Graduate Theses and Dissertations
Iowa State University Capstones, Theses and
Dissertations
2008
Quenching of particle-gas combustible mixtures
using the electric particulate suspension (EPS)
method
Hua Xu
Iowa State University
Follow this and additional works at: https://lib.dr.iastate.edu/etd
Part of the Mechanical Engineering Commons
Recommended Citation
Xu, Hua, "Quenching of particle-gas combustible mixtures using the electric particulate suspension (EPS) method" (2008). Graduate
Theses and Dissertations. 10635.
https://lib.dr.iastate.edu/etd/10635
This Dissertation is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University
Digital Repository. It has been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of Iowa State University
Digital Repository. For more information, please contact digirep@iastate.edu.
Quenching of particle-gas combustible mixtures using the electric particulate
suspension (EPS) method
by
Hua Xu
A dissertation submitted to the graduate faculty
in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
Major: Mechanical Engineering
Program of Study Committee:
Gerald M. Colver, Major Professor
Theodore J. Heindel
James C. Hill
Steven J. Hoff
Ron M. Nelson
Iowa State University
Ames, Iowa
2008
Copyright © Hua Xu, 2008. All rights reserved.
ii
TABLE OF CONTENTS
LIST OF FIGURES
iv
LIST OF TABLES
vii
ACKNOWLEGMENTS
viii
ABSTRACT
ix
CHAPTER 1. INTRODUCTION
1.1 Motivation
1.2 Objectives
1.3 Organization of Dissertation
1
1
3
4
CHAPTER 2. LITERATURE REVIEW
2.1 Electric Particulate Suspension (EPS)
2.1.1 Dynamics of an electric suspension
2.1.2 Applications of EPS technique in multiphase flows
2.1.3 Application of EPS in combustion
2.2 Metal Particle Combustion
2.2.1 Metal combustion classification
2.2.2 Metal particle combustion regimes
2.2.3 Metal combustion in oxidizer atmospheres
2.2.4 Metal combustion in microgravity
2.3 Flammability Limits and Flame Quenching
2.3.1 Gas fuel flame quenching
2.3.2 Particle-gas cloud flame quenching
2.4 Summary
5
5
5
10
14
21
22
25
28
32
37
39
45
51
CHAPTER 3. EXPERIMENTAL METHOD
3.1 Basic Experimental Setup
3.2 Particle Preparation
3.3 Light Extinction Coefficient
3.3.1 Theoretical calculation of extinction coefficient
3.3.2 Experimental determination of extinction coefficient
3.4 Experiments for Microgravity Study
3.4.1 Drop tower microgravity test
3.4.2 EPS 5-tiered ignition system and A-Frame layout
3.4.3 Drop tower control and data sequencing
3.5. Summary
53
53
57
64
64
67
68
69
70
77
82
CHAPTER 4. PARTICLE SUSPENSION CONCENTRATION
4.1 Particle Concentration Measurement in Normal Gravity
4.2 Particle Concentration Measurement in Microgravity
4.3 Particle Concentration Stratification Criteria
84
85
98
107
iii
4.4 Summary
110
CHAPTER 5. MAXIMUM SUSPENSION CONCENTRATION
5.1 Maximum Suspension Concentration by Current Measurement in Normal Gravity
5.2 Excess Electric Field Intensity
5.3 Maximum Particle Concentration Correlation
5.4 Other Observed Phenomena in EPS
5.5 Summary
112
114
123
129
136
138
CHAPTER 6. ALUMINUM PARTICLE IGNITION AND QUENCHING
6.1 Aluminum Quenching
6.1.1 Aluminum quenching at normal gravity
6.1.2 Aluminum quenching in microgravity
6.1.3 Aluminum concentration correction at microgravity
6.1.4 Quenching distance comparison
6.1.5 Microgravity combustion products
6.2 Effect of Inert Particles on Aluminum Apparent Ignition Energy
6.2.1 Aluminum apparent ignition energy without inert particles
6.2.2 Aluminum apparent ignition with inert particles (copper)
6.2.3 Apparent ignition energy correlation
6.3 Summary
139
139
140
142
150
154
156
159
160
164
167
173
CHAPTER 7. CONCLUSIONS AND FUTURE WORK
174
BIBLIOGRAPHY
177
APPENDIX: UNCERTAINTY ANALYSIS
A.1 Significant Digits
A.1.1 Electric field intensity
A.1.2 Particle number density and mass concentration
A.1.3 Quenching distance
A.1.4 Ignition energy
A.2 Regression Fit Confidence Interval
A.3 Uncertainty Propagation
195
195
195
195
196
196
197
198
iv
LIST OF FIGURES
Figure 2.1. Schematic of the particle behavior in the electric field.
Figure 2.2. Experiment layout with continuous particle feed - 1 g (Yu, 1983).
Figure 2.3. Quenching of copper-propane-air mixtures - 1 g (Yu, 1983).
Figure 2.4. Aluminum mixtures minimum spark energy – 1 g (Kim, 1986).
Figure 2.5. Quenching distance versus aluminum concentration - 1 g (Kim, 1989).
Figure 2.6. Quenching distance versus coal concentration - 1 g (Kim, 1989).
Figure 2.7. Schematic design diagram of the aerosol feeder (Shoshin and Dreizin, 2002).
Figure 2.8. Temperature profile in a flame (Williams, 1985).
Figure 3.1. Basic EPS setup.
Figure 3.2. Laser photodiode scanning system (Greene, 2004).
Figure 3.3. Laser photodiode scanner (Greene, 2004).
Figure 3.4. Size distribution and micro-photo of Al (15-20 μ m).
Figure 3.5. Size distribution and micro-photo of Al (20-25 μ m).
Figure 3.6. Size distribution and micro-photo of Al (25-30 μ m).
Figure 3.7. Size distribution and micro-photo of Cu (30-38 μ m).
Figure 3.8. Size distribution and micro-photo of Cu (44-53 μ m).
Figure 3.9. Size distribution and micro-photo of Cu (53-63 μ m).
Figure 3.10. Size distribution and micro-photo of Cu (63-74 μ m).
Figure 3.11. Size distribution and micro-photo of Gl (25-30 μ m).
Figure 3.12. Size distribution and micro-photo of Gl (53-63 μ m).
Figure 3.13. Size distribution and micro-photo of Fe (0-44 μ m).
Figure 3.14. Size distribution and micro-photo of Mg (0-44 μ m).
Figure 3.15. Size distribution and micro-photo of Ti (0-44 μ m).
Figure 3.16. Calculated light extinction coefficients curve.
Figure 3.17. Experimental light extinction coefficients.
Figure 3.18. EPS 5-tiered experiment rig.
Figure 3.19. Schematic of EPS 5-tiered system.
Figure 3.20. Dual HV circuit.
Figure 3.21. Schematic of A-Frame layout.
Figure 3.22. Actual picture of A-Frame layout.
Figure 3.23. Drop tower control flowchart.
Figure 3.24. Drop tower DAQ diagram.
Figure 4.1. Concentration profile of 0.2 gm Al (20-25 μ m) with moving laser scan - 1 g.
Figure 4.2. Concentration profile of 0.2 gm Gl (53-63 μ m) with moving laser scan - 1 g.
Figure 4.3. Concentration profile of 1.0 gm Cu (53-63 μ m) with moving laser scan - 1 g.
Figure 4.4. Concentration profile of 0.2 gm Gl (53-63 μ m) with laser array scan - 1 g.
Figure 4.5. Concentration profile of 0.1 gm Cu (53-63 μ m) with laser array scan - 1 g.
Figure 4.6. Concentration profile of 0.2 gm Cu (53-63 μ m) with laser array scan - 1 g.
7
15
15
16
18
18
19
42
54
56
56
58
58
59
59
59
60
60
60
61
61
61
62
66
68
71
72
74
76
76
79
80
86
87
88
90
91
92
v
Figure 4.7. Concentration profile of 0.3 gm Cu (53-63 μ m) with laser array scan - 1 g.
Figure 4.8. Concentration profile of 0.4 gm Cu (53-63 μ m) with laser array scan - 1 g.
Figure 4.9. Concentration profile of 0.5 gm Cu (53-63 μ m) with laser array scan - 1 g.
