• Design and Rocket-based Flight Testing of a Thermoelectric ...

Sponsored Links

Download the ebook

Thermo Electric Motor
Vanderbilt Aerospace Club: Post Launch Assessment
May 7 2010
Table of Contents
Attestation 2
I. Introduction 3
II. Background 3
III. Design Challenge 4
IV. Rocket Vehicle Summary 5
V. Rocket Flight Simulation Summary 7
VI. Payload Description
A. Data Acquisition System 8
B. Payload Design 9
C. Payload Assembly 9
VII. USLI Flight Summary 11
VIII. USLI Payload Scientific Data Analysis
A. Raw Data & first steps to Data Processing 13
B. Comparisons with Ideal TEG Engine Operation 15
C. Conclusions from Data Analysis 18
IX. Overall Experience and Lessons Learned 19
X. Outreach Summary 20
XI. Acknowledgements 20
XII. Appendix
A1. Manufacturer’s Specs on TEG devices used in the payload 21
A2. ZT values for TEG device used in the payload 21
A3. Thrust Curve for Cesaroni L610 Motor 22
A4. Wind Tunnel Measurement of CD for Rocket Model 22
A5. Payload wall thickness measurements 23
A6: Offset of the Nozzle exit plane & misalignment w/ rocket axis 23
A7: Power Generation Efficiency Spectra of various engines 23
List of Figures
Figure 1: Picture sequence showing the development of the USLI payload 4
Figure 2: Schematic of USLI payload 5
Figure 3: Competition Rocket and RockSim schematic 6
Figure 4: Predicted trajectories for March 20th & USLI launch 7
Figure 5: Thermocouple placement on the payload 8
Figure 6: Schematic of the data stream for one set of TEGs 9
Figure 7: Pictures of the two independent DAQ boards showing components 9
Figure 8: Isometric view and the actual machined, assembled payload flown at USLI 10
Figure 9: Vanderbilt Team with the post USLI Rocket & Payload 11
Figure10: MAWD Altimeter data showing apogee, flight speed & acceleration 12
Figure 11: Hot and Cold temperature data across TEG during USLI flight 13
Figure 12: Power generated by the two TEG streams 13
Figure 13: Take-off signature of exhaust jet with respect to rocket axis 14
Figure 14: Total power generated by the two TEG streams till apogee 14
Figure 15: Total power generated as a function of temperature difference 15
Figure 16: Carnot & Seebeck efficiencies for the TEG Engine USLI Payload 16
Figure 17: Overall Thermodynamic Efficiency of the TEG Engine Payload 17
Figure 18: Normalized Payload Energy Conversion improving with Flight Speed 18
List of Tables
Table 1: USLI Launch Vehicle & Motor Details 6
Table 2: USLI Launch Vehicle Cost 6
Rocket-based Studies of Thermoelectric Exhaust Heat Recovery in Aerospace Engines:
Design of a Thermoelectric Engine
I. Introduction
Thermodynamics, aerodynamics, combustion and heat transfer, to name a few fields have allowed
us to perfect engineering systems that have been continuously developed through the dawn of the industrial
era. As we move forward, we are faced with new challenges: finding renewable sources of energy, and
improving energy efficiencies of innumerable systems uniquely designed through man’s ingenuity.
Thermodynamics clearly points out that not all heat energy can be converted to work; nowhere is
this more apparent than in the excessive heat (~50%) carried away by the exhaust of jet engines. A Boeing
747 on an intercontinental flight can run for fourteen hours at a stretch and consume 50,000 gallons of fuel.
Similarly, the Global Hawk UAV can run continuously for twenty fours at a time. It is tempting to ask: can
a small portion of the exhaust heat energy be economically recovered to generate electrical power?
The Vanderbilt Aerospace Club set out to design a Thermoelectric Engine for Aerospace
applications. It is an engine with no moving parts which can convert some of the waste exhaust heat into
usable electrical power. The concept thermoelectric engine is built using thermoelectric generators (TEG):
solid state devices that can directly generate electric power across a temperature gradient. Essentially they
work on the principle of the Seebeck effect which states that free charge carrier transport results from
thermal transport. Older Seebeck-based devices used bimetallic junctions and were bulky while more
recent devices use bismuth telluride (Bi2Te3)-based semiconductor p-n junctions which can have
thicknesses in the millimeter range, and can be engineered to have higher energy conversion efficiencies
(appendix A1).
