. The test will take
the combustor-turbine side up to 0.071 kg/s and 1200 Kelvin then witness if the
compressor is capable of producing the 0.071 kg/s mass ﬂow rate needed to run the
system in closed loop.

The ﬁrst open-loop test was to see how much mass ﬂow could be generated by
the compressor when the turbine was being driven by the 0.071 kg/s ﬂow rate. This test
was without combustion. Due to losses in the system it was not possible for the
compressor to be able to move as much air as the turbine without any additional energy
input from the combustion. This test was intended to serve as a bench mark for the true

combustion tests.

68

With no combustion, the process for running this test was quite simple. The mass
ﬂow through the combustor-turbine couple was slowly increased. During testing the
mass ﬂows of both the compressor and turbine side are being acquired. As seen below
(Figure 46) a plot was generated with the resulting data that shows the difference in mass
ﬂow between the turbine and the compressor sides. With no combustion the compressor
seemed only to be able to output roughly half the mass ﬂow that was passing through the
turbine side. This mass ﬂow mismatch is the energy loss the combustion must be able to
overcome. It should also be noted that the popping sound believed to be surging was not
present during this test. This could be because the compressor was operating at a much
lower ﬂow rate than it was when combustion was present during the attempts to start the

engine as a closed-loop system.

Open Loop Test Without Combustion - No Oriﬁce Plate

+ Burner Mass Flow ‘

 

 

0'130 "‘+Compressor Mass Flow"'*""*-—- *u-ie r~-—»—~ 160000
0120 “-4 Bgmeietgeasagssme - a a v-
0.110 2. 4*: 2p 2 ;- ;:- A- -- a 2 140000 E
3
0100 F F i ' ‘ ' ” ***** * * ’ ‘ 120000 3
130.090 e- - . - - . . 3
E0080 . - 100000 2.
30.070 9;
3 —~ 80000 .,
00.060j , '5
30°05" 4 60000 g
2 0.040 ~» - E
0.030 ‘ - 40000 E
O
33%” ~20000 '-
0.000 , 0

 

 

O 1 0 20 30 40 50 60 70 80
Time (sec)

Figure 46. Open-loop turbocharger test without combustion, mass ﬂow and pressure

69

As can be seen by the results of the ﬁrst test, there was no signiﬁcant rise in total
pressure from the compressor. The total pressure in the turbine side was found to be
roughly 30 % higher than the pressure in the compressor side. In fact the compressor did
not appear to raise the pressure of the incoming air much above atmospheric pressure.
This low pressure would be an issue when attempting to run the rig in a closed system
conﬁguration. With no combustion it takes a greater amount of ﬂow and pressure to
yield the same results then if more energy was present in the form of heat addition.

Due to this lack of pressure it was determined that an oriﬁce plate should be
installed at the compressor discharge. By forcing the compressor to discharge through a
smaller oriﬁce the compressor must work harder and develop a higher static pressure.
This should match the pressure through the turbine side better.

The normal discharge area for the compressor was a standard 2-inch-diameter
pipe. Several iterations and small tests were run to determine that a l-inch-diameter
oriﬁce at the compressor discharge would be a conservative starting point. Using an
oriﬁce plate that was too small could cause damage to the system. If too much ﬂow was
blocked, the compressor would not be able to discharge air as quickly as needed. This
would result in a spike in pressure which could be catastrophic to the rig. If the
compressor does not respond as expected with an increase in static pressure, another
smaller oriﬁce plate could easily be installed instead.

The next test with an oriﬁce plate in place was run exactly like the previous
(Figure 47). The air was increased until the desired ﬂow rate was achieved. Meanwhile
static pressure, temperature and mass ﬂow were being acquired for both the compressor

and turbine. As can be seen in the results the compressor still fell short of matching the

70

mass ﬂow through the turbine. This was expected like it was in the previous test due to
the absence of combustion. What was most notable was that the compressor side seemed
to still have little pressure rise even with the introduction of the oriﬁce plate. This
suggests that the oriﬁce plate should be smaller than the 1-inch-diameter. However the
introduction of combustion could change this. It was decided that testing with

combustion should be run before a decision to reduce the oriﬁce plate would be made.

