|UFDC Home||myUFDC Home | Help ||
|Table of Contents|
|List of Tables|
|List of Figures|
|List of symbols and nomenclatu...|
|Literature review and theoretical...|
|Experimental approach and...|
|Properties of the acoustic...|
|Empirical model development and...|
|The effects of acoustics on the...|
|Conclusions and recommendation...|
|Appendix A: Computer programs|
|Appendix B: Checklists and...|
|Appendix C: Statistical design...|
|Appendix D: Experiment run...|
|Appendix E: Transfer functions|
|Appendix F: Temperature profil...|
CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
STANDARD VIEW MARC VIEW
|Table of Contents|
Table of Contents
List of Tables
List of Figures
List of symbols and nomenclature
Literature review and theoretical approach
Experimental approach and facility
Properties of the acoustic field
Empirical model development and analysis
The effects of acoustics on the reformation process
Conclusions and recommendations
Appendix A: Computer programs
Appendix B: Checklists and procedures
Appendix C: Statistical design of experiments
Appendix D: Experiment run orders
Appendix E: Transfer functions
Appendix F: Temperature profiles
ENHANCING THE STEAM-REFORMING PROCESS WITH ACOUSTICS: AN
INVESTIGATION FOR FUEL CELL VEHICLE APPLICATIONS
PAUL ANDERS ERICKSON
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
UNIVERSITY OF FLORIDA
3 1262 08554 4335
Paul Anders Erickson
This dissertation is dedicated to my wife, Dena.
Anyone with a semblance of true wisdom knows that one in and of oneself is a
very small thing. Such is the case with this study, as many people have contributed
directly or indirectly to this dissertation. I acknowledge the following individuals and
institutions who have extended themselves to give support to the present study.
I acknowledge the support, encouragement, and patience put forth by my advising
professor, Dr. Vernon P. Roan. Dr. Roan is truly a scholar and I am honored to have had
him as my mentor for the duration of my time at the University of Florida. Similarly, I
would like to thank my doctoral committee (Dr. Jean M. Andino, Dr. D.Yogi Goswami,
Dr. Ulrich H. Kurzweg and Dr. William E. Lear, Jr) for their support in this research
project and in my personal education.
I acknowledge the State of Florida, through the University of Florida and the
Mechanical Engineering Department for funding my education and allowing me to serve
as an instructor and teaching assistant. The research has been funded in part by Ford
Motor Company with their generous grant to the University of Florida Fuel Cell Research
and Training Laboratory for studies dealing with reformation of hydrocarbon fuels.
I acknowledge the support and guidance I have received from my parents Lloyd
and Virginia Erickson and my siblings. Words cannot express my thanks to my wife
Dena, to whom I have dedicated this work, and my children Timothy, Eliani, and Kari.
They provide me with inspiration to fulfill my potential. My family has sacrificed much
that I might pursue my education. Truly my family is the center of my joy and happiness.
I also acknowledge support from my in-laws Fred and Jacque Wright as well as Les
Meinzer who provided computer hardware for the publication of the project. I
acknowledge my colleagues at the Fuel Cell Research and Training Laboratory, with
special mention of Timothy Simmons, Daniel Betts, and Michael Heckwolf who gave
much needed assistance with the project and encouragement at critical times.
No acknowledgment would be complete without thanking God for enlightenment,
wisdom, and vision to see me through this project. My sincere hope is that He will find
that I have used my knowledge, time and resources to help His children here on earth, and
that I will be found worthy of His approval through the Messiah. Despite any small
amount of earthly knowledge I have obtained, I consider myself a "fool before God" and I
understand that eternal knowledge and wisdom is His to give to those who humbly and
diligently seek His will.
TABLE OF CONTENTS
LIST OF TABLES ........................
LIST OF FIGURES .......... .............
LIST OF SYMBOLS AND NOMENCLATURE
.. ... ...... ................ xv
1 INTRODUCTION ........ ........................
Background .... ................ ... ....... ......
Fuel Cells .....................................
Vehicle Applications ....... .......................
Fueling Options ..... ... ................... ...
Reform ing .. ............ .................. ... ....
Problem Definition ...... ...............................
Research Objectives ............... ....................
2 LITERATURE REVIEW AND THEORETICAL APPROACH .
Steam Reforming .. ..................................
Limiting Steps in the Reformation Process ..................
Acoustics in Past Processes ..............................
Steam-Reforming in Combination with Acoustics ............
Increased Particle Path Length due to Acoustics ..........
Increased Heat Transfer Rate due to Acoustics ...........
Increased Overall Reaction Rate due to Acoustics .........
Extension of Catalyst Life by Using Acoustics ...........
Contribution .............................. ..........
3 EXPERIMENTAL APPROACH AND FACILITY
il Facility ...................... ........
g Subassem bly .........................
zer Subassembly ........................
Subassem bly ..........................
talyst bed .................. ...........
oustic components ......................
sing Unit Subassembly .......................
entation and Control Subassembly ..........
mperature control .......................
oustic control ..........................
al of Data Points ........................
. . . . . . . . . . .
Modeling the Acoustically Enhanced Reformation Process
Factorial Experiment Design ....................
Further Data and Verification of the Model .........
4 PROPERTIES OF THE ACOUSTIC FIELD ...........
The Use of Resonance within the Catalyst Bed ...........
Transfer Functions .................................
M odal A analysis ...................................
5 EMPIRICAL MODEL DEVELOPMENT AND ANALYSIS
Experimental Results and Model
Uncertainty Analysis .........
Midpoint Curvature .........
Flow Rate .............
Sound Pressure ..........
Catalyst Bed Length ......
The Effect of Frequency .......
. . . . .
6 THE EFFECTS OF ACOUSTICS ON THE REFORMATION PROCESS
The Effect on Temperature Profile ...............................
The Effect on Conversion ......................................
The Effect on Controllability ........................ ... ........
The Effect on Heat Band Power Draw ................ .................. 93
The Effect on Power Output ......................................... 95
The Effect on Catalyst Degradation ................................... 98
Potential Benefits for a Fuel Cell Vehicle ................................ 100
7 CONCLUSIONS AND RECOMMENDATIONS .......................... 103
Conclusions ...................................................... 103
Recommendations ................................................. 104
A COMPUTER PROGRAMS .......................................... 107
B CHECKLISTS AND PROCEDURES ................................. 123
C STATISTICAL DESIGN OF EXPERIMENTS ........................... 127
D EXPERIMENT RUN ORDERS ....................................... 135
E TRANSFER FUNCTIONS ........................................ 137
F TEMPERATURE PROFILES ....................................... 147
REFERENCES ...................................................... 153
BIOGRAPHICAL SKETCH ............................................ 158
LIST OF TABLES
1.1 Types of fuel cells and their operating parameters ........................... 3
1.2 Energy densities of fuels (LHV). ........ ............. .................. 7
3.1 Control setpoint matrix. .......................................... 48
3.2 Average and standard deviation for all catalyst bed temperatures utilizing a
250C setpoint.................................................... 49
3.3 Gas chromatograph calibration levels. ..................... ............... 54
3.4 Volume percent species of the dry product outlet gas. ...................... 55
3.5 Factorial input matrix and run order. ................................. 59
4.1 Waveforms and boundary conditions for a 33 C reactor containing only air .... 72
5.1 The input matrix and methanol conversion percentage data obtained from the
factorial experiments .............. ......... ................. ..... .. 74
5.2 Total average conversion, pooled standard deviation, standard error, and total
degrees of freedom from the factorial experiments ......................... 75
5.3 The effects and interactions found from the factorial experimentation. ......... 75
5.4 Signal-to-noise t-ratios and statistical significance. ..................... .. 77
5.5 Static and kinematic Reynolds numbers and methanol conversion levels for
three frequencies. ................................................. 82
6.1 Extent of conversion for catalyst bed length with and without acoustics. ....... 89
6.2 Average and standard deviation of the centerline temperatures with a setpoint
temperature of 250C at various locations both with and without a 165dB
acoustic field. ..................................................93
6.3 Overall average catalyst bed power draw and pooled standard deviation for
conditions with and without a 165 dB acoustic field. ....................... 94
6.4 A summary of significant performance parameters with and without acoustic
enhancement for two space velocities. ................................. 97
6.5 Auxiliary requirements for generating the acoustic fields in the current facility. 102
LIST OF FIGURES
1.2 Simplified fuel cell schematic. ...................................... 2
1.2 Hondas' hydrogen-fueled experimental fuel cell vehicle ................... 5
1.3 Methanol-fueled PAFC transit bus at the University of Florida .............. 9
2.1 Simplified representation of a packed bed methanol-steam reformer. ......... 14
2.2 Simplified representation of the reactants moving to the open catalyst site and
products moving out to the bulk stream....................... ........... 16
2.3 Flow fields at different times for a plane wave superimposed onto a steady
flow field. ........................................................ 19
2.4 A 15 kW diffusion propane flame. .................................... 21
2.5 Total slip displacement and slip displacement due to acoustics for increasing
velocity ratio. ................................................ 24
2.6 Nusselt number ratio versus velocity ratio for a single particle in an empty duct. 26
3.1 Reforming facility at the University of Florida Fuel Cell Research and
Training Laboratory. ....... ....................................... 33
3.2 Simplified schematic of the methanol steam reforming processor ............ 34
3.3 Artistic representation of the pumping subassembly ....................... 36
3.4 Artistic representation of the vaporizer and liquid trap .............. ...... 37
3.5 Artistic representation of the superheater .............................. 38
3.6 Artistic representation of the catalyst bed housing. ............ ...... 4. 40
3.7 Artistic representation of the acoustic driver, driver housing. mounting plate
and fittings ....................................................... 42
3.8 Artistic representation of the condenser subassembly ....................... 44
3.9 Input and output signals for temperature and acoustical control .............. 45
3.10 Control circuit for heating elements................................... 46
3.11 Power distribution box.............................. ........ ......... 47
3.12 Heating element control logic ...................................... 47
3.13 Time plot of catalyst bed temperatures at zone exit utilizing two control
schemes with a 250C setpoint. ..................................... .49
3.14 Typical time trace of pressure in the catalyst bed while generating hydrogen
with an acoustic input of 650 Hz ................................... 51
3.15 Flow paths in the gas chromatograph. ................... ............. 53
3.16 Silica gel-packed column temperature and approximate specie elution times
for the species expected with the gas chromatograph. ...................... 53
4.1 Standing wave pattern of displacement and velocity in a waveguide and an
analogous pattern of displacement and velocity in a driven string. ............ 63
4.2 Transmission and reflection of an acoustic plane wave at a pipe junction. ...... 64
4.3 Transfer function magnitude for the catalyst bed housing containing only
33C air. ........................................................ .66
4.4 Transfer function magnitude for a catalyst bed housing containing catalyst
and air at 33C ................................................... 67
4.5 Transfer function magnitude for a reactor containing reduced catalyst and CO,
at operating temperature (250 C) .......................... ........... 67
4.6 Reactor transfer function magnitude while generating hydrogen at 2500C....... 68
4.7 Frequency response of the acoustic driver as measured in a plane wave tube
by the manufacturer. .............................................. 69
4.8 Normalized velocity magnitude for six waves found in a 33 C reactor
without catalyst. ................................................. 70
5.1 Predicted and actual conversions for various flow rates with and without
5.2 Predicted and actual conversion for maximum pressure amplitude with a
0.568 m catalyst bed length and 9.5 g/min premix flow rate. ....... ........ 80
5.3 Predicted and actual conversions for bed length at a flow rate of 9.5 g/min
premix and a 165 dB acoustic field. .................................. 81
6.1 Temperature profile for the catalyst bed at premix flow rate 4.9 g/min without
an acoustic field .... ............... ............... ..... ..... .... 86
6.2 Temperature profile for the catalyst bed at premix flow rate of 4.9 g/min with
an acoustic field at 165 dB............................................ 86
6.3 Temperature profile for the catalyst bed at a premix flow rate of 9.5 g/min
without an acoustic field .......................................... 87
6.4 Temperature profile for the catalyst bed at premix flow rate of 9.5 g/min with
an acoustic field at 165 dB ........................................... 87
6.5 Conversion versus bed length with and without acoustic enhancement. ........ 90
6.6 Difference in conversion as a function of bed length for 4.8 and 9.5 g/min
premix flow rates .................................................. 91
6.7 Conversion versus space velocity with and without a 165 dB acoustic field. .... 91
6.8 Average power draw for zone band heaters at two space velocities with and
without a 165 dB acoustic field. ..................................... .94
6.9 Average maximum pressure as a function of driver power at 650 Hz with a
0.568 m catalyst bed length while generating hydrogen ..................... 95
6.10 Average maximum pressure as a function of driver power at 650 Hz with a
0.095 m catalyst bed length while generating hydrogen ..................... 97
6.11 Catalyst degradation with and with out acoustics over a 50 hour time period
with an average premix flow rate of 9.3 g/min. .......................... ..98
LIST OF SYMBOLS AND NOMENCLATURE
Symbol Definition Symbol Definition Symbol Definition
6 Boundary layer
Rek Kinetic reynolds
Nu, Nusselt number
Nu, Nusselt number
S Cross sectional
X Input variable
n, Number of
f Frequency (Hz)
ro Pipe radius or
L Length or
Sh Sherwood number
um Acoustic velocity
Pm Acoustic pressure
rh Mass flow rate
Y Output variable
o, Pooled standard
Root mean square
SE Standard Error
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
ENHANCING THE STEAM-REFORMING PROCESS WITH ACOUSTICS: AN
INVESTIGATION FOR FUEL CELL VEHICLE APPLICATIONS
Paul Anders Erickson
Chairperson: Dr. Vernon P. Roan
Major Department: Mechanical Engineering
An investigation was made into the possibility of enhancing steam reformation
processes through superposition of an acoustic field on the catalyst bed of a methanol
steam reformer. The enhancement scheme presented is of particular interest for
application in fuel cell vehicles or similar applications where size and mass of a reactor is
to be minimized. As part of this study, background is given outlining the difficulties and
liabilities of steam-reformation for transportation applications. Proven acoustic
enhancement of various processes is reviewed and a theory-based model of acoustic
enhancement of the steam-reforming process is developed. The experimental facility
includes a steam-reforming reactor that has been modified to accept an acoustic field.