It was proposed that the thermoelectric engine performance can be evaluated through a rocket
flight, albeit only for a short time. We argued that the rocket flight would help check out the role of flight
speeds, atmospheric conditions, and launch-g effects on the design and viability of such devices and
engines. The alternate test bed of a real jet engine, while necessary, can be tremendously expensive; rocket-
based testing can provide a low-cost access to typical flight conditions and can serve as the first test bed for
Our extensive results from the April 17th USLI flight, which was a culmination of a year-long
design, testing and development process, has clearly proved that rocket flight can be used to test out the
concept design, and that flight speed has a major role in the performance efficiency of these engines.
Progress made through such studies can also have implications in the design of semi conductor-based
thermoelectric engines for interplanetary and cosmic vehicles, where power can be generated by exploiting
the temperature difference available between a source (nuclear fuel) and the cosmos.
II. Background: Thermoelectric Devices & Thermoelectric Engine
When a temperature difference is sustained across the walls of a thermoelectric device (Th & Tc for
the hot and cold sides), the device converts heat energy to electrical energy with an efficiency given by:
⎛ ⎞
⎛ T − Tc ⎞⎜ ZT + 1 − 1 ⎟ (1)
η =⎜ h
⎜ T ⎟.⎜

⎝ ⎠⎜ T ⎟
c ZT + 1 + h ⎟
⎝ Tc ⎠
(*J.G. Snyder, ‘Small Thermoelectric Generators’ Electrochemical Society Interface 2008, pp. 54-56)
The first term in equation 1 is the Carnot efficiency for a heat engine operating between temperatures
Th and Tc, and the second term is the Seebeck efficiency of the thermoelectric device with ZT being the
figure of merit of the device. The figure of merit for thermoelectric devices is defined as ZT = σS T , where
σ is the electrical conductivity, κ is the thermal conductivity, and S is the Seebeck coefficient or thermo-
power (conventionally in μV/K), and T is the average temperature across the device ((Th + Tc) / 2). Greater
values of ZT will result in greater thermodynamic efficiency. The off-the-shelf thermoelectric devices
chosen for the construction of the thermoelectric engine are made of Bismuth Telluride and have values of
ZT close to 1 (see appendix A2). Current research in thermoelectric materials has focused on increasing the
Seebeck coefficient, and reducing the thermal conductivity through manipulation of the nanostructure of
the materials.
III. Design Challenge
Designing a payload for a short-duration rocket flight which is to be mounted in the jet exhaust poses
unique challenges:
(i) There is a possibility of build up of back pressure at the supersonic nozzle exit, and through this a
tremendous loss of thrust (Krushnic effect). The Vanderbilt team conducted a series of systematic
groundbased tests to establish the right kind of venting in the payload tube, at the rocket nozzle
exit, to eliminate the deleterious effects of the back pressure.
(ii) To ensure that the right hot side temperature is available across the TEG through optimum
interference of the exhaust jet. Both our analytical modeling and two flight tests prior to USLI
showed that there is no measurable loss of thrust due to this jet interference. Further, preliminary
power generation results from our flight test on March 20th helped us to refine the position of the
TEGs on the payload tube, and to select better heat dissipation fins.
(iii) To ensure that the two ends of a TEG sustain a temperature difference all through flight, in
otherwords, heat should be convected away into the ambient, through proper selection of engine
wall thickness, heat dissipation fins and attainment of rocket flight speeds. The current TEG
devices can handle a maximum temperature difference of 2000C, with the hot side temperature
not to exceed 2250C (see appendix A1).
(iv) Quality data needs to be collected during the motor burn phase, and this requires that we select a
long-burn motor, which has sufficient thrust to get the rocket off the launch pad. We selected
Cesaroni L610 (appendix A3), which has an 8 sec burn time and adequate thrust to propel the
rocket off the launch pad.