Open Loop Test Without Combustion - Oriﬁce Plate

 

 

  

 

 

 

0.100 140000
0.090 2 , -
- 120000 ’3
0.080 ir‘ .tﬂ—‘ﬂ 5
. . :,,___,_- o
130070 . - 100000 .3
3’ 0 060 - - -
; - .- “3» 80000 3
20.050 ~~ ~ A * _«z —* a 7* -2, n 7 ~- P~H ‘5'
IL L-
- 60000 3
§0040 -4, -~- J-“ 11-!“ g
2 0.030 e f * f . n- ~~ ~77 - *7 40000 5
0-020 - *A *‘* m* r' : BTJrner Mass F1891? “—711 §
1+Compressor Mass Flow ~ 20000 1-
0.010 1 e r ' ~ Burner Pressure 1*
t . CPmPLGSSOLPVGSSUre ,1
0.000 - T‘ T“ ’T‘ 2 ”T “T" * 0
0 50 100 150 200 250 300
Time (sec)

Figure 47. Open-loop turbocharger test without combustion, mass ﬂow and pressure

The next test was to add in the combustion process and determine if the
compressor was capable of producing a matching ﬂow rate to the combustor-turbine side.
The introduction of combustion compensated for the mechanical losses in the system and
allowed the compressor to meet or exceed the mass ﬂow rate of the turbine. In order for
the system to be run in a closed loop as an engine the pressure created by the compressor

should also meet or exceed the turbine side to maintain ﬂow in the correct direction.

71

The testing process was quite different when combustion was taking place. First,
the pilot light was lit. Then the air being forced through the combustor-turbine side was
brought up to about .040 kg/s, about halfway to the ﬁnal operating point. The idea was to
start the combustion at this lower mass ﬂow rate of 0.040 kg/s to ensure sustainable
combustion. After ignition the mass ﬂow through the combustor was walked up to the
ﬁnal 0.071 kg/s operating point.

When the air reached 0.040 kg/s, the ignition process was started. As can be seen
in Figure 48 and 49 this happened at roughly 50 seconds into the test. At this point the
compressor-turbine couple quickly sped up, which was clearly audible as a large whining
sound and evident in the results. At the point of ignition the results showed that the
compressor was outputting roughly 0.035 kg/s while the turbine was outputting 0.04 kg/s.
Even though matching was not achieved, the combustion process brought the mass ﬂows
much closer. Even though small, there was an increase in pressure on the compressor

side that correlated with the introduction of combustion.

72

0.075
0.070

0.065 , ‘

0.060
A 0.055
a 0.050

2 0.045 ,

g 0.040
a. 0.035
g 0.030
g 0.025
0.020
0.015
0.010
0.005
0.000

Open Loop Test With Combustion - Oriﬁce Plate

 

 

 

- + Buinér Mass FLovv — ,, 7 .7 - ,ﬁ. 1200
r Compressor Mass Flow, ,
ILT.T.7 7 W W , 7
0 20 4o 60 80 100 120
Time (s)

Figure 48. Open-loop turbocharger test with combustion, mass flow and I.T.T.

Open Loop Test With Combustion - Oriﬁce Plate

150000 «:BurinerPressutei i 1200

140000

r Compression Pressure
I.T.T.

 

 

 

379‘ 1000

2

8 130000 ,, , , ,_, 7, ,. 7, , , ,

“a 800

9‘: 120000

2 600

3 110000

0)

2 400

E 100000

(B

19 90000 ,, 200
80000 0

 

0 20 40 60 80 100 120 140
Time (s)

Figure 49. Open-loop turbocharger test with combustion, pressure and I.T.T

73

Inlet Turbine Temperature (K)

Inlet Turbine Temperature (K)

Unfortunately, the compressor only sped up for a short while before the loud
popping sound was witnessed again. This time the popping was followed by a sudden
lock of the compressor. Once it was noticed that the compressor was no longer spinning
even though there was still ﬂow through the combustor, the system was quickly shut
down. After disassembly it was apparent that the compressor had slammed forward
against the volute. This clearly supported the idea that the compressor reached its limit
and surged. Fortunately, this surge did not break the shaft between the compressor and
turbine. Removing the volute on the front of the turbocharger allowed for the compressor
to be freed. The compressor blades were slightly damaged by coming in contact with the
volute. This damage to the blades will affect the performance of the compressor and will
most likely increase the probability of surging again.

Due to the surge, the operating point of 0.071 kg/s was never achieved. Even
though surging was again likely another test was run to determine if the compressor truly
reached its surge limit or if the previous test was a one-time irregularity.

The second test was run in the same manner (Figure 50, 51). Like in the previous
test the introduction of combustion caused the compressor to start popping loudly. Only
this time the compressor did not lock into the volute. It continued to pop once or twice
every few seconds. This popping took place for roughly 20 seconds until the mass ﬂow
in the compressor side began to drop rapidly. Correspondingly, the ﬂow through the
turbine began to rise. This would suggest that the shaft had broken. After the test was
shut down, the turbocharger was again disassembled and it was conﬁrmed that the

compressor had surged so violently that the shaft broke.