Relc\eant parameters of the acoustic field are quantified and discussed. The effect of the
acoustic field is investigated experimentally with relation to the reactor output parameters
of temperature profile, controllability, fuel conversion, species concentration, power
demand, and catalyst degradation. Additionally, an empirical model is developed that
predicts the fuel conversion in the reactor for a given set of fuel flow rate, catalyst bed
length, and acoustic pressure parameters.
Although the test facility used was not optimized for utilizing acoustic N\ aves,
significant acoustic enhancement of the steam-reformation process is demonstrated by the
study and such enhancement is statistically proven to have a positive effect on the
reformation process. Potential benefits resulting from acoustic enhancement of steam
reforming as shown by this study are an increased reactor capacity for a given size and
mass, smoothing of the temperature profile, better temperature control of the catalyst bed,
and increased catalyst life. It is expected that for different fuels and/or reforming
methods, similar results would be obtained for comparable process constraints.
Conclusions and recommendations for further study and application are drawn from the
This chapter gives the background for the present study and is an introduction to
fuel cells, including a brief historical sketch of the technology. The different types of fuel
cells are outlined and the main difficulties in applying fuel cells for transportation are
mentioned. Fueling options for fuel cell vehicles are discussed and the difficulties in
reforming liquid fuels are presented. This background allows a formulation of the
problem definition and a statement of the study's objective.
Because of high energy conversion efficiencies and low emissions, fuel cell power
plants (or "engines") are currently receiving a high degree of interest as potential
replacements for the internal combustion engines in transportation applications.
The fuel cell was first demonstrated in 1839 by Sir William Grove. However, the
majority of practical work on fuel cells did not happen until Francis Bacon demonstrated
a 5 kW working system in 1959. NASA became interested in the technology, and fuel
cells were further developed under General Electric and United Technologies Corporation
for the Gemini and Apollo space flights. The Arab oil embargo in 1973 stimulated the
United States Department of Energy (DOE) to look for ways to conserve
petroleum-derived fuels and utilize fuel cells in terrestrial applications. A major
breakthrough in fuel cell technology came in the 1980s when Los Alamos National
Laboratory and Texas A&M University found a way to decrease the platinum loading for
certain types of fuel cells (Norbeck, 1996). Since that time vehicle manufacturers have
been looking to the fuel cell power plant as a way to comply with ever-increasing
emissions standards, yet keep the high level of performance expected with the mature
internal combustion engine.
The fuel cell is an electrochemical cell that uses an oxidation reaction to generate
electricity directly from a fuel and an oxidizer. Figure 1.1 shows how an acid or polymer
electrolyte fuel cell operates. Other fuel cell types operate in a comparable manner yet
these other types of fuel cells can use conductive oxide, alkali, or molten carbonate as
electrolytes. For the acid or polymer type fuel cells as shown in Figure 1.1, hydrogen
fuel is brought to the anode of the fuel cell. At the anode, typically a platinum-type
catalyst splits the hydrogen into protons and electrons. The protons are free to migrate
Electrolyte H+ ,o
t+02+4e 2HO "
Figure 1.1: Simplified fuel cell schematic.
across the electrolyte and combine with oxygen at the cathode to form water. To
complete this reaction, the electrons that are left at the anode must be brought to the
reaction site at the cathode. A conductive path is made available through a load and the
electrons are passed through the circuit. The end products are water, heat, and direct
current electrical power. Because a single fuel cell only generates about 0.75 volts,
practical applications arrange multiple fuel cells in series to form a fuel cell stack that
yields usable voltages.
Several different types of fuel cells exist. The different types of fuel cells are
normally distinguished by their electrolyte. Table 1.1 (adapted from Norbeck, 1996) lists
several types of fuel cells along with their operating temperature, typical application, and
chemical tolerance to carbon monoxide and carbon dioxide.
Table 1.1: Types of fuel cells and their operating parameters.
Fuel cell Typical stack
operating Typical applications CO tolerance o
type temperature tolerant?
Solid oxide Steam cogenerating power
(SOFC) 1000 plants Good Yes
olten Steam cogenerating power
car e 600 plants Good Yes
Phosphoric Smaller stationary power
AF 150-205 Heavy transport (buses, Fair Yes
trucks, ships, trains)
Solid Stationary and mobile power
(PEM) 25-120 Heavy transport, light Poor Yes
Space Vehicles, submarines,
Alkali 65-220 specialty vehicles carrying Poor No
hydrogen and oxidi/er
Transportation applications place demands on the power source that currently rule
out certain types of fuel cells. Some of the more significant demands on the fuel cell
system for these applications are the ability to use air as the oxidizer, and quick start-up
and shutdown periods. Because of the development of the Apollo and Space Shuttle
Programs, alkali-type fuel cells have been thoroughly studied, produced, and successfully
operated for over 40 years. These fuel cells have provided astronauts with electrical
power, heat and drinking water. Excellent as they may be for such space vehicles
carrying hydrogen and oxidizer, these types of fuel cells are of little use in terrestrial
applications because the Alkali-type fuel cells are not tolerant of the low levels of carbon
dioxide in the air or in reformed liquid fuels. Solid Oxide and Molten Carbonate type
cells are excellent in terms of carbon dioxide and carbon monoxide tolerance; however,
their high temperatures and associated long start-up times currently prohibit their use in
automotive applications. The lower-temperature fuel cells that are carbon dioxide
tolerant such as the Polymer Electrolyte Membrane (PEM) and Phosphoric Acid Fuel
Cells (PAFC) are more suitable for transportation applications. PEM (then referred to as
solid polymer electrolyte or SPE) fuel cells were first developed in the 1960s for NASA
and were used in the Gemini Program (Prater, 1990). These fuel cells were not further
developed for NASA as alkali-type cells were used in later space missions. PEM fuel
cells can have high power densities because of the thin polymer electrolyte and thus have
an advantage over the bulkier PAFCs for terrestrial applications. On the other hand,
PAFCs which operate at a higher temperature have a better tolerance for carbon
monoxide in the fuel stream. Both of these cell types (indeed all fuel cells, except for
direct methanol) best utilize pure hydrogen as the fuel. As it has little infrastructure, this
use of hydrogen is a significant drawback to widespread implementation of hydrogen-
fueled fuel cell vehicles. As shown in Figure 1.2 (Photograph by Tom LaRocque and
.. .. .. .
f .. l
Figure 1.2: Hondas' hydrogen-fueled experimental fuel cell vehicle.
NREL/DOE), this limited availability has not prevented development and demonstration
of hydrogen prototype fuel cell vehicles by most auto manufacturers. Auto manufacturers
see fuel cell powerplants as perhaps the only feasible technology that can match the
requirements of both consumers and increasingly stringent emission regulations while
reducing dependence on petroleum and reducing the generation of CO,.
In a fuel cell chicle that uses hydrogen as a primary fuel, there are no chicle
emissions as the only end products are water, heat, and electrical power. As hydrogen is
also the most abundant element in the known universe, it may seem that hydrogen would
be the ideal primary fuel for a fuel cell vehicle. Unfortunately, virtually all of the
hydrogen on earth is bound up in compounds such as water that are at a low energy state.
Because there are no direct hydrogen sources available, with the exception of small
quantities of hydrogen found with some natural gas deposits, hydrogen fuel must be
derived from other sources. Although there are some projects studying and
demonstrating renewable hydrogen production (Erickson and Goswami, 2001), the
majority of hydrogen production currently utilizes fossil fuels, primarily natural gas, for
the feedstock. The production of hydrogen from fossil fuels creates emissions where the
hydrogen or the energy to produce the hydrogen is generated. For a hydrogen fuel cell
vehicle there are no vehicle emissions but the production of the hydrogen fuel creates
emissions upstream for the vehicle. While there are places such as the Los Angeles
Basin, where transferring pollution outside of the area would be beneficial, the overall
efficiency is changed only slightly by production of hydrogen from a hydrocarbon fuel
on-board the vehicle (Louis, 2001). Adamson and Pearson (2000) have shown that the
"well-to-wheel" efficiency and emissions for fossil-derived hydrogen-fueled fuel cell
vehicles over the entire use-phase life-cycle are only slightly better than that of fossil-
derived methanol-fueled fuel cell vehicles over the same life cycle.
With these similarities in the overall energy conversion understood, one must
consider which fuel would be a wise choice for use in fuel cell vehicles. Table 1.2
(Adapted from Ogden and Nitsch, 1993; and Turns, 1996) shows the energy densities,
based on the lower heating value available from various fuels. These energy densities
demonstrate a practical area where hydrogen fuel is clearly not the best choice. While
Table 1.2: Energy densities of fuels (LHV).
Fuel Volumetric energy Mass energy density
density (MJ/liter) (MJ/kg)
Diesel 33.3 44.5
Gasoline 32.3 43.0
Liquified propane (690 kPa) 23.6 46.3
Methane gas 5.51 50.0
Liquified methane 2. .
(-161 C)21.5 39.1
Methanol 15.9 19.9
Ethanol 21.6 27.2
Hydrogen gas (16.5 MPa) 1.76 120.2
Liquified hydrogen (-253 C) 8.4 120.2
Metal hydride 4.9 1.80
hydrogen gas has the highest specific energy (mass energy density) of all practical fuels,
it has the lowest volumetric energy density. The volumetric energy density of hydrogen is
still paltry compared to liquid hydrocarbon fuels even in cryogenic liquid form near
absolute zero temperature. This issue of poor volumetric energy density suggests
prohibitively large fuel tanks for hydrogen fuel cell vehicles to achieve current vehicle
ranges (-560 km, -350 miles). In terms of volumetric energy density, a more feasible
alternative is to use a liquid hydrocarbon fuel such as liquified methane, liquified
propane, methanol, ethanol, gasoline, or diesel.
As previously mentioned, an even more significant problem hindering the
introduction of direct hydrogen fuel cell vehicles is the lack of hydrogen fueling
infrastructure. The current United States fuel infrastructure has evolved for use of liquid
fuels, notably gasoline and diesel. Because of the infrastructure and energy density
issues, near-term true consumer (as opposed to limited fleet operation) fuel cell vehicles
will likely utilize a liquid hydrocarbon fuel. Although smaller environmental benefits are
projected for using these liquid fuels, rather than using hydrogen from a renewable
resource, the use of fossil-derived liquid hydrocarbon fuels can serve as a precursor to
enable the move toward a renewable energy economy.
Liquid hydrocarbon fuels need to be decomposed or "reformed" into a hydrogen-
rich stream before they can be used by a Polymer Electrolyte Membrane (PEM)' or a
Phosphoric Acid Fuel Cell (PAFC). This can be accomplished by a steam reforming,
partial oxidation, or an autothermal reaction. For comparisons of the reforming methods,
see Ahmed and Krumpelt (2001), Brown (2001), and Docter and Lamm (1999). The on-
board reforming process further complicates the fuel cell vehicle and increases the cost,
weight, and under-hood volume requirements of the fuel cell engine. The reforming
process also penalizes the overall on-board energy conversion efficiency because of an
associated increase in entropy. Furthermore, the liquid-fueled vehicle may require
clean-up units to purify the reformate stream before hydrogen-rich mixture can be utilized
in the fuel cell. The reforming process also may further complicate the overall vehicle
system by requiring long start-up (Erickson et al., 2000) and shutdown (IHeck\x olf et al.,
2001) periods. Although liquid-fueled fuel cell vehicles are demonstrated, as shown in
Figure 1.3 (Photograph by Daniel Betts and the University of Florida Fuel Cell Research
There are some exceptions such as the direct methanol PEM fuel cell but it is
still in the laboratory research phase.
Figure 1.3: Methanol-fueled PAFC transit bus at the University of Florida.
and Training Laboratory), the efficient reforming of the liquid fuel utilizing practical on-
board reformers is considered one of the major technological hurdles delaying widespread
introduction of fuel cell vehicles (Erickson and Roan, 1999; Romney, 2001). Steam
reforming, where fuel and steam are externally heated and catalytically reacted is
currently the most widespread method of reforming a hydrocarbon fuel into a hN drogcn-
rich gas. Although steam reforming is widely used, typical steam reformers are not
designed for transportation application and thus they have large volume with
corresponding high mass, slow transient response (especially at start-up), and are often
characterized by ha\ ing hot and cold spots throughout (deWild and Verhaak, 2000).
As introduced above, the problem of reforming liquid hydrocarbon fuels to
hydrogen is currently one of the major technological hurdles delaying the introduction of
liquid-fueled fuel cell vehicles. While it has been demonstrated that the endothermic
steam-reforming process can be accomplished in a catalytic converter (Idem and Bakhshi,
1994: Ledjeff-Hey et al., 1998: Peppley et al., 1997), the process has liabilities of weight,
size, and complexity, in addition to long start-up and transient response times (Ohl et al.,
1996). Because of these liabilities any enhancement scheme for the steam-reforming
process which could lead to reducing size, weight, or response time of the reformer
should be investigated.
A steam-reformation process that utilizes acoustic fields in critical fluid paths is
proposed. This study develops a theoretical approach for the use of acoustic fields in
conjunction with steam-reforming of methanol and demonstrates that the resulting
acoustic enhancement of reforming is one approach that can reduce the above liabilities
of the steam-reforming process.
The objectives of the present study are to develop theory-based models for
acoustic enhancement of the reformation process, to experimentally demonstrate a steam-
reformation process under the influence of an acoustic field, and to quantify the effects of
acoustic fields on the steam-reforming process. This includes an analysis of acoustic
fields and investigation of the effect of acoustic fields with relation to the reactor output
parameters of temperature profile and controllability, power demand, fuel conversion,
and catalyst degradation. The objectives also include a demonstration of the overall
hydrogen power output relative to acoustic power input. In connection with the objective
to quantify the effect of the acoustic field on the reformation process, an empirical model
is developed that predicts the conversion for a given set of space velocity, and acoustic
pressure parameters. Fulfillment of the objectives provides understanding of how
acoustic fields can be used to enhance the reformation of hydrocarbon fuels, resulting in
smaller, more effective reformers, especially relevant for transportation applications.