Figure 1: Sequence of pictures showing the stages in the development of the thermoelectric engine;
challenges were (i) eliminating back pressure-induced thrust loss, (ii) finding the optimum jet interference
for heat transport through engine, (iii) optimal engine wall thickness and selection of suitable fins for heat
mixed layer
nozzle Supersonic Thrust core
venting slot insulation TEG
Figure 2: Schematic of the USLI payload showing the optimal location of the TEG devices with respect to
the supersonic thrust core; the payload has minimal drag interference with respect to the exhaust jet
IV. Rocket Vehicle Summary
The design of the rocket vehicle was influenced by three primary factors:
(i) The rocket should be powered by a motor with relatively long burn time in order to collect
scientific data via the payload specifically designed for the mission.
(ii) The rocket should fly safely according to high power rocketry design and safety principles
(thrust-to-weight ratio, stability margin, etc.).
(iii) The rocket should fly to an altitude of approximately one mile.
This threefold challenge required that the rocket design be very closely coupled to the motor selection.
The Cesaroni L610 Pro98 reloadable solid-fuel motor was selected for its thrust curve and relatively long
burn time. The team designed the rocket around this motor selection. The final design was a 6.16" diameter
rocket that stood 10' 4" high from fin to nose. It contains two dedicated electronics compartments for
avionics and experimental electronics. The body is composed of two 6" by 48" DynaWind tubes and the
ogive nose cone is 24” in length with a 4:1 length to diameter ratio. Two PerfectFlite miniAlt/WD logging
altimeters ensure that the chutes deploy at the correct altitude, enabling a safe recovery of the launch
vehicle. The drogue was set to open at apogee, and the main at 700 feet.
Figure 3: (a) The competition rocket and (b) the RockSim schematic showing the placement of the various
flight accessories
Table 1: USLI Launch Vehicle & Motor Details
Vehicle height : 10’4” Main Chute: 12’ diameter
Vehicle diameter: 6.16” Drogue: 3’ diameter
Launch Weight: 44.02lb (19.98Kg) Nose Cone length: 24”
Payload Weight: 4lb (1.82Kg) Nosecone Type: Ogive
Rocket Motor: L610 Nosecone Shape: 4:1 length to diameter
Body Tube Material: DynaWind Altimeter: MAWD PerfectFlite (2 units)
T/O Thrust –weight: 4.76 Drogue deployment: at apogee
Stability Margin: 2.38 Main Deployment: set at 700ft
Motor Specifications Flight Results
Length: 39.4cm, Diameter: 98mm Flight 1 (Feb 13): 5467 ft-----40.01lb @ T/O
Total Weight: 4.975Kg, Propellant: 2.415Kg Flight 2 (Mar 20): 5303 ft---- 42.87lb @T/O
Average Thrust: 610N, Burn Time: 8.1 secs Flight 3 (Apr17th): -----44.02 @T/O
Table 2: USLI Launch Vehicle Cost
Body Tubes, Rail guides, Recovery Sys $522.85 Thermoelectric devices $300.00
Nosecone $103.20 Resistor Bank $4.68
Altimeters – PerfectFlite $205.40 9V batteries $9.55
Carbon fiber material – Fins $100.94 Aluminum tube for payload $70.68
L610 motor, hardware, retainer $868.00 Heat Sink Fins $90.00
DAQ thermocouple boards $293.11 Rocket Graphics $44.50
DAQ flight computer $632.63 Misc items (epoxy, etc) $75.00
Omega K type Thermocouples $42.50
Connectors for payload electronics $124.21 Launch Rocket Total Cost $3487.25
V. Rocket Flight Simulation Summary
A software tool capable of predicting rocket flight trajectory was developed, by the Vanderbilt
USLI team, in Matlab. The VU USLI agenda was to rigorously simulate rocket flight from first principles.
In addition, the team felt it necessary to establish rocket flight profiles independent of and complement to
The effects of variable air density (as per flight day ground temperature and altitude), gravity (for
a more general purpose model), and motor mass were incorporated into the trajectory model. The time
dependent thrust and mass curves for the Cesaroni L610 rocket motor were obtained from thrustcurve.org.
Velocity and altitude data was calculated using numerical integration methods with a time step size of 0.01
seconds. The governing differential equations used in this simulation are below.