Figure 4.10. Maximum number density of Cu (53-63 μ m) at given E - 1 g.
Figure 4.11. Minimum E of Cu (53-63 μ m) vs. mass loading - 1 g.
Figure 4.12. Laser intensity profile of 667 mg Al (20-25 μ m) at E=292 kV/m.
Figure 4.13. Particle concentrations of 667 mg Al (20-25 μ m) in 11 mm test cell.
Figure 4.14. Particle concentrations of 500 mg Al (20-25 μ m) in 14 mm test cell.
Figure 4.15. Particle concentrations of 400 mg Al (25-30 μ m) in 20 mm test cell.
Figure 4.16. Particle concentrations of 1789 mg Gl (53-63 μ m) in 20 mm test cell.
Figure 4.17. Microgravity videos of Gl (53-63 μ m).
Figure 4.18. Microgravity videos of Copper (63-74 μ m).
Figure 5.1. Induced currents of Al (20-25 μ m) at given voltages and heights - 1 g.
Figure 5.2. Induced currents of Cu (30-38 μ m) at given voltages and heights - 1g.
Figure 5.3. Induced currents of Cu (44-53 μ m) at given voltages and heights - 1g.
Figure 5.4. Induced currents of Cu (63-74 μ m) at given voltages and heights - 1g.
Figure 5.5. Induced currents of Gl (25-30 μ m) at given voltages and heights – 1 g.
Figure 5.6 Induced currents of Gl (53-63 μ m) at given voltages and heights - 1 g.
Figure 5.7. Induced currents of Mg (0-44 μ m) at given voltages and heights - 1 g.
Figure 5.8. Induced currents of Ti (0-44 μ m) at given voltages and heights - 1 g.
Figure 5.9. Induced currents of Fe (0-44 μ m) at given voltages and heights - 1 g.
Figure 5.10. Excess Electric Field Intensity Ratio α - 1 g.
Figure 5.11. Maximum Particle Concentration with Excess Electric Field Intensity.
Figure 5.12. Max particle concentration at given E for Al (20-25 μ m) - 1 g.
Figure 5.13. Max particle concentration at given E for Cu (30-38 μ m) - 1 g.
Figure 5.14. Max particle concentration at given E for Cu (44-53 μ m) - 1 g.
Figure 5.15. Max particle concentration at given E for Cu (63-74 μ m) - 1 g.
Figure 5.16. Max particle concentration at given E for Gl (25-30 μ m) - 1 g.
Figure 5.17. Max particle concentration at given E for Gl (53-63 μ m) - 1 g.
Figure 5.18. Max particle concentration at given E for Mg (0-44 μ m) - 1 g.
Figure 5.19. Max particle concentration at given E for Ti (0-44 μ m) - 1 g.
Figure 5.20. Max particle concentration at given E for Fe (0-44 μ m) - 1 g.
Figure 5.21. Summary of correlations with excess electric field intensity.
Figure 5.22. Interesting phenomena of glass powder after suspension - 1 g.
Figure 6.1. Quenching curve for Al (d32=19.47 μ m) - 1 g.
Figure 6.2. Quenching curve for Al (d32=19.47 μ m) - 1 g and 0 g.
Figure 6.3. Ignition development in drop tower for Al (d32=19.47 μ m).
Figure 6.4. Quenching curves Al (d32=23.72 μ m) - 1 g and 0 g.
Figure 6.5. Quenching curves for Al (d32=31.87 μ m) - 1 g and 0 g.
92
93
93
95
97
99
100
100
102
103
104
106
115
116
117
118
119
120
121
122
123
126
128
130
131
131
132
132
133
133
134
134
136
137
141
142
144
145
146
vi
Figure 6.6. Ignition development in drop tower for Al (d32=23.72 μ m).
Figure 6.7. Ignition development in drop tower for Al (d32=31.87 μ m).
Figure 6.8. Corrected microgravity quenching data for Al (d32=23.72 μ m).
Figure 6.9. Corrected microgravity quenching data for Al (d32=19.47 μ m).
Figure 6.10. Corrected microgravity quenching data for Al (d32=31.87 μ m).
Figure 6.11. Comparison of experimental data with predicted curve - 1 g.
Figure 6.12. New found formation of stringer/web - 0 g.
Figure 6.13. Microphotos of aluminum combustion product.
Figure 6.14. Apparent ignition energy of aluminum (15-20 μ m) - 1 g
Figure 6.15. Apparent ignition energy of aluminum (20-25 μ m) - 1 g.
Figure 6.16. Apparent ignition energy of aluminum (25-30 μ m) - 1 g.
Figure 6.17 Apparent ignition energy of Al (15-20 μ m) with Cu (30-38 μ m) - 1 g.
Figure 6.18. Apparent ignition energy of Al (15-20 μ m) with Cu (44-53 μ m) - 1 g.
Figure 6.19. Apparent ignition energy of Al (15-20 μ m) with Cu (63-74 μ m) - 1 g.
Figure 6.20. Log ( Δ Ei) Vs. Log (N) - 1 g.
Figure 6.21. Log ( Δ Ei) Vs. Log (D) - 1 g.
Figure 6.22. Average ( Δ Ei) Vs. ND2 – 1 g.
Figure A.1. Total uncertainty for Glass Bead using moving laser scan – 1 g.
Figure A.2. Total uncertainty of Glass Bead using laser array scan – 1 g.
147
149
152
153
153
156
158
159
161
162
163
165
166
167
168
170
172
201
201
vii
LIST OF TABLES
Table 2.1. Classification of metal particle combustion (Williams, 1997)
Table 2.2. Comparison of quenching distance and flammability
Table 3.1. Particle size range
Table 3.2. Particle diameter table-part1
Table 3.3. Particle diameter table-part2
Table 4.1. Comparison of N ave and N cal
Table 5.1. Comparison of minimum particle lift-off electric field intensity
Table 5.2. Linear Regression Equations
Table 6.1. Max mass concentration for aluminum (20-25 μ m)
24
52
58
63
63
96
127
135
151
viii
ACKNOWLEGMENTS
First, I would like to thank Dr. Gerald M. Colver, my advisor and mentor, for
continuous guidance and support during this project. I would also like to thank my POS
committee members, Drs. Theodore J. Heindel, Ron M. Nelson, James C. Hill, and Steven J.
Hoff, for their generous contribution of time and valuable comments. Financial support from
the Mechanical Engineering Department is greatly appreciated. I am also thankful to the
professors I was working with during my teaching, Drs. Theodore J. Heindel, Sriram.
Sundararajan, and Adin Mann.
This research was supported by the NASA grant NCC3-846 in response to NASA
Research Announcement Mircrogravity Combustion/Science: Research and Flight
Experiment Opportunities, NRA-99-HED-04. This author gives special thanks to our NASA
technical contacts and administrators of the project Drs. Zeng-Guang Yuan and Peter
Sunderland, who provided guidance and expertise on this project.
This author gives a special thanks to the research group members: Nate Greene,
David Shoemaker, and Ryan Kroll. I also thank Jim Dautremont for his help on the
instruments.
Last, but not the least, I would like to thank my parents for their support and love
during my lifetime. This thesis is dedicated to them.
ix
ABSTRACT
A new experiment referred to as EPS (electric particulate suspension) designed for
quenching distance measurement of combustible powder suspensions has been taken from
concept to working prototype. The method is validated in both normal gravity and
microgravity.
Particle suspension concentration profiles were determined by either a moving single
laser scan (in 1 g) or by an expanded laser sheet scan (in 1 and 0 g). Stratification of
suspensions (asymmetry of concentration profiles due to gravity) was found in 1 g.
Microgravity experiments carried out in the NASA Glenn Research Center’s (NGRC) 2.2second drop tower showed higher powder concentrations in microgravity compared to
suspensions formed in normal gravity. Powder concentration profiles were confirmed to be
symmetric at 0 g and a near-constant value under certain conditions. Abnormal suspension
instability occurred in microgravity, observed as spatially periodic “cellular” powder flow
structures at high electric field intensities.
A new external current method for determining the maximum suspension particle was
introduced using EPS. A variety of particles showed a similar two-stage development–an
initial increase stage of current followed by a constant current stage. The rising stage lends
itself to a linear regression analysis. A new concept of excess electric field intensity was
introduced to correlate the normal gravity and microgravity. Suspension concentrations in
both normal gravity and microgravity were successfully correlated using this concept.