LITERATURE REVIEW AND THEORETICAL APPROACH
This chapter provides a theoretical basis and background for the strategy of using
acoustic waves as a means of enhancing performance in conjunction with the
hydrocarbon fuel steam-reforming process. As part of the theoretical approach, the
literature dealing with steam-reforming and acoustic enhancement of related processes is
reviewed and the potential contribution of the present study is identified.
Steam reforming is the conversion of a hydrocarbon and steam into hydrogen-rich
gas. The steam reforming process was developed in the 1920s to produce a mixture of
fuel gas containing carbon monoxide and hydrogen known as "syngas" or synthetic gas
(Cox and Williamson, 1977). Steam reforming can occur with a variety of hydrocarbon
fuels including methanol and is used extensively for the commercial production of
hydrogen from natural gas. Equation 2.1 shows the ideal steam-reforming reaction for
CH3OH+ H20-+ 3H2 + CO2 (2.1)
production of hydrogen and carbon dioxide from methanol and steam. The change in
enthalpy of the reaction is +49.0 kJ/mol. This net gain in enthalpy indicates that
methanol-steam reformation is endothermic. Practical product species typically include,
in addition to hydrogen and carbon dioxide, carbon monoxide, unreacted methanol and
steam. Significant methane concentrations and solid carbon formation have also been
noted under certain conditions. The reforming process typically occurs at a temperature of
200-3000C and in the presence of a catalyst (usually copper and zinc on an alumina
substrate) that aids the reaction kinetics and shifts selectivity to carbon dioxide. Excess
water (1.3:1 -1.5:1 steam:carbon ratio) is used to inhibit formation of solid carbon (Cheng
et al., 1999; Amphlett et al., 1993). Models have been developed for methanol steam
reforming processors (Asprey et al., 1999) and reaction mechanisms have been proposed
(Amphlett et al., 1988). It has been proposed that a simplified model of the methanol
steam reforming process can be seen as two series reactions which are the endothermic
decomposition of methanol, followed by the exothermic water-gas shift reaction. These
steps are shown in Equations 2.2 and 2.3 respectively. The decomposition of methanol is
thought to be the rate-limiting step as its rate is approximately one order of magnitude
less than the water-gas shift reaction (Pour et al., 1975).
CH3OH o CO+ 2H2 (2.2)
CO+ H20 CO, + H, (2.3)
With its high hydrogen yield from a liquid fuel, the methanol-steam reforming
reaction is being studied as a possible method to provide hydrogen for fuel cell vehicles.
Methanol-steam reforming has been shown to successfully produce hydrogen for \vehicle
applications (Kaufman, 1992), yet the process is hindered in transportation applications
by characteristics of slow transient response (Dusterwald, 1997), temperature gradients
(Wild and Verhaak, 2000), and large volume and mass (Ohl et al.,1996).
Limiting Steps in the Reformation Process
For fuel cell vehicle applications, size, weight, and transient response are vital
factors. In order to design an optimal reformer for fuel cell vehicles one would desire the
smallest possible, lightweight reformer that would have enough catal st surface area to
react the fuel into a hydrogen-rich stream at the maximum design flow at the end of the
useful lifetime. Steam-reforming of hydrocarbons can be limited by heat transfer rates,
diffusion or mass transport, chemical kinetics and/or by degradation of the catalyst with
time. These limitations are evidenced by restrictions in reformer capacity, slow transient
response and/or limited lifetime.
Because of its endothermic nature, the reformation of liquid hydrocarbon fuels
requires heat transfer to the reactants and the catalyst. In a steam reformer heat comes
from an external source and is transferred through the catalyst bed housing wall and into
the catalyst bed as shown in Figure 2.1. Because of this endothermic nature and irregular
Figure 2.1: Simplified representation of a packed
bed methanol-steam reformer.
flow patterns associated with packed bed catalysts, the typical steam reformer is plagued
with temperature gradients. Typically high temperatures are found nearest to the heating
surface, and low temperatures are found toward the midpoint of the catalyst bed.
Furthermore, if the heat transfer rate to the midpoint of the catalyst bed is not high
enough, unreacted species may pass through the reactor at this location resulting in a
limited reformer capacity. Additionally the catalyst may be damaged by condensing
species where heat transfer is insufficient. Heat transfer can take place as convection,
conduction, or radiation. Radiation is typically negligible, as the local temperature
difference is relatively small. Heat conducts through the catalyst bed housing wall and
possibly into the first layer of catalyst particles coming into contact with the wall.
Because of point-to-point contact between the catalyst pellets, conduction from catalyst
pellet to catalyst pellet is minimal, and the bulk of the heat transfer to the center of the
packed bed is via convection through the fluid inside the reactor. As the explained
temperature gradient manifests, heat transfer via convection is typically the limiting
factor. Heat transfer limitations can be compensated for by an increased size of the
reformer to increase the heat transfer area (Ohl et al., 1996). Such design results in
undesirable large processors and corresponding high mass. The large size and thermal
mass then penalize the process for transportation applications.
Diffusion can also limit the reformation process. The processes of transport
within the catalyst bed and conversion to a hydrogen-rich gas is represented in Figure 2.2.
As this figure shows, the hydrocarbon and steam mixture must first be transported to the
surface of the catal st. Next the mixture must diffuse through the pores to an open
catalyst site to react. The reaction rate on the catalyst is then governed by chemical
Step 1 Step 7
Step 1- Reactants diffuse through bulk
stream to catalyst surface
Step 2- Reactants diffuse through
catalyst pores to open reaction site
Step 3- Reactants absorb onto catalyst Step 6
Step 4- Reaction time governed by Step 24
chemical kinetics on the catalyst
Step 5-Products adsorb from catalyst
Step 6-Products diffuse through catalyst
pores to surface
Step 7- Products diffuse from surface to
Figure 2.2: Simplified representation of the reactants moving to the open catalyst site
and products moving out to the bulk stream.
kinetics. Products of the reaction must then diffuse back out of the catalyst pores and into
the bulk flow stream. Increasing the amount of catalyst sites available can compensate for
limitations in diffusion. Normally this means increasing the amount of catalyst to provide
sufficient reaction area and open reaction sites closer to the surface of the catalyst,
allowing for a decreased diffusion distance. This typically requires an increase in the
reactor volume or conversely, a decrease of mass flow rate of the reactants. This is a
decrease in what is known as "space velocity" as shown in Equation 2.4. This decrease of
space velocity once again results in larger reforming processors or a more limited
processing capacity for a set reformer volume.
Volumetric Flow Rate (.
Space Velocity = (2.4)
Volume of Reactor
Slow chemical kinetics can usually be alleviated by increasing the temperature.
Because temperature gradients exist within the catalyst bed, temperature cannot be
increased at the midpoint of the catalyst bed without possibly damaging the catalyst near
the wall surfaces by overheating. Thus, restriction by chemical kinetics is normally a
manifestation of convective heat transfer limitations in the center portion of the catalyst
In methanol-steam reformation, performance decreases as the catalyst ages. The
lifetime requirements of transportation applications require the reforming processor to be
over-designed to order to compensate for performance degradation. Once again this
overdesign results in larger processors and increases under-hood volume requirements in
transportation applications. Catalyst lifetime can be affected by poisoning, fouling, or
sintering of the catalyst. Deactivation due to poisoning occurs with chemisorption of
impurities on reaction sites (Farrauto and Bartholomew, 1997). In reduced copper-zinc
catalysts, oxygen is a poison as the metals are changed to an oxidized state and thus are
unavailable for the steam reforming process. Fouling is the physical blockage of reaction
sites and can take place when fluid phase molecules contact the catalyst or as solid
material (e.g., carbon) collects on the surface (Sterchi, 2001). In pelletized catalysts
%where the interior area is dominant, fouling at the catalyst surface or pores can
significantly change the available catalyst area. Sintering or a change in catalyst structure
normally occurs with thermal stress of the substrate material. The change in structure
reduces the effective area of the catalyst. In order to avoid sintering. o\ erheating. rapid
thermal cycling, and thermal gradients should be minimized.
It is proposed that some of these limiting steps in the methanol-steam reformation
process may be minimized by imposing an acoustic field in the critical flow paths. The
proposed benefits due to this acoustic field are expected as a result of increased heat
transfer and mass transfer rates.
Acoustics in Past Processes
In several past studies (see overview by White, 1991) an acoustic wave
superimposed on a flow field has been shown to change the characteristics of the flow
field. Periodic plane waves in a wave guide can be represented by their pressure (p) and
velocity (u) components as shown in Equations 2.5 and 2.6 (Kinsler et al., 1982). In these
equations x is position, t is time, j is the imaginary constant, po is the non-oscillating
( 2nf 2nf
p=Ael c +Be c(2.5)
p- Ae+ 2ft2xf2
A j(2nt- 2fx) B (it 2+2-f x)
u -- e c + -- e (2.6)
density and A and B are found from the initial and boundary conditions. A general
representation of an acoustic wave at a frequency (f) in combination with a steady flow
component within a tube is shown in Figure 2.3. In this figure the acoustic pressure
(shown as the shading variation) drives the acoustic velocity. When the acoustic velocity
component (u') is larger than a left-to-right steady flow component (U), the motion of the
fluid periodically changes direction depending on space and time. The use of resonance
can allow for high acoustic pressures and velocities to build with the resulting standing
Steady Flow (U) with a Superimposed Acoustic Wave
Instananeous Velocity U+U U-U
Motion direction for
< .U t=0
Amplitude (U) Trace
$ Motion direction for
t= 1/2f U
Instananeous Velocity U-U U+U
Maximum ; Minimum
Figure 2.3: Flow fields at different times for a plane wave superimposed onto a steady
waves, yielding a periodic change of motion although now dependant on the spatial
position associated with the specific mode of the standing wave.
In the development of fluid mechanics it has been noted that acoustic waves can
influence several parameters of the flow field. These parameters include the boundary
layer and flow characterization. Stokes' 2nd problem solved for the velocity profile near a
flat plate with an oscillating stream (see discussion in White, 1991). The solution of this
problem allows calculation of the laminar boundary layer thickness in this oscillating
flowfield. The boundary layer is represented by Equation 2.7, where 6 is the boundary
S= 6.5 (2.7)
layer thickness, v is the kinematic viscosity of the fluid, and f is the frequency of the
wave oscillation. The acoustic field may also change the flow regime as discussed by
White (1991). The kinetic Reynolds number (Rek), which is the Reynolds number for
oscillating flowfields is shown in Equation 2.8 for pipe flow. In this equation, ro is the
radius of the pipe or a characteristic length. Like the standard Reynolds number the flow
may become turbulent when this value approaches 2000. Turbulent flow regimes are
characterized by chaotic motion of the fluid resulting in increased mixing, and for pipe
flow a flattened velocity profile as opposed to a parabolic profile for a fully developed
The acoustically thinned boundary layer in laminar flows as well as a jump to the
turbulent flow regime, due to the employment of an acoustic field has direct positive
implications on heat transfer (Fraenkel et al., 1998). In combustion studies acoustic wave
amplitude has been correlated to heat transfer (Kwon and Lee, 1985), temperature
(Crocco, 1969), emissions (Keller and Hongo, 1990), particle dispersion (Barrere and
Williams, 1969), mixing of species (Dubey et al., 1996), and flame structure (McManus
et al., 1990). An example of the effect of acoustics on flame structure is shown in Figure
2.4 (Photographs by author). These photographs of propane diffusion flames show
marked changes in the flow field due to acoustic oscillation. Enhanced mixing is
demonstrated by the compacted flame zone for the case with the oscillating flow field. A
change of flow field direction is manifested by the flame zone being moved downward
below the burner arms. Some of these changes seen in past processes can be expected
with implementation of an acoustic field in the steam-reforming process.
Figure 2.4: A 15 kW diffusion propane flame without
oscillating flow (A) and with oscillating flow (B).
Steam-Reforming in Combination with Acoustics
Superimposing an oscillating flow in the axial direction of the reformer (also
referred to as organ pipe oscillations) with the steady flow component is proposed
(Erickson and Roan, 2000). High acoustic pressures and corresponding local particle
velocity perturbations are achieved by establishing resonance in the reactor. If the
acoustic velocity component is high enough, the steady flow component carries the
overall flow forward and the oscillating flow portion incrementally steps the fluid
forward and backward across portions of the catalyst bed. This allows for a "mixing" of
the boundary layer along the reactor walls and past the catalyst particles. Under the
proper conditions, increased heat and mass transfer rates also occur as a result of the
thinned boundary layers and acoustic-induced mixing in the reactor. With increased heat
and mass transfer, temperature gradients are also reduced throughout the acoustically
enhanced reactor. The projected effects of acoustics on the limiting steps of heat transfer,
mass transfer and chemical kinetics of the overall process are discussed as potential
theory-based models are developed below.
Increased Particle Path Length due to Acoustics
A simplified model of the acoustics shows a significant increase in the particle
path length, resulting in a decreased effective space velocity, without changing the overall
bulk residence time. In order to illustrate this effect, consider an empty open-ended pipe
(without obstruction from the catalyst). Acoustic displacement amplitude is given in
Equation 2.9 (Halliday and Resnick, 1986), where Ym is the maximum acoustic
displacement amplitude (m), Pm is the maximum acoustic pressure (Pa), f is frequency
(Hz), p is the non-oscillating density of the gas (kg/m3), and c is the sonic velocity in the
gas (m/sec). The equation for maximum particle velocity due to acoustic oscillation (U,)
(Kinsler et al., 1982) is shown in Equation 2.10. This particle velocity is subsequently
Um = (2.10)
referred to as acoustic velocity. In order to derive the total path length seen by a particle
traveling through the reactor (effective particle path length), one must consider the length
of the reactor (original particle path length with no acoustic oscillation) and distance
traveled by a particle due to acoustic oscillation.