Th − W − D
dV = dt , dS = Vdt
⎛ r ⎞
m = m(t ), Th = Th(t ), W = mg , g = go ⎜ ⎟ (2)
− goh
1 rS
D = C D AρV 2 , ρ = ρoe RT
, h= , T = To − 0.0065h
2 r+S
Thrust loss due to the payload was not considered in this analysis, since benchmark static tests
indicated that no significant thrust loss would be incurred due the attachment of an appropriately designed
payload. The rocket CD was set at 0.61 (based on limited wind tunnel measurements, see appendix A4 )
and the appropriate rocket diameter 15.65 cm (6.16”) was used for the cross sectional area estimation. As
per our simulation, the rocket was supposed to reach an altitude of 1587m (5196 ft) on flight day for the
launch weight.
1600 March 20th Simulation
1200 April 17th Simulation
Height (m)
0 2 4 6 8 10 12 14 16 18 20 22
Time (s)
Figure 4: Predicted flight trajectories for the March 20th and the April 17th USLI launch.
VI. Payload Description
A. Data Acquisition System
The success of this project relied on the ability to collect accurate data during rocket flight, more
specifically, during the motor burn phase of the rocket flight. The thermoelectric engine uses
thermoelectric generators to produce power from a temperature difference. The rocket exhaust serves as
the heat source and the ambient air serves as the heat sink. These two temperatures dictate the performance
of the thermoelectric generators and thus the performance of the thermoelectric engine. The DAQ system
needs to be able to collect temperature, power, and flight characteristic data. We set up two independent
and identical data streams for redundancy.
The temperature data is needed to characterize the performance of the TEGs during flight as
compared to ground conditions. The power data will show exactly the power that is produced, and when it
is produced with respect to flight speeds and altitude. In order to collect all of the temperature and power
data and have them matched with the flight conditions it was necessary to use the RDAS flight computer
(RDAS Tiny v.4). The RDAS has a ‘G’ trigger which allows for data collection to begin recording 2
seconds prior to experiencing 2.5 G’s for 0.25 seconds. The RDAS was configured to log data from two
thermocouple boards (Aerocon 1.3H) and the voltage from the voltage divider across the TEGs. The
thermocouple boards used type K thermocouples which were placed in four different locations on the
payload (See figure 5 for these placements). Two were placed on the hot side (one on each end of the
TEG) to read the hot-side temperature range that the TEG experienced and the other two were placed on
the heat sink (cool side of TEG) to read the cold-side temperature range that the TEG experienced. Each
data stream had one hot and one cold side thermocouple wired to it.
Each data stream collected power data from a set of three TEGs wired in series. The TEGs were
wired to a resistor bank which matched their combined internal resistance (~3.3 ohms) for optimum power
dissipation. The flight computer can only read 0-5Volts and the voltage produced by the TEGs in series
would possibly exceed this. A voltage divider was designed to lower the voltage from the TEGs to
readable levels (see figure 6). After careful measurement of the exact resistance of each resistor and bench
testing the voltage divider, we reduced the TEG voltage by 4.5x on one stream and 4.6x on the other.
Figure 6 shows one of the two data stream setups and figure 7 shows the actual DAQ boards.
Figure 5: Thermocouple placement on the payload: ‘Cold’ are on fins (not shown), and ‘hot’ are
on the hexagonal tube. *Note: this is the payload unwrapped.
Figure 6: Schematic of the data stream set up for one set of three TEGs; there were two data
streams for redundancy.
Thermocouple boards load Resistor flight Computer voltage divider
Figure 7: Pictures of the two DAQ boards (mounted back-to-back), showing the various components
B. Payload Design
After substantial testing with payloads of varying size and shape to examine the payload effects on
motor thrust, the hot side temperature of the payload, and the power produced by the TEGs, an optimal
final payload design was chosen. Tests have shown that a greater temperature can be attained with a longer
body without measurably affecting the thrust. The final payload’s overall length was 8.5 inches (7.75 from
nozzle exit). This length would allow for a suitable temperature for the thermoelectric generators, closer to
2000C, where they deliver the most regenerative power. The final design’s inside circular diameter was 3 ¾
inches. In order to maximize the amount of power generated by the TEGs, a hexagonal shape was used for
the body of the payload so that six thermoelectric generators could be mounted. Figure8a shows the
payload isometric view with TEGs (colored red) mounted.
Following the March 20th flight a few improvements were incorporated. It was decided to mount all
TEGs circumferentially in a single axial location, and move them aft with respect to the nozzle by 0.5” so
that they could be exposed to a higher temperature. Further, it was decided to go for shorter but denser
higher-efficiency fins, and the vent slots were deemed too long and made shorter by about 0.5”. The
weight of the payload did not alter through these changes and remained at 4lbs.