The range of ignition tests with EPS was significantly broadened to richer powder
mixtures in microgravity. The nominal aluminum concentrations were corrected to the actual
x
suspension aluminum concentrations using the excess field intensity concept. Some
interesting newfound web/stringer formations, comprised of aluminum particles, were
observed following powder ignition during drop tower experiments. The quenching effect of
inert copper particles on aluminum/air mixtures was also investigated in normal gravity. It
was confirmed that more ignition energy was required in the presence of inert particles. The
average additional ignition energy difference was successfully correlated with ND2 (adopted
from previous studies) using an exponential equation.
1
CHAPTER 1. INTRODUCTION
This chapter introduces the motivation and objectives of the current research. It
covers the fundamental subjects utilized in the research and the organization of the
dissertation.
1.1 Motivation
Many solid particles utilized in industrial processes are combustible and will burn in
air under certain circumstances of fuel-air ratio. Examples of combustible particles include
coal, grain, flour, sugar, starch, plastics, metals, pharmaceuticals, wood dust, textile fluff, and
other dusts generated when solid surfaces are polished, cleaned, or cut. Generally, dust
clouds comprised of fine particles can ignite rapidly, presenting an increasing explosion
hazard (Palmer, 1973; Field, 1982; Nagy and Verakis, 1983). In particular, the study of flame
propagation and quenching of both dusts and pure gases are beneficial for the following areas:
optimum design of furnaces and coal-fire power generators, analysis of combustion
phenomena in internal combustion engines, development of new fuel efficient engines
satisfying new emission standards, control of combustion for propulsion, and fire safety and
reduction of explosion hazards in industrial processes.
The flame-quenching distance of dust clouds (this study) is not only an important
parameter for understanding flame propagation, but is itself a fundamental parameter applied
to laminar flame theory, as it closely relates to both the characteristic flame’s thickness and
structure. Reliable data on flame quenching distance (and flammability limits) also serve as a
basis for theoretical flame modeling of flame propagation in dust suspensions.
2
From the particle point of view, the quenching distance is important in understanding
the flame-holding phenomenon, the design of flame traps and the assessment of explosion
hazards in the flow of combustible mixtures in industrial processed. And it is also relevant to
the general evaluation of explosion hazards.
Another important and recent application of flame propagation–quenching distance
and flammability testing of solid particles–is in microgravity environments. The behavior of
flame propagation in microgravity is of importance for fire safety in space environments
including the International Space Station (ISS). The National Aeronautics and Space
Administration (NASA) is also interested in the properties of flame propagation in
microgravity environments. Previous to the 1980s, metal rod burn data were collected at
NASA White Sands Test Facility (NWSTF). Data from 1-g tests were extrapolated to predict
how metals would behave in microgravity (0-g). During the 1980s, NWSTF conducted the
metallic materials combustion tests in the NASA Glenn Research Center (NGRC) drop tower
and later aboard the NASA KC-135 aircraft for 0-g combustion data. NWSTF continues to
build on its 0-g burn database for metals and metal alloys in solid rods.
Achieving a uniform dust concentration has been identified as a major technical
obstacle to be resolved in dust flame studies. Studies using conventional pneumatic
dispersion methods are problematic in achieving cloud uniformity especially in attaining rich
mixtures. These dispersion methods usually require large samples of powders and can only
reach limited dust concentrations (Ballal, 1980, 1983a, 1983b; Goroshin, 1996; Jarosinski et
al., 1986).
A new technique referred to as EPS (electric particulate suspension) was developed at
Iowa State University to overcome the difficulty of achieving uniform dust concentration
3
(Kim, 1989; Colver et al., 2004). The technique supports NASA’s space science effort in
microgravity and fire safety. By using the EPS method, one can hopefully advance the
current limited theory on the behavior of powder suspension formation and burning in
microgravity under the influence of electric fields.
1.2 Objectives
A new experimental EPS technique designed for the study of quenching distance
measurement and combustible powder suspension concentrations is taken from concept to
working prototype, and validated under both normal gravity and microgravity conditions.
The experiment focuses on development of the particle suspension and the measurement of
quenching distance and flame propagation limits. This study includes the following specific
objectives:
•
To validate the EPS method for quenching and combustion testing in
microgravity by evaluating performance and extended ranges of testing compared
to normal gravity. The particles studied include aluminum, copper, and glass
beads. The quenching distances of various sizes of aluminum particles in air are
measured in microgravity and compared with results found in normal gravity.
•
To determine the relationship between particle suspension concentrations and
electric fields for maximum suspension concentration at a specified electric field
strength. The particles tested include aluminum, iron, magnesium, titanium, nickel,
glass beads, and copper. Correlations of particle concentration and electric field
are obtained.
4
•
To evaluate the effect of inert particles (copper) on aluminum ignition using the
EPS method in a normal gravity environment. Mixtures of combustible particles
(aluminum) and inert particles (copper) in air are ignited in a closed system test
cell. The ignition energies are compared to results observed for combustible
particles in air.
1.3 Organization of Dissertation
The motivation, objectives and dissertation organization are introduced in Chapter 1.
A literature review is presented in Chapter 2 along with studies of electric particulate
suspensions (EPS) as well as studies of combustible mixtures of solid particle/air and solid
particle/CO2. Chapter 3 outlines the experimental program for the study. Details are included
that describe the experimental facility, test conditions, measurement system, experimental
procedure, data reduction and experiment uncertainty. Chapter 4 presents the experimental
results for the measurement of particle suspension concentration under normal gravity and
microgravity environments using the laser extenuation method. Chapter 5 presents the
experimental results of maximum particle concentration using an external current method
along with correlations. The initial part of Chapter 6 presents the results of aluminum
quenching distance measurements in both normal gravity and microgravity environments.
Chapter 6 also summarizes the effect of inert copper particles on the ignition of combustible
aluminum particles with correlations. Chapter 7 summarizes the conclusions of this study.
Recommendations and for further study are also included.
5
CHAPTER 2. LITERATURE REVIEW
In this chapter, studies related to this research project are surveyed and summarized.
The first section gives a brief introduction to the electric particulate suspension (EPS)
technique, related studies, and results. The second section provides a summary of quenching
distance and flame propagation measurements from the literature for combustible dust
mixtures. The third section focuses on burning mixtures for various kinds of combustible
particles and oxidizers.
2.1 Electric Particulate Suspension (EPS)
The EPS utilizes a parallel plate capacitor with particles suspended and driven
between the plates by electrostatic forces. A DC voltage is applied across the plates creating
an electric field (E). The particles act as small capacitors, while in contact with an electrode
being charged, thus taking on the same sign and potential. The like charges of the particles
and electrodes as well as the electric field drive the particles oscillate between the positively
and negatively charged electrodes from the combined effects of the applied electric field and
charges on surrounding particles resulting in a particle cloud. This particle cloud attains a
steady state mode of operation as long as the DC electric field remains applied.
2.1.1 Dynamics of an electric suspension
The charging process occurs as a result of the “capacitance” effect of the particle,
while in electrical contact with the wall. During the time of charging, the particle becomes
part of a large capacitor, the wall, and therefore charges to the same sign and potential. If the
externally applied electric field intensity, E, remains unchanged in sign, electrostatic force, Fe,
6
will tend to drive particles away from the wall. Image force, Fi, will attract particles towards
the wall. A particle confined between two parallel plates and possessing sufficiently large
charge will continue in motion indefinitely, once set in motion, oscillating between the
parallel plates and sustained by the DC electric field. Thus, a unique charge exists on a
particle for a specific electric field intensity.
Body forces such as gravity, inelastic collisions with walls and viscous air drag serve
to limit the maximum possible velocity of the particles and the maximum particle suspension
concentration. The natural independent variable controlling the motion is the externally
applied electric field.
Any particle-wall-gas properties, which affect the dynamic discharging process at the
walls such as surface conductivity, dielectric constant, relative humidity, and geometry will
limit the maximum charge transfer to the particle. The schematic of single particle behavior
in the electric field is shown in Figure 2.1.