A simulation of the particle path can be made by investigating the velocity
components during the residence time. The slip velocity, or velocity relative to a
stationary catalyst pellet or wall is the sum of the oscillating velocity component Um
(multiplied by a spatial factor (Sf) resulting from the mode shape of the standing wave)
and average velocity (Vs). The spatial factor is assumed to follow a cycle sine wave
pattern within the reactor. The absolute value of the total velocity multiplied by the
residence time yields the slip displacement (Xsii,) past the catalyst or wall. If the time is
decreased to a small time step (At), then summed over the residence time (t,) the slip
displacement can be represented by Equation 2.11. The slip displacement can be
nondimensionalized by dividing by the reactor length. Plotting this dimensionless slip
displacement versus the velocity ratio (Um/Vs) Equation 2.11 yields the total theoretical
slip displacement in Figure 2.5
Xsp = I(SU,m sin2ntft + U,) At (2.11)
This figure shows that an slope asymptote of 0.407 is dee eloped at large velocity
ratios. This limiting slope can be developed through investigation of the displacement
due to the acoustic field within the reactor during the residence time. As before,
assuming that a standing wave pattern equivalent to a single cycle sine pattern for
velocity is developed in the reactor, a spatial average displacement amplitude of 0.64 of
E Total Theoretical
S4 Slip Displacement
SQ 2 Slip Displacement
S 1 due to Acoustics
0 5 10 15 20
Velocity Ratio (Um/Vs)
Figure 2.5: Total slip displacement and slip displacement due to acoustics for increasing
the maximum amplitude will occur. The equation showing the displacement of the
oscillating portion while residing in the reactor is shown as Equation 2.12 and further
simplified in Equation 2.13, where Xs,,p is the slip displacement (m), L is the length of the
Xsi = (4) x x 0.64 x f
21tfp oc V
X,,p 0.407 x
reactor (m), and Vs is the steady flow component of the reactor (m/s). One can see that
the slip displacement is independent of frequency in this model. This results from the
fact that while an increased frequency does cause more oscillations during the residence
time the amplitude of those oscillations is decreased proportionally for a given acoustic
As Figure 2.5 indicates, the model predicts large increases in particle path length
for decreasing average velocity or increasing particle acoustic velocity by higher acoustic
pressure levels. For a reasonable example, a sound pressure level of 165 dB (re 20pPa) in
air and a steady flow component of 1 m/s yields a particle path length of 7 times the
particle path length with non-oscillating flow. If employed in a reforming process the
effective space velocity would be 7 times smaller than the actual space velocity or the
space velocity without acoustics. This effect is diminished with the longer flow path and
wave attenuation due to obstruction from catalyst pellets. Also because of the fact that
the catalyst is not continuous throughout the catalyst bed these results must be scaled by
the catalyst packing factor and a factor accounting for the effective area of the catalyst
bed seen by the acoustic wave (ie. Exterior of the catalyst pellet). However this model
yielding the effective space velocity gives an upper boundary of the expected benefits due
to an acoustic field in conjunction with methanol-steam reformation in an other\\ ise
undisturbed and unobstructed (without catalyst pellets) flow field.
Increased Heat Transfer Rate due to Acoustics
A more realistic expectation may be that of modeling an increase in capacity due
to increased heat transfer in the reformer. As presented above, heat transfer rate is seen as
a fundamental resistive step in the fuel reformation process. Recent papers have
addressed this issue (Ohl et al., 1996; Daister\nald et al., 1997; Nagano et al., 2001) and
certain methods of increasing heat transfer rate have been reported, including the use of
fins to increase the heat transfer area (Shiizaki et al., 1999). However, increasing heat
transfer rates with acoustic fields for other applications has been well documented over
the years (Fraenkel et al., 1998: Erickson et al., 1997: Zinn, 1992: Yavuzkurt et al., 1991)
Space and time averaged Nusselt numbers have been shown by Ha and Yavuzkurt
(1993), for a single spherical particle in a pulsating flow with a steady flow component, to
be represented by Equations 2.14 and 2.15. In these equations, V, is velocity ratio
for V, > 1 =0.73839V, +0.16161 (2.15)
for V < 1 = 1+0.009608 V, -0.109608 V2 (2.14)
(previously defined as acoustic velocity U, divided by average velocity Vj), Nu, is the
Nusselt number due to pulsating flow, and Nus is the steady flow Nusselt number. Nusselt
number is a direct indicator of convective heat transfer rate as shown in Equation 2.16,
where h is the convective heat transfer rate coefficient, k is the thermal conductivity of
Nu =hL (2.16)
the gas, and L is a characteristic length. Figure 2.6 is a plotted representation of
Equations 2.14 and 2.15. While interactions between particles in a fixed bed reactor are
0 10 20 30 40 50
Acoustic Velocity IAverage Velocity
Figure 2.6: Nusselt number ratio versus velocity ratio for a single particle in an
clearly much more complex than the single particle example, Figure 2.6 shows the
potential for increasing convective heat transfer rates with acoustic fields. For the
previous example of a SPL of 165 dB and a 1 m/s steady or average velocity component,
the convective heat transfer rate is increased by a factor of 4 by using an acoustic wave.
Fraenkel et al. (1998) have found similar results for velocity ratios above I when flow
oscillations are used in grain drying.
The increase in heat transfer rate, by definition, allows a faster response to
transients including load following and start-up. Mass flow capacity can also been seen
to increase by referring to a simplified energy balance within the catalyst bed. From an
energy balance in the reformer (taking into account convective heat transfer from the
walls to the gas in the catalyst bed), allowable reactant mass flow rate is directly
proportional to the heat transfer rate from the hot wall to the gas as shown in Equation
2.17 where rh is mass flow rate, h is the convective heat transfer rate coefficient, A is the
m = l oca' (2.17)
heat transfer area, T is temperature, and AH is the change in enthalpy between reactants
and products. This relation applies for heat transfer limited cases that are not limited by
diffusion or chemical kinetic rates.
Increased Overall Reaction Rate due to Acoustics
With a higher heat transfer rate, the rate limiting steps of diffusion and chemical
kinetics become more important to the reformation process. Although mass transfer rate
has not been a perceived limitation in previous work (Asprey et al.. 1999: Pcppley et al.,
1997), smaller reactors with high flow rates will deviate from the plug flow reactor
assumption. As described and quantified by Ha and Yazavkurt (1993), acoustic mixing
increases the mass flow rate at the surface of a particle; thus in a reformer, acoustics
would reduce the diffusion time of a mixture to reach the outside of a catalyst. Equations
2.14 and 2.15 apply to the Sherwood number ratio as well as to the Nusselt number ratio
as presented. The Sherwood number (Sh), as shown in Equation 2.18 is analogous to the
Sh = (2.18)
Nusselt number in relation to convective mass transfer rate. In Equation 2.18 hm is the
convective mass transfer rate coefficient, DAB is the diffusivity of gas A into the gas B,
and L is a characteristic length. This increase in convective mass transfer should
increase the diffusion of both the reactants to the catalyst particle surface and the products
to the bulk stream. Internal diffusion or mass transfer into the core catalyst area is not
expected to be significantly affected by the acoustic waves due to extremely small
dimensions and velocities, thus overall diffusion rate enhancement by acoustics may be
limited. If the internal diffusion is not the limiting factor, but rather the process is limited
by external diffusion, acoustics may have a significant effect.
While increased mixing from acoustics may decrease a rate limiting diffusion step
to a catalyst surface, there is probability that the chemical kinetics will be increased as
well. Considering Arrhenius type rate equations, reactions will proceed more quickly for
higher temperatures. With the reduction of local temperature gradients due to acoustically
induced mixing and increased heat transfer, the reactor controlled by local temperature
feedback (as discussed by Sterchi, 2001 and Amphlett et al., 1999) would have higher
temperatures near the entrance of the reactor and thus higher chemical kinetic rates would
result. Because of the rate limiting step of methanol decomposition occurs toward the
entrance of the catalyst bed, minimization of the temperature gradient at this point is vital.
Additionally, Tonkovich et al. (1999) have found that for the water gas shift reaction.
intrinsic reaction times can be on the order of milliseconds rather than the observed
reaction times of 6-9 seconds. They exploit the fast intrinsic reaction times by using
micro-channel reactors where fast heat and mass transfer rates govern. In a similar
approach, acoustic vibrations should increase heat and mass transfer rates at the exposed
catalyst surface and thus may decrease the required contact time on the catalyst surface.
Although it is obvious that not all of the area in a catalyst would be affected by the
acoustic wave (due to internal pores etc.), bounds of the reaction rate can be established.
The lower bound would be the case where the reaction rate is limited only by internal
diffusion rate and thus the overall reaction rate would not change when acoustics are
applied. The upper bound would be where the rate is wholly limited by diffusion rates at
the surface and the total effective area (derived from the effective length) would be used
in the reaction rate models. It is expected that with catalyst pellets, internal area and
diffusion to those sites should dominate, thus deviating from the upper boundary
predicted by use of the effective area. With screen or monolith type catalysts, where
acoustic waves would have more of an effect on the active reaction sites, the actual rate
may follow predicted rate using the effective area (derived from effective length in Eq.
2.11) more closely.
IExension of Catal st .ife by Using Acoustics
In some aspects, catalyst life is also expected to be positively affected by using an
acoustic field in the reformation process. Acoustics may hinder catalyst degradation due
to poisoning, fouling and sintering. While all types of poisoning may not be affected by
the acoustic wave, poisoning by oxygen with reduced catalysts could be mitigated by the
oscillating flow under the proper conditions. With an oscillating flow, hydrogen
produced in the reformer would be recirculated to the upstream catalyst particles,
effectively re-reducing any oxidized catalyst. This recirculating characteristic of the flow
field would happen at velocity ratios (V,) above 1. Fouling is also a degradation
mechanism that would theoretically be minimized by employing an acoustic field. Much
like the oscillations of a sonic cleaning application, clogging at the surface reaction sites
and catalyst pores could be reduced by the oscillating action of the passing fluid.
Reduction of certain types of sintering may also occur as spatial temperature gradients
throughout the reactor are alleviated and temperature fluctuations are minimized by
There also may be a mechanism for catalyst degradation due to acoustic fields.
Theoretically, oscillating pressure fields may break apart the catalyst structure or drive
particles into the pores of the catalyst increasing the fouling mechanism or creating the
effects of catalyst sintering.
In theory, the possible benefits which arise from employing acoustic waves in the
reforming process are an increase in capacity for a given catalyst mass, a decrease in
transient response time, an increase in the overall efficiency of the reformation process,
and potential extension of catalyst life. The benefits of using acoustics in related
processes including combustion reactors have been shown in the literature as increased
convective heat transfer rates, increased mass transfer rates, and increased specie mixing.
While the acoustic field may not significantly affect diffusion and chemical kinetics for
all catalysts, especially those with high amounts of internal area not affected by the
acoustic wave, the theoretical investigation points toward potential for enhancing
reformation processes with acoustic waves.
The present study contributes to the technical field in that it is both a theoretical
development as presented above and an empirical quantification of the benefits that arise
from using an acoustic field in conjunction with the steam-reforming of a hydrocarbon
fuel. As shown by the literature review, a properly applied acoustic field can benefit other
processes. The acoustic field in the reformation process results with similar enhancement
and is expected to allow smaller size and lower weight of the reforming reactor, which in
turn would yield quicker start-up times and faster dynamic response. The lifetime of the
catalyst is expected to be extended by utilizing acoustic waves. The present investigation
quantifies these benefits and may contribute towards a practical method of reforming a
liquid fuel with high efficiency for fuel cell vehicles.
EXPERIMENTAL APPROACH AND FACILITY
This chapter discusses the experimental approach and facility. A description of
the methanol steam reforming facility is presented. Various subassemblies of the facility,
as well as instrument and control interfaces are described. This includes discussion of the
acoustic field generator and schemes for controlling the temperature and acoustic field.
The instrumentation is reviewed along with the methods used for collecting and analyzing
data. Modeling of the reformation process is discussed, which discussion includes the set
up of a factorial experimental design.
The experimental approach was designed to quantify the effect of the acoustic
field on the reformation process. This included adaptation of a reformer facility to accept
acoustic waves, development of procedures for operation of the facility, establishment of
temperature and acoustic control in the catalyst bed, the acoustical analysis of that
facility, and the design of an experiment from which, a statistically valid empirical model
could be developed. Further data were taken to show curvature of the model, verify the
model results, demonstrate the effect of acoustics on catalyst degradation, power demand.
as well as on the reactor temperature profile and controllability. Analysis of the data
enabled construction of a hydrogen power output per acoustic power input ratio which
quantifies the energy benefit/cost on the reformation process from the acoustic field.
A reforming facility was previously constructed at the University of Florida Fuel
Cell Research and Training Laboratory. This reforming facility, shown in Figure 3.1
(photograph by J. Patrick Sterchi) has been used in the past to study the effects of
Figure 3.1: Reforming facility at the University of Florida Fuel Cell Research and
impurities in methanol on the steam-reforming process (Sterchi, 2001). The reforming
facility hardware consists of pumps. vapori/ers, superheaters, two electrically heated
reforming reactors, a condensing unit. a po\cer relay box, and thermal instrumentation in
connection with computer control for the power relay box and pumps. As the facility \was
originally designed to simultaneously operate two reforming reactors, the present study
utilized only one side of the total facility. Figure 3.2 is a simplified schematic of the
Figure 3.2: Simplified schematic of the methanol steam reforming processor.
system used. In this system methanol:water premix is pumped through an electrically
heated vaporizer. The vaporized gas is routed through an electrically heated liquid trap or
secondary vaporizer. This liquid trap insures complete vaporization of the premix
entering the reformer at high flow rates. The gaseous premix then may be routed to the
exhaust or directed to the superheater by manipulation of the valves shown. If routed to
the electrically heated superheater the premix follows the path shown to the catalyst bed.