C. Payload Assembly
Starting with a 3.75 ID, 4.25 OD standard aluminum stock, a hexagonal external shape and the
venting slots were milled with slot-slot circumferential spacing done using a rotary chuck. No finishing
work was done on the interior of the tube. The exact length of the machined tube was then faced off on the
lathe, along with machining the excess OD at the vent holes. The nozzle end of the tube was TIG-welded
to the retainer ring.
Each TEG was carefully epoxied at the right location on the hexagonal outer surface using
manufacturer supplied thermally-conductive high-temperature epoxy. Following this, a fin was epoxied on
the outer cold surface of each TEG using the same method. Once the entire assembly was completed, the
fins were further constrained to the payload tube by circumferentially running two 18G galvanized steel
wires in pre-cut tracks on the fins. The non-conductive surfaces of the TEG engine were thermally
insulated using high-temperature ceramic tape. Following this, three TEGs that belonged to a single data
stream, were connected in series. The appropriate thermocouples were embedded in their pre-determined
locations using thermal epoxy. Figure 8b shows the payload flown for the USLI launch on April 17th.
TEG insulation
Figure 8: (a) Isometric view (with dimensions in inches), and (b) the actual machined, assembled payload
flown for the USLI launch on April 17th.
VII. ULSI Flight Summary
The USLI Launch Director requested that all competition rockets be aligned at a 5degree angle
away from the crowds. The rocket took off and cleared the launched pad well. Once it reached apogee the
drogue was deployed. On the descent it was noticed that the main had deployed early. The descending
rocket did not seem to catch much of the side breeze and it landed safely within the field, with neither the
rocket nor the payload suffering any damage (figure 9).
Our team’s post flight interpretation of the MAWD altimeter beeps indicated that the rocket had
reached an altitude of 5466ft as relayed to NASA. This was a surprise that despite an angled launch, and a
predicted maximum of 5196ft, due to various mass additions to the rocket, the rocket had overshot the
apogee! Post-flight analysis shows that both the MAWD altimeters had recorded pressure spikes at apogee
during drogue deployment (figure 10a). The two MAWD altimeters project an average apogee at 5032ft
(official altitude), and the two RDAS altimeters project an average apogee at 5096ft. If an average of all
the four altimeter data were taken, then the rocket reached an apogee of 5064ft (1547m). The vehicle
attained a maximum speed of 152.5 m/s (500ft/s) and experienced at least 4g at take-off. (figure 10 b,c,d).
The safe descent rate was at 15.9ft/s. The slightly angled launch could have resulted in the slighter lower
altitude than estimated.
Despite a successful flight and main deployment at 700ft on our March 20th flight, with the exact
same rocket and the same number of harnessing shear pins, the USLI flight threw up the surprise of the
early deployment of the main parachute. The nose cone had a good interference fit (in fact it was a snug
fit), and it was additionally harnessed with two shear pins; the team felt that four shear pins would be too
much and played it safe, but the flight results proved otherwise.
Our post-flight analysis has established that changing the parachute to a 12ft one from the 10ft
one, for the USLI flight may have added sufficient extra mass to dislodge the nose cone during the drogue
deployment; additionally strong winds at altitude may have enabled the process, as few of the competing
teams had similar deployment of the main. All in all, the scientific payload and the rocket were recovered
without any damage. In hindsight, four shear pins would have given us a perfect flight!
Figure 9: The Vanderbilt Team with the post-USLI recovered rocket and payload; neither suffered
any damage and were certified to fly immediately.
Height (m)
Projected Flight Altitude
Actual Flight Altitude
0 2 4 6 8 10 12 14 16 18 20 22
Time (s)
160 4
Specific Acceleration
Velocity (m/s)
80 1
0 -2
0 5 10 15 20 25 0 5 10 15 20 25
Time (s) Time (s)
Figure 10: (a) MAWD PerfectFlite altimeter data from USLI flight; dashed line shows the
possible trajectory if main had deployed at 700ft (b) average apogee reached was 1537 m (5032 ft) in
comparison to the estimated (5196 ft), (c) the deduced flight velocity showing that the vehicle reached a
maximum speed of 152.5 m/s (500 ft/s), and that the take-off acceleration was minimally 4g.