Colver (1976) conducted an experiment on the dynamic charging of metallic particles
against a conducting wall by a DC electric field. Metallic particles were electrically charged,
while in dynamic and stationary contact with either wall of a charged parallel plate capacitor.
The particle charge distribution tended to remove particles away from the wall. The forces
from the interaction of the induced charge and the externally applied electric field tended to
move the particles away from the plate, overcoming the image attraction. The dynamic and
stationary charged particle motions were determined theoretically and experimentally by
considering the particles as capacitors in themselves. The charge of a spherical conducting
aluminum particle was determined from the following equation,
7
Figure 2.1. Schematic of the particle behavior in the electric field.
Q = 4πεa 2 EK
(2.1)
The K value was confirmed to be equal to 1.64 theoretically and experimentally for
conducting particle possessing equilibrium charge. Therefore, the theoretical equilibrium
force required to lift a single conducting sphere from a plane in a uniform electric field can
be determined by
Fe = πε 0 D 2 E 2 (1.37)
(2.2)
Inelastic particle-wall collisions and conservative body forces, such as gravity,
determine the critical lower limit electric field intensity for sustained particle motion. It
follows that in the absence of contact effects, the applied electric field is entirely responsible
for the resulting lifting force on a particle, since it alone determines the electric field intensity,
8
E, and also controls the amount of charge, Q, accumulated on the particle. The minimum
electric field intensity is given as
⎡⎛ 1 − e 2
⎤
⎞ 6
2
⎟
+
= ⎢⎜⎜
mg
F
D
ε
⎥
0
D
2
⎟π 3
⎠
⎣⎝ 1 + e
⎦
FLL
1
2
(2.3)
where e is the coefficient of restitution and F D is the drag force per particle, given as
FD = 3πμdV [1 + 3DρV / 16μ ]
1
2
(2.4)
where the above equation is valid for particle Reynolds number below 100. d , μ , and ρ are
the particle diameter, viscosity, and density, respectively.
Continuous cyclic motion resulted as a particle impacts, discharges, and recharges
with each electrode. Colver (1976) found that particles in clouds in the presence of gravity
and standard atmospheric air also demonstrate a continuous motion between the parallel
plates. The motion of a particle oscillating between parallel electrodes, responding to the
influence of an electric field intensity, E, can be described by
d 2xp
dt
=
2
QE
±g
m
(2.5)
From the above equation, the average velocity of the particle oscillating between the
parallel plates can be described by
V = [ (1 + e)]
1
8
1
2
1
1
⎧⎪⎡ QE 1
g ⎤ 2 ⎫⎪
g ⎤ 2 ⎡ QE 1
+
+⎢
−
⎬
⎨⎢
2
2
1 + e 2 ⎥⎦
1 + e 2 ⎥⎦ ⎪
⎣ m 1− e
⎪⎩⎣ m 1 − e
⎭
(2.6)
For multiple particles suspension, the charge transfers by collisions (with solid wall
and other particles) are very important. The transferred charge also indicates the particle
number density and the local mass flux. For the mono-dispersed dilute suspension of gas–
9
solid flows, Cheng and Soo (1970) presented a simple model for the charge transfer in a
single scattering collision between two elastic particles, by measuring the ball probe current
produced by electrostatic charge separation from impaction. The number of simultaneous
impacts by probe on a ball probe is given by
Nc =
1
λB n
1
(2.7)
3
where n is the number density of particles, λ B is the mean free path of particle to probe
collisions which can be expressed as
λB =
4
nπ (d b + d p ) 2
(2.8)
and db is the ball probe diameter.
The effect of multiple scattering, sliding, and charge saturation on the ball probe
current can be taken into account in a theory via the modification of the electric potential
difference between the ball and each single colliding particle (Zhu and Soo, 1992). The
current for dense suspensions can be described as
⎛
Jp
3
ib = AU p 5 J pexp⎜ − C c 2
⎜
Up
⎝
⎞
⎟
⎟
⎠
(2.9)
where A and Cc are determined experimentally, J p is the mass flux of particles, and U p is
the particle velocity.
10
2.1.2 Applications of EPS technique in multiphase flows
Cotroneo and Colver (1978) applied the EPS technique to pneumatic transportation of
copper particles. It their study, 85 μ m copper spheres were suspended electrostatically
against gravitational forces and were pneumatically transported horizontally in a parallel
plate flow system of 1 by 5 cm cross–section. With the assistance of electrical forces, the
particle motion in pneumatic transport was extended into low Reynolds numbers (both flow
system and particle) regimes. Since particle and duct Reynolds numbers are small ( Re P <13:
Re C <1870), vortex shedding was minimized, resulting laminar flow around the particle and
through the duct. An equation for vertical current flux J was derived,
[
J = fnQV e − nσlα + γ (1 − e − nσlα )
]
(2.10)
where the first term in the bracket accounts for particles moving a distance l without collision
and the second term accounts for the remaining fraction undergoing collisions. The
parameters, f , α , and γ , are suggested to account for particle history effects, randomization
as a result of collisions, irregular bounces, and particle rotation. Sarhan (1989) continued the
application of EPS to the pneumatic transport of solids. An experimental investigation was
carried out using an electric field to assist particle lift in a rectangular duct. Copper and
spherical glass particles were pneumatically conveyed at small particle Reynolds number.
With electric field assist, conveying particles flow at a rate below the saltation velocity in
laminar flow. The average particle drift velocity along the duct was measured by LDA,
together with the duct pressure drop. By comparing the pressure drop with the assisting
electric field and without electric field, it was observed that the pressure drop increases with
increasing electric field strength. Thus the conveying efficiency also increased.
11
Colver and Howell (1980) measured particle number densities experimentally by
three independent methods–by electrical current density, laser beam attenuation, and direct
count. It was shown that in an electric suspension the diffusion process is significant and
furthermore can be isolated experimentally in the absence of any fluid dynamic driving force.
Also, they traced the origin of electric suspension diffusion to one or more of the following
processes: (1) particle concentration gradients in the electric field intensity along the plate as
a consequence of spatial variations in net charge concentration, and (2) random particle
motions, due to particle-particle collisions or particle-wall collisions.
Colver (1980) reviewed naturally occurring and electrically-induced particle cohesive
forces regarding their inhibiting effects on the formation of an electrical suspension. The
naturally occurring cohesive and adhesive particle interactions are generally important for
particles below 200 μm . The naturally occurring particle forces include Van der Waals forces,
capillary forces, and electrostatic contact forces (contact potential or tribolectric effects).
Induced particle interactions will result from applied electric fields (ac or dc) acting on the
particles within a fixed or fluidized bed, which at least include dipole forces and electrostatic
forces associated with charge separation as a result of current flow in a resistive bed. Colver
points out that it is the naturally occurring forces and not the electric filed induced cohesive
forces that are basically responsible for the failure of an electric suspension to form. In
general, the EPS technique can be applied to any powder exhibiting a finite surface or
volume conductivity, including normal insulating powders. Bulk powder resistivities < 109
Ω m are recommended. For cohesive powders such as coal, acoustic vibration of the powder
may be required to form a suspension.
12
Colver and Ehlinger (1988) measured particle speed distribution of an electric
particulate suspension in the direction of the applied electric field, using spherical copper
particles of mean diameters 48.5, 69, and 115 μ m by leaking particles from a small hole
located on top of the suspension test section. Different ranges of particle speeds were
determined by capturing the particles on epoxy-coated glass slides placed at various heights
above the sampling hole. Curve fits of the data suggest that a Maxwellian speed distribution
applies to the particle motion in the direction of the applied electric field. Eimers (2002)
applied the EDA technique to measure the average particle slip velocity of 96 μ m copper
particle suspension. The particle velocity distribution is also similar to a Maxwellian-type
speed distribution.
The EPS technique was also applied to electrostatic precipitation (Liu and Colver,
1991). The target copper spheres were first electrostatically charged, using the EPS method
and propelled out of the chamber using a dc-applied electric field. The charged particles were
then directed upward into an adjacent duct and scrubbed a cross-flowing dust-laden gas as a
result of impaction and electrostatic attraction. The copper spheres were larger by a factor of
10 to 100, compared to the captured dust. The total collection of dust particles (per target
particle) and the target collection efficiency was found to increase with the charging electric
field intensity, the target particle diameter, and the duct air velocity. The results obtained in
the study showed target efficiency in excess of unity (up to 1.8) is possible with the improved
performance attributed to the effect of electrostatic charge induced on the target copper
particles. It is noteworthy that the values obtained in the experiment exceed, in every case.