An oscillating pressure is directed through the catalyst bed from a nitrogen-purged and
water-cooled acoustic driver attached at the lower end of the catalyst bed. After traveling
through the electrically heated catalyst bed the reacted mixture can be routed through the
condensing unit or to the exhaust by manipulation of the valves shown. After passing
through the condensing unit the dry gas may be sampled through a gas sampling port as it
travels to the exhaust. A carbon dioxide purge unit attaches to the superheater assembly
to keep the superheater and catalyst bed free from oxygen during non-operating periods.
The overall system was divided into separate subassemblies. The methanol:water
premix reservoir, pump, and flow metering devices were grouped into the pumping
.%uba..sembhly. The vaporizer, liquid trap (or secondary vaporizer) and the superheater
comprise the vaporizer subassembly. The reactor subassembly consists of the catalyst bed
and acoustic components. These acoustic components are the acoustic driver, a nitrogen
purge, a water cooling system, a pressure transducer, and adapters. The condensing unit
and gas sample port make up the condenser subassembly. Computer control and logic for
both temperature and acoustic field regulation is explained in the control uha~vremhly.
Each subassembly is more fully described as a separate section below.
The pumping subassembly consists of a reservoir placed on a scale, a peristaltic
pump and associated tubing as shown in Figure 3.3. The scale and a timing device was
used to calculate the actual mass flow rate of premix fed into the reactor. The scale used
was an Acculab VI-4kg that had a resolution of 1 gram. The peristaltic pump was a
lasterflex L/S Model 7519-20 cartridge pump. Rough control of premix mass flow rate
%\as enabled by changing the volumetric flow setting on this peristaltic pump. The
volumetric flow setting could be set at increments of 0.1 ml/min. The actual mass flow
rate was found by measuring the mass change of the reservoir per unit time.
Figure 3.3: Artistic representation of the pumping
The vaporizer subassembly consisted of a vaporizer and a liquid trap or secondary
vaporizer as shown in Figure 3.4, and the superheater shown in Figure 3.5. The vaporizer
was a 1.27 cm (0.5 in.) O. D. stainless steel tube cut to 85 cm (33.5 in.) in length. The
length of the tube was wrapped in a 244 cm (96 in.) 650 Watt heat tape. The liquid trap
or secondary vaporizer was also made of 1.27 cm (0.5 in.) O. D. stainless steel tubing
that was cut to 64 cm (25.2 in). This tube was wrapped in a 183 cm (72 in.) length 435
Watt heat tape. The temperatures at the pipe surfaces and the exits were monitored by
stainless steel sheathed k-type thermocouples. The superheater assembly as shown in
Figure 3.5 is a 2.68 cm O.D. (nominal 3/4" schedule 40) stainless steel pipe cut to a 38
cm (15 in.) length. Receptacles for fittings (7/8 in.-14) were machined into both ends of
the pipe. The pipe was wrapped in a 366 cm (144 in.) 430 Watt ceramic beaded heater
85 cm length
650W Heat Tape
1.27 cm O.D.
64 cm length
435W Heat Tape
Liquid Trap or
0.635 cm O.D.
*Not to Scale
Figure 3.4: Artistic representation of the vapori/er and liquid trap.
and outer surface temperature was measured near the bottom of the assembly by a
sheathed k-type thermocouple. The exit of the pipe was made by allowing a 15.25 cm
CO, i -:
2.68 cm O.D. Stainless
Steel Pipe 38 cm length
430W Ceramic beaded
To Catalyst Bed
0.635 O.D. Stainless
Figure 3.5: Artistic representation of the superheater
(6 in.) long 0.635 cm (0.25 in.) O.D. stainless steel tube to be placed at the bottom of the
assembly recessing through the lower fitting into the superheater cavity. This recessed
outlet insured that no liquid phase fluid accumulating at the bottom of the superheater
would be allowed into the catalyst bed. The exit temperature was monitored at the exit of
*Not to Scale
the recessed outlet pipe by a stainless steel sheathed k-type thermocouple inside the
fitting directly above the catalyst bed.
The reactor subassembly consisted of the catalyst bed, and acoustic components.
These two sections are discussed individually below.
As shown in Figure 3.6, The catalyst bed was housed in a 2.68 cm O.D. (nominal
3/4 in. schedule 40) stainless steel pipe cut to a 63.5 cm (25 in.) length. Receptacles for
fittings were made at both ends of the pipe. The bed was filled with a pelletized low-
temperature water-gas shift catalyst manufactured by ICI Katalco (ICI catalyst 51-3). The
catalyst was designed for a temperature range of 200-3000C. This catalyst is a
proprietary combination of copper oxide, zinc oxide, and aluminum oxide and had to be
reduced before use. The reduction process for this catalyst takes place over 40 hours at
1800C with a 2.3% hydrogen, 97.7% nitrogen gas mixture, at a gas hourly space velocity
of 300 hr'. The 40 hour period is followed by an hour at 210C with a 6.5% volume
fraction of hydrogen. The reduction finishes with a final hour at 2100C with a 12 %
volume fraction of hydrogen. This reduction process is explained by Sterchi (2001). The
pelletized catalyst was dimensionally 0.52 cm in length and 0.54 cm in diameter and had
a bulk density of 1.2-1.3 kg/liter. In order to avoid blockage of the exit, the catalyst bed
was supported by a stainless steel screen fixed at the top of the adapter for the acoustic
generator. The catalyst bed was divided into 6 zones. At the end of each zone and before
the first zone a 0.93 cm diameter (1/8 in. NPT) port was machined into one or both sides
1.0 cm O.D. Stainless
12.7 cm length
Kistler 206 Piezoelectric
Water Cooling Jacket
placed 9.525 cm
apart on center
250W Heating bands
2.68 cm O.D. Stainless
63.5 cm length
*Not to Scale
To Condenser or Exhaust
To Condenser or Exhaust
Adapter for Acoustic Field
Figure 3.6: Artistic representation of the catalyst bed housing.
of the stainless steel pipe. Fittings could be placed into the machined holes to extract
temperature or pressure information at each respective location in the bed.
The fittings were sealed by a high temperature silicone sealant. Above the first
zone, two ports are used for monitoring dynamic pressure and monitoring the exit
temperature of the superheater. Each zone was individually heated by a 250 Watt band
heater. Control for these heaters was determined by the temperature at the respective
zone exit as explained in the control subassembly below. Temperature was monitored
using k-type stainless steel sheathed thermocouples. Each fitting was machined to be
flush with the interior surface of the pipe at the zone exits. The control thermocouples
were aligned as to be near the interior wall location in the fitting. These temperatures
represented the interior wall temperature at the specified axial location. At the bottom of
the pipe an adapter was threaded into the machined receptacle to allow an acoustic driver
to modulate the catalyst bed pressure.
Transmission of acoustic pressure and associated velocity fluctuations are a
product of geometric conditions, and also the fluid properties. Additionally, transmission
is typically inhibited by obstruction in the acoustic path. In a packed catalyst bed, the
catalyst itself partially obstructs the propagation of acoustic waves. Notwithstanding this
difficulty, standing acoustic waves can build up within the bed as a result of wave
reflection at the ends of the reactor.
An acoustic field was generated by pulsating an acoustic driver attached to an
adapter at the bottom of the reactor as shown in Figure 3.7. The acoustic driver used was
a Klipsch Pro 1133 Horn Driver typically used in amphitheater applications. The method
of using loudspeakers to modulate pressure in a chemical reaction has been pro\ cn by
several researchers including McManus et al. (1990), Poinsot et al. (1989), Heckl (1988),
.ang et al. (1987), Zikikiout et al. (1986), and many others. The horn driver was attached
to mounting devices and directed into the catalyst bed as show n. As the horn driver was
*Not to Scale
Water Cooling \
Adapter (7/8 in.-14
To Exhaust x 3/4 in. NPT Pressure
x 1/8 in. NPT) AcotEnclosure
Figure 3.7: Artistic representation of the acoustic driver, driver housing, mounting plate
not designed to be used in relatively high temperature chemical applications, a water
cooling jacket was designed around a fitting to eliminate conduction from the catalyst bed
walls. This water-cooled section was also a heat sink for the reactor as it pulled heat
away from the bottom zone. A nitrogen purge was directed to both the front of the
diaphragm and the back of the driver housing. The nitrogen at the front of the diaphragm
was used to inhibit hot reacted species from flowing into the driver. The nitrogen purge
at the back of the housing was a safety precaution as there was possibility of electrical
ignition in the voice coils in the presence of hydrogen. The entire housing and fittings
were tilted upward at a 100 angle to allow any condensed species forming on the water
cooled section to flow down toward the exit of the reactor rather than condensing and
pooling in the driver or fittings. The driver housing and fittings were sealed with the same
silicone sealant as the port fittings in the high temperature zone.
The acoustic wave was monitored by a Kistler piezoelectric A206 pressure
transducer. This piezoelectric transducer used a Kistler piezotron coupler and was factory
calibrated at 15.3 mV/kPa. As previously shown in Figure 3.6, the pressure transducer
was mounted at the port above the first zone in the catalyst bed for most of the
experiments performed. This position was selected so that the acoustic wave would have
to travel through the entire length of the catalyst bed. The mounting took place through a
1.0 cm O.D. (1/8 in. schedule 40) stainless steel pipe cut to a 12.7 cm (5 in.) length and
an adapter to connect the pressure transducer. A water cooling jacket was constructed for
the mounting section but after monitoring temperatures at the pressure transducer it was
determined that water cooling was not necessary. Preliminary experimentation found that
water cooling at the transducer condensed liquid species directly above the catalyst bed.
Experiments performed for catalyst degradation required the elimination of this
condensation and possible intrusion thus the pressure transducer was removed for
experiments dealing with degradation. When the pressure transducer was absent in the
degradation experiments, acoustic driver feedback came from the voltage input to the
Condensing Unit Subassembly
A condensing unit was also used that cooled the exit gases from approximately
2500 C to 0 C. As shown in Figure 3.8, This unit consisted of a coiled 0.635 cm (0.25
in.) O. D. copper tubing through which, the exit gases flowed. The exterior of this copper
tubing was flooded with water from an ice bath. Reacted species would condense in the
Unit Dry Gas
Catalyst To Exhaust
0.635 cm O.D.
Water IceBath Condensate V
; Pump Trap
*Not to scale
Figure 3.8: Artistic representation of the condenser subassembly.
tubing and follow the flow path to a liquid trap where the liquid species would remain
and the dry product gases would be directed to the exhaust. The above condensing
components were housed in a refrigeration unit that decreased the heat losses of the ice
bath. Temperatures were monitored at the dry gas exit and at the interior of the
refrigeration unit with sheathed k-type thermocouples. A dry gas sampling port was
available as the dry gas products were directed to the exhaust.
Instrumentation and Control Subassembly
Control of the temperatures and acoustic pressure took place in the form of an
active closed loop and open loop schemes, respectively. The input and output signals for
these two control schemes are found in Figure 3.9. The instrumentation for these systems
were the output signals coming from the reactor. A description of the control schemes is
given in the following subsections.
Output Signal ..."".....
Thermocouple Signals (14)
ADC/DAC .. Distribution
.. ........... .
Computer 1 Computer 2
LabVIEW User LabVIEW Us
I-* |1|* II* II ** l -l -l^lIlll I ***
Figure 3.9: Input and output signals for temperature and acoustical control.
The heating elements discussed previously were computer controlled by feedback
from the thermocouple outputs. The 14 thermocouple signals were processed using a
lotech DBK-52 Thermocouple module that was linked to a data acquisition and control
program that was built by the author in LabVIEW. This temperature control program can
be found in Appendix A. Control of the 9 heating elements took place through an lotech
DBK-24 Digital Output module linked to the same program. Depending on the
indi\ idual location and temperature history sensed by the respective thermocouples, the
output module switched individual relays that in turn switched 120 V to the heating
elements through a solid state rela\ as shown in the control circuit display ed in Figure
3.10. The relays were contained in a power distribution box exhibited in Figure 3.11. As
shown by the control logic diagram found in Figure 3.12, control occurred as the
thermocouple signals were compared to a setpoint temperature. The slope of the
temperature signals was also monitored and compared to predetermined set points as was
the duration of the on position for each heating element. The ability to override the
control was given to the user but this option was not used for any of the experiments
presented. The setpoint matrix used for the automatic control of the temperatures is
found in Table 3.1. The temperature control scheme with the matrix in Table 3.1 was
found to regulate temperature more precisely than other control schemes previously
utilized with this steam reforming reactor. While previous studies did not require precise
Figure 3.10: Control circuit for heating elements.
+ Relay State 120 VAC \
5V DC Input Relay
Figure 3.11: Power distribution box.
> dT/dt < dT/dts, ,mt
Duration < Durationpm,,
Figure 3.12: 1 leading element control logic.
Table 3.1: Control setpoint matrix.
Setpoint DT/dt Maximum Duration
C (C / sec) (% time)
Vaporizer Surface 420 0.3 100
Vaporizer Exit 290 20 100
Trap Surface 480 0.3 100
Trap Exit 250 20 100
Superheater Surface 305 0.2 100
Superheater Exit 250 0.1 100
Zone 1 Exit 250 0 100
Zone 2 Exit 250 0 100
Zone 3 Exit 250 0 100
Zone 4 Exit 250 0 100
Zone 5 Exit 250 0 100
Zone 6 Exit 250 0 100
control of the instantaneous temperature, it was expedient in the present study to
eliminate possible variations to clearly identify factors affecting the reforming process.