VIII. Payload Scientific Data Analysis & Results
A. Raw Data and first steps in Data Processing
The voltage generated by a thermoelectric generator is directly proportional to the temperature
difference between the hot source and cold sink. The constant of proportionality, S, is termed the Seebeck
V = SΔT (3)
In order to determine the temperature difference established during flight, four thermocouples
were installed on the payload; two embedded in the wall of the exhaust tube (hot source) and two mounted
on the fins (cold sink). Temperature and power data was recorded at 50 Hz throughout the flight.
Temperature (C)
Cold Side Temp.
20 Hot Side Temp.
Temperature Difference
0 5 10 15 20 25
Time (s)
Figure 11: Hot side (jet) and cold side (fin) temperature data across TEG devices during flight
Ground-based testing established a non-linear increase in hot source temperature along the axial
direction of the exhaust tube. As a result, a weighted average of the thermocouple temperatures was used
to establish the average hot- and cold-side temperatures. In figure 11, the average hot- and cold-side
temperatures and the average temperature difference are plotted to flight apogee. As expected, the
temperature gradient generated by the hot rocket exhaust and ambient air flow generated electrical power.
Voltage (V) across the load resistors (R) for the two independent TEG streams was measured and the
generated power (P) was calculated (P=V2/R). This is plotted in figure 12.
18 TEG Stream 1
TEG Stream 2
Power (W)
0 5 10 15 20 25
Time (s)
Figure 12: Power generated by the two TEG streams; clearly one of the streams generated more
power than the other.
As shown in figure 12, one power stream generated substantially more power than the other. We
have hypothesized that this discrepancy was caused by a slightly skewed jet angle with respect to the
payload wall. Ground-based studies have shown that even miniscule asymmetries in jet impingement can
result in large temperature variations. To ascertain this hypothesis, we meticulously examined the flight
videos from our three launches (Feb 13th, March 20th, and USLI April 17th). From figure 13, it is apparent
that the jet axis is slightly skewed with respect to the rocket axis. We established that the motor tube is
slightly misaligned with respect to the rocket. While this is probably true for most rocket assemblies, the
high sensitivity of our payload has clearly pointed out this effect. We further examined this asymmetry by
measuring the offset of the nozzle face with a reference surface and the results are plotted in appendix A6.
a b1 b2 c1 c2
Figure 13: The take-off signature of the exhaust jet with respect to the rocket axis for (a) Feb 13th, (b)
March 20th, and (c) April 17th USLI launches. Clearly the jet axis is slightly off compared to the rocket
The combined power generated would be a better representation of the performance of the payload
and therefore the total power generated is plotted in figure 14 till time to flight apogee. Maximum power
generation is clearly centered about the rocket motor burnout time. Nearly 35 watts of power was
generated at peak power production.
Total Power
Total Power Generated (W)
0 2 4 6 8 10 12 14 16 18 20
Time (s)
Figure 14: Total power generated by the two TEG streams till apogee. The rocket burnout is at 8 seconds.
B. Comparisons with Ideal TEG Engine Operation
This section examines the comparisons between the experimentally generated power to that
theoretically possible under ideal conditions, and what design factors could assist in approaching the ideal
conditions. The analysis has been limited to the motor-powered phase of the rocket flight (t < 8 seconds),
where the quality of the temperature data is high. While electrical power is being continuously produced
during the entire rocket flight as the TEG devices cool, the heat transfer problem following motor burnout
is complicated by the fact that the TEG devices are cooled not only on the cold side, but also on the hot side
due to the air entrained through the venting slots at the nozzle exit.
Combining equations (3) and the expression for power, we get
S 2 ΔT 2
P= (4)
The relationship in (4) is readily observed when total power generated is plotted as a function temperature
difference (figure 15).
Total Power Generated (W)
Total Power Generation
Projected Power Generation
0 20 40 60 80 100 120 140 160 180 200
Temperature Difference (C)
Figure 15: Total power generated as a function of temperature difference across the TEGs
The thermoelectric generator manufacturer specification sheet (appendix A1) indicates that each
thermoelectric generator will produce 14.7 W given a temperature difference of 200°C. Further, the hot
side temperature should not exceed 225oC. In the current flight, the hot side temperature reached 170oC at
motor burnout, and the maximum temperature difference achiev

Use: 0.1197