74

Total Pressure (Pa, absolute)

Open Loop Test With Combustion - Oriﬁce Plate

0.110 éié—iBJrnerh/lasﬁs Flow 1 n
0,100 - r Compressor Mass Flow‘ .2
7 7|.T7.T.7 7 ‘

 

 

0 20 40 60 80 100 120 140 160
Time (sec)

Figure 50. Open-loop turbocharger test with combustion, mass ﬂow and I.T.T.

Open Loop Test With Combustion - Oriﬁce Plate

 

 

200000 r-o-FBuirner Pressure 7 —— 2-, ,, 1400
180000 + Compressor Pressure , 7 77 I
I.T.T. 1200
160000 , : f 7’ ’ -, 7 , , , ‘
140000 , . , ,, w ’11 I 7 7 1000
120000 , , - , , , 7 ,
100000 , _ 7
80000 I ' 600
60000 , ,, , , 7 i i 400
40000 '7 7 W ,
i 200
20000 ' ~ ,, ,,
0 0
0 20 40 60 80 100 120 140 160

Time (sec)

Figure 51. Open-loop turbocharger test with combustion, pressure and I.T.T.

75

   

Inlet Turbine Temperature (K)

Inlet Turbine Temperature (K)

WAVE ROTOR TESTING

The other half of the combustion chamber and test rig design was for the
evaluation of the wave rotor. The wave rotor was tested in the same open-loop
conﬁguration as the turbocharger, meaning there was a separation of the compression and
expansion ﬂows of the system (Figure 52). Like in the turbocharger tests the combustion
chamber was routed upstream of the expansion side of the wave rotor. The compression
inlet was allowed to pull air freely from ambient. The hot combustion gas expanded in
the wave rotor driving the compression process. Fresh air was pulled in and compressed.

This compressed air was ejected from the wave rotor.

 

Figure 52. Wave rotor test apparatus
The ﬁrst step was to light the pilot light. Once the pilot light was lit, the wave

rotor was sped up to 13,000 rpm by an external electrical motor. The high pressure line

76

was again used to force air through the combustor and to the expansion ports of the wave
rotor. The air in the combustor was ﬁrst brought up to the desired mass ﬂow rate. At this
point the fuel was added. As expected, there was a low thud audible and combustion was
apparent. Being the ﬁrst ever test run with the wave rotor, it was determined to ﬁrst
operate the wave rotor with combustion gases no higher than 900 Kelvin. This operating
point was sustained for roughly 30 seconds and then the test was systematically shut
down. Fuel was removed and then the wave rotor was gradually returned to zero rpm
while the air ﬂow through the combustor was being shut off. When safe to approach the
rig, the last step was to turn the pilot off.

The types of parameters acquired for the wave rotor are identical to previous tests.
The pressure, temperature, and mass ﬂow of the expansion side were acquired as well as
the pressure, temperature, and mass ﬂow of the compression side. In the results below,
the mass ﬂow through the expansion side was much higher than the mass ﬂow through
the compression side, roughly 8 times as much. When at the proper operating condition,
the mass ﬂows should be roughly equal. Data suggest the point at which this particular
test was conducted was not the optimal one (Figure 53, 54). However, the wave rotor did

show that it was capable of ingesting and compressing some amount of ambient air.

77

0.075

0.070 ,

0.065

0.060 »

... 0.055
a 0.050

5 0.045 -

g 0.040
u—_ 0.035

'0 0.030 <

“ 0.025
0.020

0.015 .

0.010
0.005
0.000

150000 - Compression Pressure

Total Pressure (Pa, absolute)

80000 -

Open Loop Test With Combustion - Wave Rotor

 

 

+ Burner MassiFLow
' ' + Compressor Mass Flow“ T

   

* 1000
900
800
700
*H 600
500

P 400

300

,, — 200

- 100

 

 

140000
130000
120000
1 10000 -

l

1
100000 ’
90000 [w
o

0 5 10 15 20

Time (s)

25

30

Figure 53. Open-loop wave rotor test with combustion, mass flow and I.T.T.

Open Loop Test With Combustion- Wave Rotor

+ Burner Pressure

 

7 I.T.T;

  

 

5 10 15 20
Time (s)

25

30

35

Figure 54. Open-loop wave rotor test with combustion, pressure and I.T.T.