Figure 3.13 shows 15 minute time scans of the Zone 1 exit temperatures with a control
scheme previously developed by Sterchi (2001) and the control scheme used in the
present study. Both temperature plots were acquired under similar operating conditions
with a control setpoint of 2500C with no acoustic wave, and a premix flow rate of 4.78
g/min. The major differences between these two control schemes are the control matrix,
the location of the feedback thermocouples, and the use of the liquid trap (or secondary
vaporizer) for the present study. Previous operation utilized feedback thermocouples at
the radial centerline of the catalyst bed while the control scheme used in the present study
obtains feedback from thermocouples at the interior wall positions. This position near the
0 2 4 6 8 10 12 14
x Cortrol Scheme by Stech : Corro Scheme for Present Study
Figure 3.13: Time plot of catalyst bed temperatures at zone 1 exit utilizing two
control schemes with a 250C setpoint.
wall allows a faster response time to the heat input and thus the overshoot in temperature
is reduced. The fluctuations in temperature over the period shown in Figure 3.13 for the
present study are representative of temperature measurements over an entire experiment
period and throughout the catalyst bed. These temperatures and fluctuations are also
characteristic for all the experiments performed in this study. The average and standard
deviation for all zone exit temperatures under the conditions described above are shown
in Table 3.2.
Table 3.2: Average and standard deviation for all catalyst bed temperatures utilizing a
Average Temperature Standard Deviation
Zone 1 Exit 251.4 1.6
Zone 2 Exit 251.2 1.5
Zone 3 Exit 250.9 0.9
Zone 4 Exit 251.2 1.3
Zone 5 Exit 254.1 1.0
Zone 6 Exit 251.8 3.4
__ Zone 6 Exit 251.8 3.4
An acoustic field was generated with an open loop control scheme. As previously
shown by the input and output signals in Figure 3.8, a monotone sine wave or swept sine
wave reference signal was induced with a Wavetek Sweep function generator. This
reference wave then was directed to the channel input of a QSC Mx-700 amplifier. The
amplifier was in turn connected to the Klipsch Pro 1133 Horn Driver through a pressure
housing. The amplified AC voltage seen by the driver was monitored by connecting a
handheld Fluke multimeter in parallel with the driver. The acoustic output was probed by
observing the conditioned output voltage from the Kistler piezotron coupler that was
connected to the piezoelectric pressure transducer as described previously. Both the input
(from the function generator) and the output signal (from the piezotron signal
conditioner) were simultaneously monitored by an analog oscilloscope and a digital data
acquisition unit programmed in LabVIEW by the author. These digital data acquisition
programs can be found in Appendix A. These programs included a spectral analysis
program for measuring the acoustic properties of the reactor via transfer functions and
anoscilloscope program for measuring and recording the sound pressure level data. The
frequency of oscillation was held constant throughout the experiment. The resonant
frequencies and acoustic field properties within the catalyst bed are presented and
discussed in Chapter 4. As shown in Figure 3.14, which is a typical time trace of the
acoustic pressure, there were both fluctuations in the acoustic pressure amplitude levels
and 60 Hz noise present in the transducer output signal. Because the noise fluctuations
increased the observed maximum amplitude, the average noise level amplitude found
after an experiment run was removed from the average maximum found during the
1/60 Hz 1/60 Hz 160 Hz
m 0 j -* .. -- --- --
.4000 4U -. ---. -4 -( -- 4-- -- U-- -
0 001 002 003 0.04 0,05 0 000 C7 006
Figure 3.14: Typical time trace of pressure in the catalyst bed while generating hydrogen
with an acoustic input of 650 Hz.
experiment run. This post processing yielded a more accurate representation of maximum
acoustic pressure and the corresponding sound pressure level found inside the reactor.
The sound pressure level (SPL) in dB was found by applying Equation 3.1 where P,,s is
the root mean square of the pressure wave amplitude and Pref is a standard reference of
20~Pa used throughout the results.
SPL= 20log :ns
Analysis of Data
There were multiple steps and sub-processes required for analysis of the collected
data. The sub-processes required for analysis were use of gas chromatography,
calculations of conversion, and calculations of space velocity.
An SRI model 8610-C Gas Chromatograph (GC) was used to analyze dry
products of the reactor. The GC was computer controlled using Peak Simple software as
provided by the manufacturer. The GC was equipped with a 91.44 cm (36 in.) molecular
sieve-packed column and a 182.88 cm (72 in.) silica gel-packed column. Tedlar bags (1.5
liter) were used to collect dry gas samples. After the tedlar sampling bag was connected
to the sampling valve on the gas chromatograph, a 40-50 ml gas sample was drawn
through the sampling valve with a syringe. The automated gas sampling valve was kept
heated to 100C and was used to inject a 1 ml sample into the columns upon command
from the computer user interface. A Bourdon type pressure gauge was used in connection
with the gas sampling valve to ensure that the sample in the valve was near ambient
pressure before the sample was injected. The flow paths for the GC sampling valve,
columns and detectors are shown in Figure 3.15. As the gas sampling valve would rotate
from inject back to load position (at 9.0 min), the flow in the columns would reverse.
This allowed separation of H, 02, N, CH4, and CO specie passing through the mol sieve
and separation of CO2 and higher hydrocarbons trapped in either the molecular sieve-
packed column or silica gel-packed column after the flow reversal (SRI, 1997).
Separation occurred and the silica gel-packed column was heated (molecular sie\ e-
packed column was kept at 100C for all experiments) as shown in Figure 3.16. The
detectors used were a thermal conductivity detector (TCD) and a flame ionization
detector (FID) in conjunction with a methanizer. The voltage traces from these detectors
and a calibration file were used by the software to calculate the volume percent of species
in the gas sample. The TCD was used to detect H2, 0, and N,. as well as high levels
TCD Methanizer- FID
Figure 3.15: Flow paths in the gas chromatograph.
E Flow Reversal
SH2 0 N2 CH4 CO CO2 Higher Hydrocarbons
0 5 10 15 20 25 30
Figure 3.16: Silica gel-packed column temperature and approximate specie elution times
for the species expected with the gas chromatograph.
. A -
(above 1%) of CH4, CO, and CO,. The FID was utilized to detect low levels of
carbonaceous compounds including CH4, CO, CO, and non-methane hydrocarbons
(NMHC). All of the above specie (except hydrogen) were detected using a ultra high
purity grade helium carrier gas and a high TCD current setting. Detection of H, levels
with the GC required use of an ultra high purity grade nitrogen carrier gas and a low TCD
Calibration of the GC took place as standardized gas samples were tested in the
GC. The calibration levels for each gas is shown in Table 3.3. With the exception of the
higher levels of hydrogen, these calibration mixtures were acquired from calibrated
sample gas cylinders.
Table 3.3: Gas chromatograph calibration levels.
Specie H2 02 N2 CH4 CO CO2 NMHC
1%, 4%, 10%,
S5%, 20 10%,
Calibration -65%,-70%, 5%, 20% PPM 100 PPM, Ethylene
levels -75%,-80%, 21% 5% 20 PPM
79% 4% 25%
Rough calibration of the higher hydrogen volume percentages was made by filling
a tedlar bag with metered hydrogen and nitrogen. As there was some fluctuation in the
metered values and these points were not independently verified, this calibration was
considered approximate. This rough calibration did however verify that the balance of
products not acquired using the helium carrier was hydrogen. This is shown by an
example of the volume percentages found for the dry product gases during preliminary
data at various operating conditions in Table 3.4 below. Table 3.4 also demonstrates the
variability due to the gas chromatograph sampling as Experiments 1 and 2 N\ere gathered
under the same operating conditions. Because of this demonstrated variability in o\ crall
species concentration, the overall average concentration of hydrogen was used to
calculate hydrogen output power in the analysis. Thus the hydrogen output power was a
function only of the dry product gas flow rate rather than being affected by the variable
concentrations during a particular experiment.
Table 3.4: Volume percent species of the dry product outlet gas.
Specie H2 02 N2 CH4 CO CO2 NMHC Total
Experiment 1 68.76 ND 6.52 ND 0.54 22.83 ND 98.65
Experiment 2 72.65 ND 4.04 ND 0.55 23.92 ND 101.16
Experiment 3 71.19 ND 5.21 ND 0.50 23.11 ND 100.01
Conversion of the methanol is defined as shown in Equation 3.2 where X is the
conversion of the methanol fuel, and methanol is determined on a mass basis. The
calculated conversion is then independent of the sample time period. The time period and
SCH3OHnp,,, CH3OH,,p (3.2)
corresponding amounts of condensate did, however, affect the variability associated with
the conversion. Shorter time periods yielded less condensate and thus these small
amounts of condensate were more prone to contamination from water droplets that
formed on the liquid trap when exposed to the atmosphere prior to analysis or \\aler that
would leak into the liquid trap upon opening. It was determined from comparing small
amounts of condensate from shorter experiment times to those of longer experiments
under the same conditions that a minimum of 40 grams of condensate should be collected
to insure negligible biasing of conversion data by sample collection time. The conversion
equation shown above in Equation 3.2 required that the methanol mass fraction in the
condensate be known. The amount of methanol in a condensate sample was determined
by utilizing the condensate density. The condensate density was found by way of a Anton
Paar density meter that had a 0.0001 g/cm3 resolution. The density (p) was input to an
emperical formula developed by Sterchi (2001) that yields methanol mass fraction
(mfCHIOH) of the condensate. This empirical relation is found in Equation 3.3 and was
verified for this study. From the methanol mass fraction of the condensate, the mass of
unused methanol (output) is found, and the conversion can then be determined.
mfcHnOH = 3859.9p2 823 .lp + 4370.0 (3.3)
Space velocity for this investigation was calculated from the mass flow rate of
methanol:water premix. Flow rate data was gathered as masses of both premix consumed
and condensate accumulated could be known over a unit of time. Space velocity is found
by measuring the volume of the reactor (measured catalyst bed length was used for
determination of the reactor volume), and the volumetric flow rate of the reactants. Space
velocity, as previously presented in Equation 2.4 can be defined in several ways. This
includes gaseous space velocity or liquid space velocity. Space velocity can also be
determined with regards to a single specie or all reactants traveling through a processor.
Additionally, considering an acoustic field within a processor, an effective space \ clocity
can be defined basing an effective reformer volume on a typical particle path length. In
order to avoid confusion, the space velocity with regards to liquid methanol is used.
Liquid Hourly Space Velocity of Methanol (LHSV-M), as defined in Equation 3.4 will be
used throughout the presentation of the results.
mhr liquid methanol input
mL reactor volume (3
Removal of Data Points
As multiple sets of data were used in the present study with variability in both
input and output parameters, output data were analyzed to identify outlying points using a
first application of Chavenets criterion (Holman, 1994).
Specific procedures and checklists for start-up, data collection, and shutdown
were made in order to collect all pertinent data as well as to avoid hazardous situations
and damage to equipment. These procedures were made by both consulting past studies
(Sterchi, 2001) foresight in the operating process, and experience. The checklists used,
which outline the general procedures for start-up, data collection, and shutdown can be
found in Appendix B.
Modeling the Acoustically Enhanced Reformation Process
As discussed in the theoretical approach, there are several analytical models that
may be applied to the various steps of the reformation process with acoustics. These are
increase of effective space velocity (Equation 2.11), heat transfer (using Equations 2.12
and 2.13 for Nusselt number), mass transfer (using Equation 2.12 and 2.13 for Sherwood
number) and chemical kinetics (using Equation 2.11 for the effective area). While each
of these models may represent a portion of the overall enhancement due to the acoustic
waves, an actual reactor does not fit the ideal models developed and other reactor
parameters must be considered. This is especially true in the packed bed reactor utilized
for this investigation because the models developed do not account for obstruction in the
acoustic path or the volume that the catalyst pellets occupy. While a comprehensive
model for the actual conditions may not be currently possible, empirical modifications of
the ideal model for the experimental facility can be developed. One should note that the
facility is not an optimal design for employment of acoustics or reforming and that the
resulting model is specific to the facility used. Although this is a significant limitation,
one gains understanding of how acoustics could be applied in different reactors and just
what the demonstrated benefits are in a particular application.
Factorial Experiment Design
A fraction of the data collected was used as a standard 2" factorial experiment
design. This type of experiment design is common for statistically identifying effects of
certain factors on a process and yields a model of the process with respect to the variables
investigated (Lawson and Erjavec, 2000). A brief overview of factorial experiment
design and analysis is given in Appendix C.
This factorial experiment design is used to develop an empirical model.
Although the model developed yields linear relationships for the variables investigated,
residuals or curvature of the output with relation to the inputs was also investigated with
further data. The output (dependent variable) for the factorial experiment design was
conversion of methanol. After consulting previous experimentation with the facility,
analysis of preliminary data, and alignment with the objectives of the study, the input (or
independent) variables were chosen to be fuel (premix) flow rate, acoustic pressure, and
catalyst bed length. Other possible variables were held constant for the experiments
including control setpoint temperature (2500C), frequency (650 Hz), and water-methanol
stoichiometry (1.5:1 on a molar basis). High and low levels were determined for each of
the varying parameters. The levels for fuel flow rate were a high of 9.5 g'mnin. and a low
of 4.8 g/min. The levels for acoustic pressure were a high of 5000 Pa, and a low of 0 Pa.
The levels for catalyst bed length were a high of 0.576 m, and a low of 0.279 m. This
allowed for a 23 factorial experiment design yielding 8 combinations of the input
variables at high and low settings. The input matrix and run orders can be found for the
factorial experiments in Table 3.5.
Table 3.5: Factorial input matrix and run order.
Flow Acoustic Catalyst Bed Run Order
Flow Run Order
+ + + 3,6,10
-+ + 1,4,11
+ -+ 2,7,9
+ + 14,18,19
+ -- 15,23,24
Although other parameters were held "constant", variability in these settings of
high and low levels was apparent. These fluctuations yielded variability in the measured
output of conversion. Two replications of the study were performed to allow statistical
analysis and identification of outlying points. As shown by the run order, the ariables
of fuel (premix) flow rate and acoustic pressure were randomized while catalyst bed
length was taken sequentially at a high then low level. The sequential order of catalyst
bed length was required because of the time-intensive reduction process between
changing bed lengths. Inadvertently the model includes variability from both the
reduction process and catalyst arrangement as a bed. This came as a result of rupturing
the seals of the facility halfway through the full (high level) catalyst bed experiments.
The consequences of this disturbance are that the variation in fuel conversion may be
higher in the empirical model than would be observed normally.
Further Data and Verification of the Model
Further data were required in order to verify the empirical model constructed and
to demonstrate different effects of the acoustic waves on the reformation process.