Theoretically-predicted values.
13
In a related study to EPS, elutriation control using ac and dc electric fields was
carried out in an air-fluidized bed of sand in the bubbling regime (Wang and Colver, 2003).
Reductions of sand fines concentration up to 96% in the bed freeboard were measured using
a real time laser-optical technique together with a solids flux model. A unique submerged
electrode fluidized bed was used to test the bed retention mechanisms of fines and to
measure elutriation constants and electrostatic charge of the fines. Real time particle
sampling using a Faraday cage in the bed freeboard confirmed that net charge (per mass) on
particles leaving the bed first increases and then decreases over time. Together with electric
field bubble control, this charge is thought to contribute to improved particle retention in the
bed. The experimental variables studied for elutriation included electric field intensity and
frequency, superficial velocity, bed temperature, and distance of the (submerged) electrode
below the surface of the bed. Some important conclusions from this research were:
•
Both ac and dc fields give similar percentage reductions in concentration of
•
fines in the freeboard and similar values of elutriation constants ki.
Reduction of fines concentration up to 96% is produced with the application
of ac electric fields.
•
Elutriation constants for sand are independent of temperature up to 500 C.
•
Elutriation constants decrease with increasing electric field intensity.
•
Elutriation constants can be evaluated using a solids flux model together with
a real time laser-optical measurement in the freeboard (i.e. the bed does not
have to be shut down).
•
Mechanisms associated with a submerged high-voltage electrode are
responsible for controlling elutriation.
14
•
Peak values in net charge/mass ratio are observed during elutriation for both
natural and electric field.
•
An ac field moves the peak charge/mass ratio earlier after startup of elutriation
compared to the peak charge/mass ratio for natural charging.
•
Both ac and dc electric fields can induce electrostatic charge that overrides the
sign from natural charge and even reverse the sign.
2.1.3 Application of EPS in combustion
The EPS technique was applied to study the spark energy of combustible mixtures
(mixtures of combustible particles with air and mixtures of inert particles with combustible
gases) by Yu (1983). The EPS system, in conjunction with a high-speed moving needle, was
used to study the electrical breakdown and ignition energy of dust mixtures. The
experimental setup is shown as Figure 2.2.
The high-speed “injected” moving needle is visible at the upper plate and the Pyrex
glass and copper electrodes give reflections from the spark. Streaks of particles are apparent
following the ignition from the spark between the high-voltage electrodes. The system was
used to investigate spark ignition energy and quenching of propane-air mixtures in the
presence of copper particles. It was found the parameter ND 2 (N, particle number density,
and D, particle diameter) is important for gas ignition. The quenching effect of particles is
shown in Figure 2.3. The energy required to ignite mixtures becomes bigger for either higher
values of particle concentration or particle diameter.
Minimum ignition energy curves were obtained in various admixtures of oxygen,
nitrogen, and carbon dioxide at ambient conditions of temperature and pressure (Kim, 1986).
15
Figure 2.2. Experiment layout with continuous particle feed - 1 g (Yu, 1983).
Figure 2.3. Quenching of copper-propane-air mixtures - 1 g (Yu, 1983).
16
Also, flammability curves for aluminum dust mixed with oxygen and a diluent gas of either
nitrogen or carbon dioxide were developed for lean mixtures. These studies confirm that
carbon dioxide is a more effective diluent than nitrogen in suppressing the flammability of
aluminum powder. Typical data of minimum ignition energy for an EPS mixture of
aluminum (25-30 μ m), oxygen, and carbon dioxide at ambient conditions are shown in
Figure 2.4.
Figure 2.4. Aluminum mixtures minimum spark energy – 1 g (Kim, 1986).
17
The oxygen/total gas fraction (moles/moles) was held at φ =0.21, while the parallel
plate electrode spacing was 2.1 cm at a voltage of 20 kV. It was not possible to test the rich
branch of the curve due to the concentration limitations of the EPS method, although a welldefined curve minimum could usually be found, called the lowest minimum ignition energy.
An experiment correlation was found to determine the minimum ignition energy at specified
oxygen-diluent concentrations.
E min = A + B(log10 C al ) + F (log10 C al ) 2
(2.11)
where constants A, B, and F were determined by experiments at different fuel/air ratios.
The EPS method is well suited for small-volume testing of powders and,
consequently, can be used to measure quenching distance (Kim, 1989). For parallel walls, the
quenching distance is the smallest characteristic separation distance that will permit a flame
to propagate against a zero velocity flow. As such the quenching distance is a safety criterion
that can be used to arrest flame propagation. The EPS system used in this study was equipped
with variable capacitances to achieve variable spark energies. This system also used a
permanent needle electrode to replace the moving needle to achieve better spark. To confirm
the experiment, quenching distances of methane-air mixtures at various fuel/air ratios was
compared to the well-established quenching distance curves (Lewis and Elbe, 1961). It was
found that the quenching data obtained with EPS system are consistent with previous data.
Furthermore, spherical aluminum particles and coal particles with different diameters were
tested for quenching distances. Because of the cohesiveness of coal particles, an excited table
was used to help to break down particles and assist with the suspension. The results are
presented in Figures 2.5 and 2.6 for aluminum and coal particles, respectively. The
quenching tests indicate that the quenching distance and the lean flammability limit increase
18
with particle size. The quenching distance of coal was observed to decrease with increasing
volatile content.
Figure 2.5. Quenching distance versus aluminum concentration - 1 g (Kim, 1989).
Figure 2.6. Quenching distance versus coal concentration - 1 g (Kim, 1989).
19
The EPS technique was adopted to produce metal aerosol (Shoshin and Dreizin,
2002). The modified system is shown in Figure 2.7. This improved design eliminated some
reported problems, such as charged particle deposition on side walls and particle
agglomeration. It was shown aerosol number density was controlled by an electric field,
independent of the gas flow rate. A produced stable small-scale, laminar, premixed, lifted
aluminum-air flame was successfully achieved. Individual particle flame zones were
visualized in this stable aluminum-air aerosol flame for the first time.
Figure 2.7. Schematic design diagram of the aerosol feeder (Shoshin and Dreizin, 2002).
20
A theoretical model to calculate the aerosol maximum concentration was derived
based on following assumptions:
1. A one-dimensional problem is considered.
2. The local electric field is averaged over a small volume that still contains
many particles.
3. All particles have the same radius, r.
4. The particle velocity relaxation time is small, compared with the time of
particle travel between electrodes. In other words, the particle acceleration at
each instant between the collisions is negligible and the particle velocity is
equal to the local settling velocity.
5. Particle charging during collision occurs quickly, τcharging> τcollision.
6. Particles do not collide with each other.
7. Gravity force is much less than the initial electrostatic force QoEo>>mg,
where Eo is the electric field in the absence of powder, Qo is the charge
received by a particle upon collision with an electrode at electrical field Eo, m
is the particle mass, and g is the gravitational constant.
8. Adhesion forces between particles placed at the bottom electrode and between
the particles and electrodes are neglected.
With these assumptions, the aerosol mass concentration can be found as
16 εε 0V02
B=
27 gh03
⎛ h
⎜⎜
⎝ h0
⎞
⎟⎟
⎠
−
1
3
(2.12)
where B is the aerosol mass concentration, V0 is the applied voltage, and h0 is the separation
between two electrodes.
21
Recently, Kroll (2006) used the EPS technique to study the burning velocity of
combustible dust mixtures (aluminum dust cloud, natural gas/copper, propane/air, and natural
gas/air). The experiment was carried out by monitoring the current flowing through the flame
front at different locations. The bottom electrode was divided into small sections,
individually monitored to determine current flow, and then the burning velocities were
calculated by the section’s distances and time intervals. Various concentration ratios of
propane and natural gas were measured and compared to theoretical laminar flame values.