78

Inlet Turbine Temperature (K)

Inlet Turbine Temperature (K)

The test rig proved capable of running the wave rotor at high speeds and
delivering the combustion gases needed to evaluate the technology, which was one of the
primary goals of this test rig. With the test rig procedure set, the next step for the up
coming generation of tests is to determine a series of operating ranges and conditions to

explore and optimize the capabilities of the wave rotor technology.

CONCLUSIONS

In design and implementation, the test rig is considered a success. The
combustion chamber proved to be capable of the operating ranges demanded by the
turbocharger and wave rotor test plans. Decelerating the combustion air over a larger
cross sectional area, introducing a continuous pilot ﬂame, and introducing the central
ﬂame holder proved to be crucial aspects to the success of the combustion chamber
design.

In the closed-loop conﬁguration the turbocharger was not capable of accelerating
up to the speed and mass ﬂow necessary to operate the test rig as a 5 kW gas turbine
engine. An open—loop test showed that the engine did accelerate in the presence of
combustion. However, this acceleration did not carry the engine up to the desired
operating point. Before the operating point could be achieved the compressor repeatedly
surged causing a sudden deceleration and the eventual destruction of the compressor —
turbine shaft. This turbocharger ran close to its surge line at the operating point. Another
type of turbocharger where the operating point could be placed further away from the
surge line would be more desirable.

The wave rotor testing conﬁguration has been established. This preliminary

testing proved that the test rig and procedure were capable of providing the wide ﬂow

79

rates and combustion temperatures needed to test the wave rotor. Several test runs have
been conducted at this point. The next step is to run the wave rotor at various operating
conditions and map the results. These results are crucial for understanding and validating
the performance enhancing ability of the wave rotor technology by determining the
operating conditions at which the beneﬁts exist. The rig and testing procedures have

been set. This next phase of testing can now be conducted.

RECOMMENDATIONS

There are several reasons that the current turbocharger was experiencing
compressor surge before achieving the operating point. The ﬁrst was that the operating
point chosen was very closely to the surge limit for this chosen turbocharger. This surge
limit was deﬁned by the compressor map, which comes as product data that was
produced by testing from the turbocharger manufacturer. One reason the turbocharger
may be experiencing surge at a lower operating point than indicated by the factory data
could be due to inlet distortion. This means that the incoming ﬂow to the compressor
was probably much straighter and conditioned by a special shaped inlet bell mouth, or
other geometry, when the compressor map data was generated by the manufacturer. The
current test rig does not utilize any type of inlet geometry to limit inlet distortion other
than the volute itself. For example, incorporating an A.S.M.E. bell mouth design would
drastically improve the incoming ﬂow of air to the compressor and decrease the
likelihood of surging the compressor [A.S.M.E., 2007]. The A.S.M.E. standard bell
mouth shape is widely proven and used in the aerospace industry. The speciﬁc geometry
of the bell mouth is a simple elliptical shape. The exact dimensions of the inlet are

deﬁned characteristically by the diameter of the inlet within the A.S.M.E standard. It is

80

highly recommended that this bell mouth be manufactured and implemented for any
further engine testing. A view of what an A.S.M.E. standard bell mouth would look like
for the current Garrett turbocharger being used can be seen in Figure 55. If it is
determined that a different turbocharger will be used for subsequent testing, then the
exact dimensions will simply change proportionally relative to the increase or decrease of

the inlet diameter.

 

Figure 55. ASME bell mouth scaled for use with gas turbine engine, cross section view

During testing the ﬁrst and last step was always to manually light and then
extinguish the pilot light for the combustor. It would be an advantage to be able to
automate this process because lighting the pilot requires coming in contact with the test
rig. The ultimate goal for this test rig is to be run from one central computer station
safely away from the rig itself. Therefore a device that could replace the current MAPP
gas pilot light that could be controlled electronically from a remote location would be
best. A device that could achieve this is a hot surface igniter. This igniter is commonly
used for igniting residential furnaces for home heating. The igniter is a simple heating
element that acts as an electrical resistor and heats up when energized with standard 110

Volt AC power. There are several variations of this type of product on the market.

81

Figure 56 below shows a purchased igniter with the element energized. This igniter is
already located in the lab and simply needs to be wired and mounted within the
combustion chamber and routed to the computer control station with an appropriate
switch for control. This igniter has been tested and proven to be capable of igniting the

propane.