Curvature for the input variables of the model was found by investigating intermediate
sound pressure levels, flow rates and various bed lengths including those bed lengths
lower than the low level in the factorial experiment. This bed length data enabled one to
not only find the curvature and limits of the proposed model but also to estimate how the
local conversion of methanol progressed though the full catalyst bed.
Other data not directly accounted for in the empirical model were also collected.
As frequency was suspected to change the boundary layer thickness for laminar flow, and
possibly induce the flow into the turbulent regime, conversion at two frequencies was
investigated. Because the temperature within the processor was expected to significantly
affect the conversion of fuel, temperature data were collected for acoustic field and flow
conditions. While holding the temperature control setpoint at 2500C, actual temperature
data at certain locations throughout the catalyst bed were collected. These temperature
data were collected by utilizing thermocouples extending into the bed at specified lengths
through the unused sampling ports shown in Figure 3.6. Other parameters of interest,
including conversion change with time (indicating catalyst degradation), hydrogen power
output, and acoustic driver input electrical power were also collected. Specific
configurations of all controllable parameters for these further data is explained in the
discussion of those data in Chapter 6. The run orders for the additional data collected is
found in Appendix D.
PROPERTIES OF THE ACOUSTIC FIELD
This chapter is a presentation of the results found by investigating the acoustic
field within the catalyst bed and the analysis of those results. As the acoustic velocity
must be higher than the average velocity in order to enhance the reforming process,
standing acoustic waves built at resonance with their associated pressures and velocities
can be utilized. The analysis of the acoustic field provides the basis for understanding
what resonant acoustic modes can be employed in enhancing the reformation process.
Transfer functions were gathered for frequency analysis. These data were compared to
theory and enabled an acoustic modal analysis to be performed.
The Use of Resonance within the Catalyst Bed
As explained in the theoretical approach, enhancement of the reforming process is
expected from increased mixing within the catalyst bed when the acoustic velocity is
sufficiently high. High amplitude sound waves are made possible by building up standing
or resonant waves within the reactor. Resonant frequencies and sound pressure levels in a
wave guide such as the catalyst bed is a function of acoustic boundary conditions, the
speed of sound in the wave guide and the damping characteristics of any obstruction
within the waveguide. A simple example of a standing wave pattern of an acoustic field
within a wave guide and the analogous resonant motion of a string is shown in Figure 4.1.
The boundary conditions shown in these simple examples are the same as those expected
within the catalyst bed studied. The acoustically closed condition was expected at the
Displacement Amplitude trace in a wave
guide with closed-open boundary conditions
2nd mode resonant oscillation
A Open or Driven
String displacement analogy with
fixed-free boundary conditions
2nd mode resonant oscillation
Displacement and Displacement and
Velocity Anti-node Velocity Node
Figure 4.1: Standing wave pattern of displacement and velocity in a waveguide and
an analogous pattern of displacement and velocity in a driven string.
entrance to the catalyst bed. This entrance is where a 0.635 cm (0.25 in.) O.D. tube opens
to a 2.68 cm 0. D. (0.75 in. schedule 40) pipe previously shown in Figure 3.5. Acoustic
waves propagate upward from the catalyst bed to this junction. By investigating the
theoretical acoustic power transmitted from a large pipe opening to a small pipe opening
as shown in Figure 4.2, the expectation of a closed acoustic condition at the catalyst bed
entrance can be justified. The transmitted acoustic po\ er is shown in Equation 4.1,
%\here Tp is the transmitted acoustic power, S, is the large pipe cross-sectional area and S,
in the small pipe cross-sectional area (Kinsler et al., 1982). For the pipe junction
described above, the acoustic power transmitted through the entrance is calculated to be
Free or Driven
S < Pipe Cross- > S, Acoustic Wave
Figure 4.2: Transmission and reflection of an acoustic plane wave at a pipe
T (s, + s)2 (4.1)
less than 18% of the incident power. Thus acoustically the entrance location acts as a
closed end. The other end of the waveguide is where the acoustic driver is attached.
Although physically this is not a straight through path, acoustically it appears to be so. It
is expected that because of the displacement associated with the voice coil of the driver
that this location would act as an acoustically open condition with associated large
acoustic displacements and velocities. There are several resonant modes and frequencies
associated with these boundary conditions. The predicted resonant frequencies for these
closed-open conditions are shown in Equation 4.2 where n is the mode, c is sonic velocity
and L is the length of the tube. The pressure and velocity equations for any location and
time for the closed-open acoustic boundary conditions in an idealized a\ ve guide are
f= 4L n= 1, 2, 3,... (4.2)
given in Equations 4.3 and 4.4 where x is the position from the closed end (entrance) and
t is time. With the theory established one can measure the actual resonant properties of
the catalyst bed via transfer functions and perform a modal analysis to verify the predicted
(2n 7) nx
P = P co (2n- 1)n sin(2ft) (4.3)
= P2L )
u = U, sin (2n- )n) cos(2nf.t) (4.4)
In order to observe what resonant frequencies were available in the catalyst bed
several transfer functions were measured in the reactor. These transfer functions were
measured from the signal generator or driver input signal to the pressure transducer
output signal. Excitation of a range of frequencies was accomplished by a swept sine
method. Frequencies were input from 10 Hz up to 15 kHz. The spectrums of the input
and output were measured and compared. This resulted in a transfer function magnitude
plot. The transfer function phase plot and coherence plot were also measured. The
coherence was checked to verify that the output signal was due to the input signal and not
due to noise. The number of samples acquired for the transfer function was 8192. These
samples were taken at a sampling rate of 35 ksamples/sec resulting in a frequency
resolution of 4.27 Hz. Thirty single transfer functions were averaged to ensure the
excitation and capture of all desired frequencies. In this manner, transfer functions were
measured for a variety of temperatures and reactor conditions. These conditions were
ambient (-33 C) and operating temperature (2500C Control Setpoint), empty and full of
catalyst, as well both stagnant (no premix flow) and while generating hydrogen (4.9 g/min
premix flow)(see Figures 4.3- 4.6). The transfer functions for these conditions presented
in Figures 4.3 4.6 are a composite of two transfer functions for each figure. One transfer
functions is for the frequencies from 40 to 2100 Hz and the second transfer function is for
2100 Hz to 15 kHz. There are no significant frequency peaks beyond 7000 Hz in all
cases. The most basic of the configurations tested was an air filled ambient temperature
reactor with no catalyst. The transfer function for this condition is found in Figure 4.3.
0.0- I 1 I 1 I I 1 I ,
40 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
Figure 4.3: Transfer function magnitude for the catalyst bed housing containing only
This transfer function shows the frequencies at which resonance is theoretically possible.
Transfer functions were also found for a 33 C reactor containing catalyst and air as
shown in Figure 4.4. The amplitude peaks are lower than without the catalyst because of
the acoustic attenuation of the obstructing catalyst particles, yet the frequency peaks
corresponding to the highest amplitude modes were still well established. After reducing
500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
Figure 4.4: Transfer function magnitude for a catalyst bed housing containing catalyst
and air at 33C
the catalyst and holding carbon dioxide in the reactor at 250C, the transfer function, as
shown in Figure 4.5 revealed a small shift in frequencies due to a slight change of the
sonic velocity of the gas within the catalyst bed. The sonic velocity in 33 C air is 351 m/s
40 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
Figure 4.5: Transfer function magnitude for a reactor containing reduced catalyst and
CO, at operating temperature (250 oC).
while the sonic velocity in 2500C carbon dioxide is 357 m/s. With this slight difference
in sonic velocity only a small change in frequency positioning is expected. These
frequency peaks, under the conditions described were also found positioned at similar
relative locations as the air filled reactor with catalyst, indicating no shift in primary
oscillation modes. With a flow of reactants and products in the catalyst bed (premix flow
of 4.8 g/min) under operating conditions (250C control setpoint) Transfer functions
were acquired as displayed in Figure 4.6. As seen in this figure, the transfer function
shifted slightly in magnitude and frequency positioning when water/methanol premix was
being reacted because of both the slight differences in overall reactor temperature,
pressure, and gas composition (and thus sonic velocity) due to the reaction. The highest
amplitude peak previously near 550 Hz was shifted to 650 Hz and the peak previously at
1700 Hz shifted to 1900 Hz, while two low amplitude peaks became evident near 1400
Hz and 2300 Hz. An example of the time domain pressure signal within the catalyst bed
0.1- - -
0.0- ,l l li l
40 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
Figure 4.6: Reactor transfer function magnitude while generating h1 drogen at 250'C.
while being driven at 650 Hz and while reaction is taking place was presented previously
in Figure 3.14. It was determined that while generating hydrogen rich products, an
acoustic sound pressure level of 165 dB (re20 iPa) could be maintained in the catalyst
bed if driven at 650 Hz and approximately 9 watts of electrical power to the driver.
One can compare these experimental results of resonant frequencies found by the
transfer functions with theoretical resonant frequencies for the assumed boundary
conditions. Allowable resonant frequencies (f) for an closed-open waveguide have been
given in Equation 4.1. Noticeably, the frequencies as predicted by Equation 4.1 are not
all present in the measured transfer functions. This deviation from the analytical solution
is due to insufficient driving power at certain frequencies and/or to the actual boundary
conditions present in the reactor. The midrange driver has a frequency response as shown
in Figure 4.7 (Klipsch, 2001). Because the lower frequencies (below 200 Hz) are not
well excited by the midrange horn driver, it was not expected that the lower frequencies
would be excited in the reactor.
100 1000 10000
Figure 4.7: Frequency response of the acoustic driver as measured in a plane
wave tube by the manufacturer.
The actual resonant frequencies from the transfer functions can be translated to
their associated wavelengths (1) through Equation 4.2 With the wavelengths and at least
one boundary condition known, modal shapes can be plotted and the location of actual
boundary conditions can be found.
A .- (4.2)
Wavelengths in the 330C air-filled catalyst bed housing without catalyst were
calculated from the measured transfer function previously shown in Figure 4.3. With the
acoustically closed condition assumed as the catalyst bed housing entrance, the velocity
and displacement waveforms for six frequencies (amplitude has been normalized) were
plotted. These plotted waveforms are shown in Figure 4.8. This figure plots normalized
velocity magnitude versus position from the entrance, and has the temperature and
pressure ports along with suspected physical boundary conditions marked along the x-
MODAL SHAPES IN REFORMER
1.0- wave 328 Hz
0.8-I wave 2 550Hz
0.6 wave3- 756Hz
0.4- wave 4 982 Hz
02- \ wave5-1200Hz
Voi 0.0-e wave 6 1414 Hz
-0.6- -OPEN 57500 000 00i
/ OPEN. a05000 00 E
**-to-, 1, 1 1, ,,1 1 OPEN 87500 000 00 VfS
0 50 100 150 200 250 300 350 0 0 450 500 60 o s650o 750 80 80 eSo0 OPEN 87500 000
P1 P2 P3 P4 PS P6 and o refrm Adapter drive
20.30 115-125 210-220 305-315 395-405 490-500 epaion oa throat
575 800-815 875
Figure 4.8: Normalized velocity magnitude for six waves found in a 33C reactor without
From Figure 4.8, it is evident that an acoustically open condition exists near
800-810 mm inside the reactor. This open condition directly corresponds to the driver
mounting plate where the driver is attached to the pipes leading to the catalyst bed. This
location is termed the adapter throat. This acoustically open condition at this point
indicates that the mass of fluid inside the acoustic driver oscillates in unison with the
voice coils and diaphragm. Considering the large diameter (7.62 cm) of the diaphragm
and the relatively short distance from the diaphragm to the adapter throat (8.9 cm), this
unison in oscillation is not surprising. In addition to this boundary condition at the
adapter throat, there are other contributing boundary conditions for some of the frequency
peaks found in the transfer functions. Most notably at 550 Hz, 1700 Hz, and 2100 Hz for
the empty air filled reactor there is contribution from an open condition at the pressure
transducer (PT) tube and housing (see Figure 3.6). As expected, the frequency peaks with
contributing boundary conditions are higher than those without. This amplification of
certain frequencies with contributing boundary conditions explains why the real reactor
deviates from the idealized case presented by Equation 4.2. As can be seen from Figure
4.6, the contributing conditions also widen the frequency peaks which allows for better
control of the sound pressure level. As noted in the explanation of the experimental
facility there are fluctuations in overall temperature and gas composition with time that
can change the resonant characteristics of the catalyst bed. The \wide bands of resonance
found for frequencies with contributing boundary conditions allow for minimization of
the changes in sound pressure when acoustically dri\ ing at these frequencies.
The modal analysis was verified by moving the pressure transducer down the
reactor length through the six access ports in the air-filled reactor without catal) st. The
transfer functions, measured through the ports are shown in Appendix E for the lower
frequencies. As expected from the standing wave pressure amplitude trace, which can be
thought of as 90 leading the velocity trace in Figure 4.8, the transducer picked up higher
magnitudes pressures at the corresponding pressure anti-nodes. These shifts in the
transfer function magnitude and also in the corresponding phase plots were noted and
found to follow the expected modes as predicted by the waveforms in Figure 4.8. A
summary of the frequencies, wavelengths and boundary conditions for the air-filled 33 C
reactor without catalyst is shown in Table 4.1 below.
Table 4.1: Waveforms and boundary conditions for a 33 C reactor containing only air.