The measured velocities were larger than the predicted value, possibly because of turbulent
effects. Burning velocities of various concentrations of natural gas with copper spheres were
measured. It was found that the burning velocities increased throughout the range of natural
gas concentrations at low concentrations of copper spheres. For the same natural gas/air ratio,
the burning velocities increased, following the increased particle concentrations compared to
the burning velocities in the absence of inert copper particles. Clouds of aluminum particles
in natural gas and air mixtures were also combusted and the burning velocities were
measures. Difficulties with data acquisition limited the information collected; however,
initial results suggested a burning velocity greater than found in comparable mixtures of
natural gas/air at the same ratio.
2.2 Metal Particle Combustion
Metal combustion is a challenging scientific research topic that also has important
practical applications. Because of the high energy densities of many metals, the combustion
of metals is commonly used in solid-propellant rocket motors. High temperature metal
combustion is important to self-propagation high-temperature synthesis (SHS) of materials
22
and to the production of nano-sized metal oxide and nitride particles, as well as to spectacular
displays of pyrotechnics. Metal fires can be extremely dangerous and have often led to
unexpected explosions. The concerns are particularly important to oxygen handling systems,
oxygen separation plants and nuclear reactors. In propulsion systems, the combustion of
metals typically occurs via small diameter particles. Bulk metal combustion and flame spread,
as well as the combustion of particle clouds, are common to metal fires.
2.2.1 Metal combustion classification
For oxygen-containing environments, in which the final product is a refractory metal
oxide, studies (Price, 1984; King, 1993; Glassman, 1996) recognized: (1) the importance of
the volatility of the metal relative to the volatility of the metal oxides and (2) the relationship
between the energy required to gasify the metal or metal oxide and the overall energy
available from the oxidation reaction. Metal combustion in oxygen is typically classified by
the way the metal is first oxidized to its smallest sub-oxide. This process can either occur
with the metal and oxidizer in the gas phase (a vapor phase reaction) or with the metal as a
condensed phase (a heterogeneous reaction). If the oxide vaporization-dissociation
temperature is less than the boiling point of the fuel, combustion must proceed
heterogeneously on the particle’s surface. This concept is well known as “Glassman’s
criterion” for the vapor-phase combustion of metals. The behavior of metal combustion can
be determined by comparing the boiling point temperature to the temperature at which the
metal oxide product is decomposed or dissociated to gas-phase molecules. This concept was
also applied to different metal-nitrogen systems (Glassman and Papas, 1994) by using a
thermal equilibrium calculation. Their results showed that metal particles, such as Al, Be, Fe,
23
Cr, Hf, Li, Mg, and Ti, should have the ability to burn as vapor-phase diffusion flames at 1
atm in pure oxygen. In contrast, B and Zr would be expected to burn heterogeneously.
Consequently, the existence of “limiting” temperature results noted above puts a constraint
on the temperature distribution in metal combustion in where a “flat-top” temperature
distribution may prevail over a significant range of radial distances above the particle’s
surface.
The limiting temperature concept was only recently verified experimentally for the
combustion of Al particles in oxygen environment (Bucher et al., 1996, 1998). In the
experiment, a stream of mono-disperse, burning particles having reproducible size was
produced by continuously chopping wire strands by mechanical shearing and laser ignition.
The particles were formed with zero initial velocity and accelerated to Reynolds numbers of
the order of 0.1 during a free fall. In the experiment, the flow of particle cylinders attracted to
each other (end-to-end) was moved into densely focused (0.1 mm beam diameter) radiation
from a 150W CO2 laser. Upon contact with the laser beam, the temperature of the cylinder
with a free end rapidly increased above the melting temperature of aluminum. During
exposure to laser radiation, the small cylinders contracted into a spherical particle having a
diameter around 210 μ m, ignited, and began to traverse vertically downwards through a
glass column filled with air. The sequence of events was repeated individually for each
cylinder at a rate of typically one particle per second resulting in a highly dispersed stream of
mono-sized, burning particles. Time-exposed, natural luminosity images of burning particles
were obtained using a gated CCD camera. Temporally-resolved temperature measurements
were obtained using a two-camera, two-extinction line, planar laser-induced fluorescence
(PLIF) technique. Bucher et al.’s experimental results of temperature distribution confirmed,
24
for the first time, the existence of a limiting flame temperature in metal particle combustion.
As shown in their results, the nearly flat temperature profile over a wide range of nondimensional particle radii was similar to that inferred by equilibrium calculations.
Although the general classification of metal combustion is well-defined, it is clear
that the particular mode of combustion of a metal is dependent upon the oxidizer and
environmental conditions. Williams (1997) summarizes several of the dominant criteria in
classifying metal combustion in Table 2.1. The three columns of this table contain three
criteria in the overall classification. The first discriminator determines whether the available
energy exceeds the energy required to heat and volatilize the final metal oxide. The second
discriminator is also an energy statement to ascertain if the available energy exceeds the
energy required to heat and vaporize the metal. The third discriminator, intersolubility of the
metal and its product, is also relevant to combustion behavior. For volatile metals, certain
solubility combinations are known to break up the original particle. For non-volatile metals,
purely condensed phase combustion may result.
Table 2.1. Classification of metal particle combustion (Williams, 1997)
Volatile Product
Non-volatile Product
Volatile metal, gas phase
combustion
Non-volatile metal, surface
combustion
Volatile metal, gas phase
combustion
Soluble
Non-soluble
Soluble
Non-soluble
Soluble
Non-soluble
Product may
dilute metal
during
burning and
cause
disruption if
its boiling
point
exceeds that
of metal
No flux of
product to
metal
Product may
build up in
metal during
burning
No product
penetration
into metal
If product
returns to
metal it
may dilute
it and cause
disruption
Disruption
strongly
favored
product
returns metal
25
2.2.2 Metal particle combustion regimes
From the metal combustion classification (Table 2.1 and discussed above), it is
obvious that metal combustion can occur either heterogeneously at the particle surface or
homogeneously in the surrounding gaseous environment. The combustion of metal particles
introduces a length scale into the problem and, hence, time scales for mass and energy
transport. Transport time scales may be compared with chemical time scales to define the
mode of combustion controlling macroscopic features, such as burn rates and ignition delays.
In a kinetically-controlled regime, the reaction rate is slow compared with the rates of mass
and energy diffusion so that spatial non-uniformity is eliminated. For the kineticallycontrolled regime, the chemistry is slow, occurring through a heterogeneous reaction at the
particle surface, and consequently, near spatial uniformity is observed in the surrounding
environment. The oxidizer concentration at the particle surface is nearly that of free stream.
The changes of temperature, fuel concentration, and product concentration along with
distance away from particle surface remain small.
When reactions are fast, the spatial non-uniformities of temperature and composition
fail to be eliminated in the available combustion times. For this diffusion-controlled regime,
the oxidizer concentration approaches zero at the flame, whether the flame is located at the
particle surface or in the surrounding gas phase. As a result, gradient of temperature and
species are established in space. Such gradients cause conduction of heat and diffusion of
species towards lower temperatures and concentrations, respectively. Reactants diffuse into
the flame zone, whereas, combustion products diffuse away from the flame zone. Such
unmixed combustion is said to be diffusion controlled.
26
Vapor phase combustion models have included accumulation of the oxide in a
spherical surface at a reaction sheet or distributed over an extended reaction zone, as well as
surface condensation of the oxide (Law, 1973; Law and Williams, 1974; Brooks and
Beckstead, 1995; Marion et al., 1996). Analytical models for heterogeneous combustion of
metal particles have also developed. Most of them were developed to describe boron particle
ignition and combustion (Li and Williams, 1991; King, 1993; Yeh and Kuo, 1996).