 

'- m . -. .20‘
Figure 56. Hot surface igniter commonly used in residential furnaces

Another step towards automating the test rig would be to incorporate a fuel
control system that could be controlled from the command computer. The current fuel
control meter can be computer controlled but was found to not have the capacity to
deliver the mass ﬂow rate of fuel needed. A ﬂow meter, similar in technology to the
current meter, should be obtained but capable of a larger mass ﬂow rate should be

obtained.

82

Also, the oil pump that feeds oil to the bearings for lubrication and cooling could
be automated. A simple electric motor with a potentiometer for speed control could be
coupled to the shaﬁ of the oil pump. Currently, the oil pump is driven by an AC drill and
the speed can not be altered with out ﬁrst coming in contact with the test rig. A new
motor with a potentiometer could be wired to the control station and modulated from
there.

During testing it was noticed that there were some small air leaks around the
ﬂanges of the combustion chamber. To prevent this leaking the ﬂanges should be sealed
with some type of gasket material. The combustion chamber is capable of reaching 1200
K near the outer wall; therefore any gasket chosen must be capable of this.

The last modiﬁcation recommended would be to install a ventilation system in the
lab capable of removing the exhaust gas from the lab area and expelling it outside.
Currently, the exhaust gas is routed out of the test lab through metal ducting. However, it
is inevitable that some of the exhaust gas does not completely leave the testing area and
this ventilation system should be capable of circulating these gases out.

For the wave rotor the ground work for testing has been set and ﬁiture testing with

this rig can now be conducted to further develop and understand this novel technology.

83

WORKS CITED

. Akbari, P, Nalim, R, Mueller, N. “Performance Enhancement of Microturbine

Engines Topped With Wave Rotors.” Journal of Engineering for Gas Turbines
and Power. 128 (2006): 190-202.

. ASME MFC-3Ma—2007, “Measurement of Fluid Flow in Pipes Using Oriﬁce,
Nozzle, and Venturi.”

. Dempsey, E, Mueller, N, Akbari, P, Nalim, M R. “Performance Optimization of
Gas Turbines Utilizing Four-Port Wave Rotors.” AIAA Pap. 2006-4152.

. Garrett. Garrett Product Catalog.

<http://www.turbobvgarret.com/turbobvgarrcttr’catelog/ turbochargers/ gt 1 2/gt l 24 l
_756068 l.htm>

. Kusner, M, “Design of a 5 kW Microturbine Generator.” East Lansing, MI:
Michigan Stat University, 2006.

84

BIBLIOGRAPHY

. Akbari, P, and Szpynda, E, Nalim, M R. “Recent Developments in Wave Rotor

Combustion Technology and Future Perspectives: A Progress Review.”
AIAA2007-5055.

Beer, F P, Johnston Jr, E R, DeWolf, J T. Mechanics of Materials. 3rd ed. New
York: McGraw-Hill, 2001.

Cumpsty, N. Jet Propulsion. Cambridge: Cambridge University Press, 1997.
Davison, C, and Birk, A M. “Set Up and Operational Experience with a Micro-

Turbine Engine for Research and Education.” Proceedings of ASME Turbo Expo
2004, GT2004-53377.

Dixon, S L. Fluid Mechanics, Thermodynamics of Turbomachinerjy. 4th ed.
Massachusetts: Butterworth-Heinemann, 1998.

Incropera, F P, DeWitt, D P. Introduction to Heat Transfer. 4th ed. New York:
Wiley, 2002.

Kuo, K. Principles of Combustion. New Jersey: Wiley, 2005.

Moran, M J, Shapiro, H N. Fundamentals of Engineering Thermodvnamics. 5'h
ed. New Jersey: Wiley, 2004.

Mueller, N and F rechette, L. “Performance Analysis of Brayton and Rankine
Cycle Microsystems for Portable Power Generation.” ASME International
Mechanical Engineering Congress & Exposition. New Orleans, LA, 1—10.

10. Nijeholt, J, Komen, E, Verhage, A, van Beek, M. “First Assessment of Biogas

ll.

12.

13.

Co-Firing on the GE MS9001F A Gas Turbine Using CF D.” Proceedings of
ASME Turbo Expo 2003, GT2003-38804.

Potter, M, Wiggert, D. Mechanics of Fluids. 3rd ed. California: Brooks/Cole,
2002.

Treager, 1. Aircraft Gas Turbine Engine Technology. 2nd ed. New York: McGraw-
Hill, 1979.

Tsai, L. “Design and Performance of a Gas-Turbine Engine from and Automobile
Turbocharger.” Cambridge, MA: Massachusetts Institute of Technology, 2004.

85

IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII

1000000