Frequency (Hz) Wavelength Boundary Contributing
(m) Condition Conditions
145 2.41 Open at reformer end
328* 1.06 Open at adapter throat
550* 0.63 Open at adapter throat Open at PT Tube
756* 0.46 Open at adapter throat
982* 0.36 Open at adapter throat
1196* 0.29 Open at adapter throat
1414* 0.25 Open at adapter throat
1550 0.23 Open at adapter throat
1717 0.20 Open at adapter throat (1650 Hz) Open at PT Tube
2050 0.17 Open at adapter throat
2285 0.15 Open at adapter throat
2555 0.14 Open at reformer end
2896 0.12 Open at adapter throat Open at PT Tube
3123 0.11 Open at adapter throat
3358 0.10 Open at adapter throat
3708 0.09 Open at driver throat
3900 0.09 Open at driver throat Open at PT Tube
5383 0.06 Open at adapter throat Open at PT Tube
6669 0.05 Closed-closed transverse wave Open at PT Tube
velocity amplitude mode shape is plotted in Figure 4.8
EMPIRICAL MODEL DEVELOPMENT AND ANALYSIS
This chapter presents and discusses the results of the 23 factorial experiment
design used to develop empirical correlation factors to model the reformation process
under the influence of a resonant acoustic field. The reactor was investigated for the
input variables of premix flow rate, sound pressure amplitude, and catalyst bed length at
high and low levels as described in the experimental approach. This investigation allowed
for correlation factors of an empirical model to be found. Model correlation factors are
presented, along with an uncertainty analysis of those correlation factors. Midpoint
curvature for each variable and the effect of frequency on the conversion level of the
steam-reforming process were also examined.
Experimental Results and Model Correlation Factors
By using a standard factorial design of experiments one can develop an empirical
model of a certain design space with relatively few experimental trials. An overview of
the general setup of such a process, formulation of the correlation factors, and statistical
analysis of hypothetical results are presented in Appendix C. Actual values for the high
(+) and low (-) levels investigated and the randomized run order were previously
presented in Chapter 3. The results of the experiments for the configurations specified
yielded the output data shown in Table 5.1. In these experiments the output or
independent variable is conversion of methanol in percent. The temperature control
setpoint was held constant as \Ncre all other controllable parameters not specified. As
shown in Table 5.1 the standard deviations are relatively low indicating good
repeatability for each configuration. The data also show that with only a few exceptions
the conversion decreases over the trials run. Because the trial run is correlated to the
relative sample time, these data suggest that catalyst degradation was causing the bulk of
Table 5.1: The input matrix and methanol conversion percentage data obtained from the
Flow Sound Catalyst Conversion
Bed Trial Trial Trial Standard
Run Rate Pressure % Average
(X) (X2) Length 1 2 3 Deviation
(X1) (X2) (Y)
1 + + + 96.33 96.28 94.94 95.85 0.79
2 + + 99.74 99.61 99.44 99.60 0.15
3 + + 93.06 92.22 91.84 92.37 0.62
4 + 99.13 98.81 98.90 98.95 0.17
5 + + 70.58 70.19 70.34 70.37 0.20
6 + 92.57 91.46 91.33 91.79 0.68
7 + 61.74 60.82 59.77 60.78 0.99
8 87.61 86.45 86.37 86.81 0.69
Equation 5.1 is used to find the standard error (SE) from the
deviation (op) and the total number of factorials (nf) for these data.
The total average
conversion ( Y ), pooled standard deviation, standard error, and total degrees of freedom
for the data are given in Table 5.2. These data suggest that mean values within the
conditions studied are expected to deviate less than or equal to the standard error (0.251
%) and any individual point should not deviate by more than twice the pooled standard
deviation (1.22 %) from its mean value using a 95% confidence interval. The correlation
Table 5.2: Total average conversion, pooled standard deviation, standard error, and total
de rees of freedom from the factorial experiments.
Total Average Pooled Standard Deviation Standard Error Total Degrees of
Conversion % Freedom
87.06 0.61 0.251 16
factors or effects (E) and interactions (I) were calculated for each variable (X) and set of
variables by taking the average of the average conversion outputs (Y) when the variable
or set of variables were high and subtracting the same average when the variable or set of
variables were low. This statement can be simplified to yield Equation 5.2 for the effects.
Sun Xi run (5.2)
Interactions are found in a similar manner where the combination of the level X,,ru
multiplied by Xj,rn is used instead ofX,,, in Equation 5.2 find the Interaction (I,). This
calculation, using the data presented in Table 5.1 resulted in the correlation factors for the
model as the effects and interactions displayed in Table 5.3
Table 5.3: The effects and interactions found from the factorial experimentation.
Effect of Flow Rate (E,) -14.44
Effect of Sound Pressure (E2) 4.67
Effect of Catalyst Bed Length (E3) 19.26
Interaction of Flow Rate and Sound Pressure (I12) 1.86
Interaction of Flow rate and Catalyst Bed Length (I,3) 9.28
Interaction of Sound Pressure and Catalyst Bed Length (I23) -2.61
Interaction of Flow Rate, Sound Pressure and Catalyst Bed Length (I,23) -0.45
These correlation factors show the magnitude and direction each variable has on
the conversion of methanol in the reformer. These model correlation factors yield the
equation for methanol conversion (C) as shown in Equation 5.3. However, not all effects
El E2 E3 I2 I3 123, 1.3
C = Y + X2 X, + X3 + X+ X, X + XX X3 + XXX X+' X3 (5.3)
2 2 2 2 2 2 2
or interactions presented above are necessarily statistically significant or need to be
included in the working model equation. In order to find the statistical significance of the
correlation factors, one must perform an uncertainty analysis on the effects and
In order to formulate the proper model correlation equation one must test each
effect and interaction for statistical significance. This uncertainty analysis can also be
used to test the significance level of the individual correlation factors. This is particularly
important for statistically proving that the acoustic wave has a significant effect on the
conversion in this reactor. The signal to noise t-ratio (t*) was found using Equation 5.4.,
E or I
t* = (5.4)
As with a standard t-test, statistical significance for each effect and interaction was found
at the desired confidence level (both 95% and 99.9% are presented) for the total degrees
of freedom, by comparing the absolute value of the signal-to-noise t-ratio to the Student t-
value. The Student-t value yields the maximum noise ratio at the specified degrees of
freedom and confidence interval. The effects and interactions with absolute values of the
signal to noise t-ratios larger than the Student t-value are considered to be significant at
the confidence interval stated. The results of this analysis are shown in Table 5.4.
Table 5.4: Signal-to-noise t-ratios and statistical significance.
t* Significant at Significant at 99.9%
95% Confidence? Confidence?
Student t-value at 95% Confidence 2.120
Student t-value at 99.9% Confidence 4.015
Effect of Flow Rate (E,) -57.60 Yes Yes
Effect of Sound Pressure (E2) 18.64 Yes Yes
Effect of Catalyst Bed Length (E3) 76.80 Yes Yes
Interaction of Flow Rate and Sound
7.42 Yes Yes
Interaction of Flow Rate and Catalyst 37.02 Yes Yes
37.02 Yes Yes
Bed Length (1,3)
Interaction of Sound Pressure and -11
-10.41 Yes Yes
Catalyst Bed Length (123)
Interaction of Flow Rate, Sound 78
-1.78 No No
Pressure and Catalyst Bed Length (I,23)
These data also show that all of the variables investigated including sound
pressure has a statistically significant effect on the conversion for this reactor at a
confidence level of 99.9%. The fact that the interaction of flow rate. sound pressure and
catalyst bed length can be neglected allows for a slight simplification of the con\ crsion
equation. The statistically valid conversion model for the inputs of flow rate, sound
pressure amplitude and catalyst bed length is found in Equation 5.5. This equation is
used by finding where the actual flow rate, sound pressure and bed length variables fit
into the high (+1) and low (-1) settings for each parameter. These values can be found by
C(%) = 87.06 7.22X, + 2.34X, + 9.63X3 (
+ 0.93XX,2 + 4.64X,X, 1.31XX,
utilizing Equations 5.6 5.8 for specific premix flow rates (g'min), sound pressure
amplitudes (Pa), and catalyst bed lengths (m), respectively. These values are then utilized
in Equation 5.5 as X,, X2, and X3, where C is the conversion of methanol in percent.
SFlow(g / min)- 7.13(g / min)
2.35(g / min)
Pressure (Pa)- 2680(Pa)
X2 = 2680(Pa) (5
Bed Length (m) 0.424(m)
While sound pressure (X2) does not influence the conversion as much as the flow rates
(Xi) or catalyst bed lengths (X3) studied, the equation does show how much an effect
sound pressure can have in relation to the other variables. In order to demonstrate the
some of the limitations of the model, midpoint curvature data were obtained and the
effect of frequency was observed.
The model presented in Equation 5.5 was further examined for curvature near the
center points of the variables investigated.
Flow rate curvature in the model was investigated at two sound pressure levels
keeping catalyst bed length constant as shown in Figure 5.1. For these investigations,
catalyst bed length was set at 0.568 m and the high level of sound pressure was
approximately 5000 Pa or 165 dB. Four intermediate points were taken between the high
flow setting (+1) of 9.5 g/min and the low flow setting (-1) of 4.8 g'min. Comparing the
conversion at various flow rates with the conversion predicted by the model, for cases
99 --Model Prediction No
9 97 a Actual No Acoustics
S95 Model Prediction
E 94 With Acoustics
93 e Actual with
4 6 8 10
Premix Flow Rate (glmin)
Figure 5.1: Predicted and actual conversions for various flow rates with and without
with and without the 165dB acoustic field show some visual curvature near the
midpoints. Although one can see this slight trend upward in conversion at the midpoints,
no statistically significant curvature is found. This is because differences between the
predicted and actual conversion data (also known as the residuals) are in the acceptable
noise range ( 1.22 %) for the individual data points collected.
Curvature for the sound pressure was investigated at a high level of premix flow
rate (9.5 g/min) and a high level of catalyst bed length (0.576 m) at a frequency of 650
Hz. The actual conversion at various sound pressure amplitudes, together with the
conversion predicted by the model, are shown in Figure 5.2. Although some of the actual
Sales are slightly lower than expected, these values do not show significant curvature as
the residuals are less than 1% for the cases investigated.
c 96 _
.9 Model prediction
4 w Actual
0 2000 4000 6000 8000
Maximum Pressure Amplitude (Pa)
Figure 5.2: Predicted and actual conversion for maximum pressure amplitude with a
0.568 m catalyst bed length and 9.5 g/min premix flow rate.
Catalyst Bed Length
Curvature in the catalyst bed length variable was investigated at the high level of
premix flow rate (9.48 g/min) with a 165 dB (re20Pa) acoustic field as shown in Figure
5.3. This figure shows the predicted conversion of the model and actual conversion data
points gathered at different catalyst bed lengths. Note that the conversion predicted by the
model has been extrapolated beyond the low level of catalyst bed length (0.278 m) in the
figure. Data shown has significant curvature as it deviates from the model predictions
away from the high (0.568 m) and low test levels (0.278 m). The linear model prediction
is in error for catalyst bed length and it appears that a logarithmic fit, shown as the
regression line on the figure is more suitable and realistic, especially considering the
actual entrance conditions. Further data on conversion with respect to catalyst bed length
is presented in Chapter 6.
S60 model prediction
S50 au actual
40 -Log. (actual)
S30 0 C% = 38.286Ln(length) + 119.61Log(actual)
0 20 R2 = 0.9612
0 0.2 0.4 0.6
Bed Length (m)
Figure 5.3: Predicted and actual conversions for bed length at a flow rate of 9.5 g/min
premix and a 165 dB acoustic field.
The Effect of Frequency
As previously described in the theoretical approach, for an ideal setting, frequency
has been linked to boundary layer thickness for the laminar regime (Equation 2.7) and the
kinetic Reynolds number (Equation 2.8). Because of this, the acoustically enhanced
methanol-steam reforming process was also investigated to find any correlation bet\\ een
frequency and conversion of fuel. Frequency effects were measured at a high leI el of
premix flow rate (9.5 g/min) and a high level catalyst bed length (0.568 m). With the
acoustic driver power held constant at 9.3 Watts, the average observed sound pressure
level dropped to 1434 Pa when the reference frequency was increased to 1980 Hz. High
acoustic pressures (5000 Pa Max. Amplitudes) were not possible at this frequency with
the same driver power due to the acoustic resonant properties as previously presented.
Conversion. as predicted by the model created from 650 Hz data for this sound pressure
amplitude %was 93.2%. This value matches the experimental conversion found at 1980 Hz
exactly. However, care should be taken in the interpretation of this data point because
this value cannot necessarily be considered out of the noise band of the data without
acoustics. Notwithstanding, this data suggests that frequency was not a significant
variable for the acoustic fields investigated.
Reynolds numbers were estimated for different frequencies using the properties of
ambient pressure steam at 2500C (v = 4.4018 x 10-5 m2/s) and the diameter of a catalyst
pellet (0.54 cm) as the characteristic length. The average velocity was estimated at 0.89
m/s from the flow rate, density steam at 250 C, and volume of the reactor. The Reynolds
numbers are shown in Table 5.5. Static Reynolds number (Res) for the non-oscillating
flow indicates a laminar flow regime with no acoustics. Kinetic Reynolds numbers
(Rek)were calculated for an oscillating field at the two frequencies investigated using the
properties above in Equation 2.8. The kinetic Reynolds number at 650 Hz is 2705
suggesting that the flow is in the transitional or turbulent regime at this frequency. An
increase of frequency to 1980 Hz increases the kinetic Reynolds number to 8241 which is
also in the turbulent regime. A comparison of the turbulent regime to turbulent regime
Table 5.5: Static and kinematic Reynolds numbers and methanol conversion levels for
Hz Reynolds Number Conversion %
0 110 (Re,) 92.4
650 2705 (Rek) 93.2 (predicted from model at 1434 Pa)
1980 8241 (Rek) 93.2 actual (1434 Pa)
with the two frequencies yields no difference in conversion as shown by the data.
Furthermore, increased frequency yields lower pressure for the same driver input power
thus the overall power benefit was adversely influenced by increasing acoustic frequency
in this system.
Factorial design of experiments has been used to find the effects of the parameters
of premix flow rate, sound pressure amplitude, and catalyst bed length. Sound pressure
amplitude in the study was statistically significant at a 99.9% confidence level. From the
data collected a model was formulated that predicts methanol conversion for the above
parameters. The model had negligible curvature for fuel flow rate and maximum pressure
amplitude and significant error for catalyst bed lengths. Conversion for catalyst bed
length is better modeled by a logarithmic relationship. Acoustic field frequency did not
produce a noticeable change in conversion.