Under a diffusion-controlled condition and for Lewis number of unity, the mass
consumption rate of a particle per unit mass burning in a quiescence environment is
(Williams, 1985; Glassman, 1996)
ρD
m&
ln(1 + B)
=
2
4πrp
rp
(2.13)
where B is the mass transfer number (Spalding, 1955), ρ is the gas density, D is the gas
mass diffusivity, and rp is the particle radius. For a vaporizing metal droplet, a convenient
form of B for a particle with heterogeneous surface reactions is obtained from the coupling
function of the fuel-oxidizer species equations:
BOF =
(iYO , ∞ + YF , S )
(1 − YF , S )
(2.14)
Since there is no volatility of fuel, YF , S = 0 , and BOF = iYO , ∞ , and the consumption
rate per unit particle mass reduces to
ρD
m&
=
ln(1 + iYO , ∞ )
2
rp
4πrp
The combustion time for a particle with a heterogeneous surface reaction can be
deduced as
(2.15)
27
tb , diff =
ρ p d 02
8 ρD ln(1 + iYO , ∞ )
(2.16)
Under a kinetically-controlled combustion mechanism, the oxidizer mole fraction at
the surface, X O , S , is approximately equal to X O , ∞ . Therefore, the mass consumption rate of
the particle per unit mass is
m&
= ( MW ) p kPX O , S ≈ ( MW ) p kPX O , ∞
4πrp2
(2.17)
where X O is the oxidizer mole fraction, (MW)p is the molecular weight of particle, P is the
pressure, and k is the surface reaction rate with the oxidizer. Thus, the combustion time from
the initial particle size to burnout is
tb , kin =
ρ p d0
2( MW ) p kPX O , ∞
(2.18)
To determine the dominant combustion mechanism, the Damkohler number, Da , for
surface reaction is defined as
Da =
tb , diff
tb , kin
=
( MW ) p kPd0 X O , ∞
4 ρD ln(1 + iYO , ∞ )
(2.19)
These results show that if Da =1 is defined as the transition between diffusioncontrolled and kinetically-controlled regimes, an inverse relationship exists between the
particle diameter and the system pressure at fixed Da . The above equation also shows that
large particles at high pressures probably experience diffusion-controlled combustion, and
small particles at low pressure often lead to kinetically-controlled combustion.
28
2.2.3 Metal combustion in oxidizer atmospheres
Metal combustion has been of great research interest because of the high energy
density produced. It was found that the addition of metal particles, such as aluminum,
magnesium, titanium, and boron metal particles, to solid propellants could significantly
increase the performance of rocket systems. Efforts have been made to propose the detailed
combustion models for aluminum (Bucher, 1998; Liang and Bechstead, 1998), magnesium
(Abbud-Madrid et al., 1999, 2001), and boron (Zvuloni, 1991; Cho et al., 1992; Zhou et al.,
1998, 1999; Foelsche, 1999) particles in an oxygen-environment.
Metals may be important fuels for the establishment of a lunar mission base and the
exploration of Mars as well. The atmosphere of Mars is 95% carbon dioxide and is the likely
choice of an in situ oxidizer. Many metals can be directly oxidized by CO2 to produce energy
without the need of further processing, e.g., to produce O2. Consequently, metal combustion
in oxidizers other than air has produced a lot of scholarly attentions, such as aluminum in
CO2 (Yuasa and Sogo, 1992; Legrand et al., 2001; Rossi et al., 2001), magnesium in CO2 or
CO2/CO mixture (Shafirovich and Goldshleger, 1992; Shafirovich et al., 1993; Yuasa and
Fukuchi, 1994; Fukuchi et al., 1996; Legrand et al., 1998), and lithium in CO2 (Yuasa and
Isoda, 1992).
Aluminum has been widely used in solid propellants and explosives, and, thus,
aluminum particle combustion is of great practical interest and has been extensively studied.
Based on “Glassman’s criterion”, aluminum combustion was considered a vapor-phase
combustion. However, recent studies (Dreizin, 1996, 1999b, 2003; Marion et al., 1996;
Yagodnikov and Voronetskii, 1997) showed evidence of oxygen buildup in burning
aluminum particles, which suggested that heterogeneous metal/oxygen interaction does occur
29
on the metal particle surface. The experimental results revealed that there are three distinct
stages in the aluminum combustion history. Spherically symmetric vapor phase combustion,
consistent with the conventional metal vapor-phase burning model, occurs during the first
stage. The second stage of aluminum particle combustion is associated with an increase in
the size and density of the smoke cloud surrounding the particle, a shift to a nonsymmetrical
combustion regime, and initiation of particle spinning. A finite content of oxygen builds up
in the burning particles at this time. The particle temperature is close to the boiling point of
the metal during the first two combustion stages. In the third combustion stage, an “oxide
cap” forms and grows on the burning particle, which continues to spin and burns
nonsymmetrically. The particle temperature decreases after it reaches the Al2O3 melting point.
Recently, aluminum flakes with submicron thickness are used instead of regular
aluminum powders (Tadahiro et al., 1997; Kosnake et al., 2000). It is expected that because
of the high specific surface of flakes, they could ignite easier than spherical particles of
similar mass. An accelerated ignition should result in an acceleration of the overall burn rate
of aluminum flakes as compared to regular powders. Eapen et al. (2004) compared the
ignition and combustion behavior of aluminum flakes with regular spherical aluminum
powders, based on the comparisons of pressure rise rates, consumption of oxygen from the
air in vessel, and completeness of the aluminum conversion to oxide based on the analysis of
the condensed combustion products. However, an unexpected trend of less complete and
slower combustion for the aluminum flakes as compared to the regular aluminum powders
was observed. They speculated that the relatively slow rate of combustion and incomplete
reaction observed for aluminum flakes are due to agglomeration of flakes before their
30
ignition within the small scale eddies produced as a result of flakes’ interaction with the gas
flows induced by the propagation flame.
With the prospect of using CO2 breathing engines on Mars, and because of the high
level of CO2 present in the post-combustion gases of hydrocarbons and conventional and
nitramine-based propellants, equilibrium calculations (McBride and Gordon, 1996; Bucher,
1998) were performed on the Al-CO2 system to illustrate the effects of a change in heat of
reaction. The Al-CO2 system was observed to burn with a vapor-phase diffusion flame
(Marion et al., 1996; Yuasa et al., 1996). The equilibrium calculation results confirmed that
no condensed-phase carbon is formed in fuel-lean mixtures. The maximum adiabatic
temperature occurs at stoichiometry with respect to the over reaction. Bucher et al. (1999)
studied the combustion of Al in CO2 and water vapor at a pressure of 1 atm. Compared with
combustion in O2, the gas-phase flame zone was significantly smaller for CO2.
Although aluminum is characterized by a higher heat of combustion than magnesium,
magnesium offers an important advantage over aluminum. Namely, it ignites much easier
than aluminum, especially in CO2 containing atmospheres. In some cases, this advantage
may be the deciding argument in favor of magnesium. Compared to different combustion
stages of aluminum, magnesium was found to combust in the vapor phase (Legrand et al.,
2001). The difference was due to the fact that the boiling point for magnesium is much lower
than for aluminum. It was also confirmed that the main source of heat release is
homogeneous condensation of MgO formed by the reaction of Mg vapor and the oxidizer.
They also found that magnesium particles ignite in CO2 if the partial pressure of CO2 in the
atmosphere exceeds some critical value. The critical pressure of ignition increases with the
decrease in particle size. The dependence of the ignition probability on the partial pressure of
31
CO2 implies that ignition of Mg particles is controlled by chemical kinetics. In contrast to
magnesium, the ignition probability of aluminum particles in CO2 is less than 50%. The
difference may be associated with the different properties of oxide films on the surfaces of
particles.
A new proposed Al-Mg mechanical alloy was studies by Schoenitz et al. (2003). This
high energy density material ignites at a temperature comparable to that of pure magnesium,
much lower than the temperature required for pure aluminum. For example, the Al0.9Mg0.1
mechanical alloy ignited at 1150 K, close to 1000 K for Mg, compared to 2200 K for Al. The
comparison of explosion parameters and combustion products of the Al-Mg alloy and pure
Al or Mg powders of different sizes showed much faster combustion rates and higher
explosion pressures. Thess improved combustion parameters appear to be attractive for
practical applications: highly energetic, readily ignitable, and easy to burn completely.
Another complex metal particle consisting of an aluminum core covered by a nickel
layer (nickel coated aluminum) was proposed because of reduced agglomerate and a lower
ignition temperature compared to pure aluminum (Shafirovich et al., 2002, 2005; Andrzejak
et al., 2007). It was demonstrated that the flame propagation velocity is significantly higher
than that for pure aluminum. The reduced ignition temperature (~ 1000 C), lower than the
nickel melting point (1455 C), suggests that the oxidizers react with solid phase Ni
heterogeneously on the surface of a coated particle. However, the intermetallic reaction
between Al and Ni play the main role in the ignition mechanism, especially the product
NiAl3 plays a critical role in ignition, leading to self-acceleration of the reaction.
32
2.2.4 Metal combustion in microgravity
The set of partial differential equations for natural convection problems even without
the existence of combustion is coupled, non-linear, and usually elliptical, and the equations
must be solved simultaneously. When the combustion process and the associated finite-rate
chemical kinetics (exponential in temperature and therefore computationally sensitive) are
added to the problem, tractable solutions of the basic equations to even the simplest of
combustion system is often beyond the state-of-the-art. As a result, in most practical
combustion applications, natural convective influence is necessarily neglected, often without
scientific justifications.
To minimize the effect of the buoyancy-driven flow in the combustion system, low
pressure was utilized (Chung and Law, 1986). However, low pressure may not diminish the
buoyant flow to a completely negligible level, despite the nearly spherical flame shape that
was observed experimentally. The low pressure also increases the mean free molecular path
and reduces the number of molecular collisions in the reaction zone and broadens the
reaction zone thickness. This is an undesirable consequence of trying to reduce the buoyancy
force. A better way to reduce the influence of buoyancy is to reduce the gravitational body
force. A 1 mm droplet can be studied experimentally and be free of buoyancy influences, if
the body force is reduced by approximately 10-4, a level readily achieved in low-gravity
experiment facilities. The effects of the earth’s gravity on an object can be reduced to very
small levels in a number of ways such as drop towers and free-falling aircraft during
parabolic flight. The following conditions are achieved when buoyancy-induced flows are
nearly eliminated:
1. The influence of weaker force and transport processes can be isolated.
33
2. The sensitivities to ambient oxygen concentration, pressure, and fuel flow can
be enhanced.
3. Flame flicker, due to hydrodynamic instability, is eliminated. Thus, a more
steady flame can be studied experimentally.
4. Near-extinction behavior of the flame can be probed at lower flow strains,
requiring less extrapolation for fundamental burning velocities in the “zerostrain” environment.
5. Settling or sedimentation is nearly eliminated in microgravity. In concept,
unconstrained suspensions of monosize, stationary, large fuel droplets or
particles may be created and sustained in microgravity during combustion and
enable a high degree of symmetry.
6. Characteristic time scales may be increased in microgravity.
Only a very limited number of studies have been reported on metal combustion under
microgravity conditions. The earliest studies appear to have been reported in the Russian
literature in 1978. Zenin et al. (1999) have continued this work on single aluminum particles.
Other studies include the metal rod combustion studies by Steinberg et al. (1992), the single
aluminum particle studies by Dreizin (1999a), the magnesium aerosol combustion studies by
Dreizin and Hoffman (1999, 2000), and the bulk magnesium combustion studies by AbbudMadrid et al. (1999, 2001). In addition, Goroshin et al. (1995, 1999) and Pu et al. (1998)
have reported on flame propagation through metal particle clouds at microgravity conditions.
In the studies by Steinberg et al. (1992) metal rods of iron, stainless steel, titanium,
and aluminum were burned in high pressure oxygen in a 2.2s drop tower. In their
experiments, the rods were ignited at one end and combustion was observed as the flame
34
propagated along the rod. As in normal gravity, a molten ball was formed at the end of the
wire. However, in microgravity, detachment of the ball did not occur. The following general
observations were made: (1) the absence of the buoyant forces did not extinguish combustion,
(2) the regression rate of the melting interface of the cylindrical rods is significantly greater
in microgravity than in normal gravity, (3) volatile combustion products are produced, an
event that did not occur under similar conditions in normal gravity, (4) the regression rate of
the melting interface is dependent on the oxygen pressure, as was observed in normal gravity,
and (5) excess oxygen above stoichiometric requirements is contained in the molten ball
formed, as is also observed in normal gravity.
The single aluminum particle studies by Dreizin (1999a) were conducted onboard a
DC-9 aircraft with nearly motionless 100-150 μ m aluminum particles evolving from the tip
of a wire subjected to a microwave discharger in air. The results from this study reported
similar combustion times and temperatures for normal and microgravity conditions. A nonsymmetric flame structure and brightness oscillations were observed to develop at the same
combustion times whether in normal or microgravity. It was concluded that flame nonsymmetry is an intrinsic feature of aluminum particle burning rather than the result of forced
or natural convection flows.
The work of Zenin et al. (1999) on aluminum particle combustion was conducted in a
variable pressure chamber, mounted on a free falling platform. The particle was ignited by a
ruby laser prior to free fall. Data were collected on the particle combustion time, sizes,
evolution of combustion zones, and densities and sizes of combustion products accumulated
on the burning particles and in the surrounding environment. Combustion was studied in
mixtures of 20% O2 in Ar and N2 and in 100% CO2 for pressures ranging from approximately
35
1 to 70 atm. For combustion in the 20% O2/80%Ar mixture, burn times were found to
correlate to initial diameter squared. For combustion in either 20%O2/80%N2 or the 100%
CO2 environment, the burn times were found proportional to d 1.5 . The lower burning rate
exponent was attributed to increased oxide coverage of the particle surface. The sizes of
luminous flame zones found in microgravity were approximately 24-28 particle diameters,
which are considerably larger than in normal gravity. The results also showed that the
relative amount of accumulated Al2O3 on the burned Al particle increased significantly with
pressure. The collected particle after combustion has the normal density of Al2O3 only for the
mixture with Ar. Lower densities of burned particles were obtained for other mixtures, which
also varied with pressure. The change in density was attributed to the difference in porosity
of the solid combustion products, which results from different controlling combustion
mechanisms for the three different oxidizing environments. The effect of pressure on the
surrounding oxide particles was to shift the peak in the size distribution from 230 nm to
about 300 nm in going from 1 to 40 atm.
The bulk metal combustion of magnesium was reported by Abbud-Madrid et al.
(1999, 2001). Cylindrical magnesium samples of 4 mm diameter and 4 mm height were
ignited in pure oxygen or CO2 at 1atm. The ignition source consisted of a 1000 W xenon
lamp. Flame propagation rates were compared with theoretical results from fire spread
analyses and were found to be significantly less in low gravity than in normal gravity. In
contrast with the heterogeneous and non-symmetric combustion mechanism of aluminum
combustion, a gas flame was clearly evident for magnesium. In the absence of gravity and
buoyancy-induced convection, both the molten metal specimen and the surrounding flame
exhibit a spherical shape during the burning process. In addition, the condensed oxide
36
particles formed as products of the combustion accumulated and agglomerated on the outer
edge of the visible flame front. Whereas, in normal gravity, the products are swept upward by
buoyancy induce currents, condensed oxides rapidly accumulate and agglomerate in the
reaction front in microgravity, producing a highly radiant flame front. Here, a high particle
density in the flame front generates a large heat flux to the sample, which raises the surface
temperature and increases metal evaporation. Consequently, the flexible oxide membrane,
which keeps magnesium particle at temperatures below its boiling point expands as vapor
pressure builds up inside the metal core. As evaporation increases, so does the flame front
diameter to accommodate greater oxygen flux and maintain the stoichiometry. At the peak of
the cycle, the amorphous specimen is transformed into a spherical core having a diameter
twice the size of the original cylinder.
Microgravity also presents a unique opportunity to create a cloud consisting of
relatively large diameter metal particles so that both cloud flame structure and individual
particle combustion behavior can be characterized simultaneously. In normal gravity, sustain
flows greater than gravitational settling force are necessary to maintain a particle suspension
for long periods of time. These sustained flows often reach a turbulence regime very easily,
so only low concentrations of small particles can be stabilized in normal gravity. High dust
loading leads to gravitational instability of the dust cloud and to the formation of
recirculation cells in a confined volume or to sedimentation of the dense cloud in an
unconfined volume (Goroshin and Lee, 1999).
Dreizin and Hoffman (1999, 2000) studied the combustion of magnesium aerosols in
a 2.2s drop tower. Particles with sizes between 180-250 μ m were aerosolized in a 0.5 L
combustion chamber and ignited in a constant pressure, microgravity environment. During
37
the experiments, two flame images were produced simultaneously, using interference filters
separating adjacent MgO and black body radiation bands at 500 and 510 nm, respectively.
The characteristic MgO radiation was used as an indicator of the gas phase combustion.
Experiments also showed that in microgravity the flame speed depends on their initial
particle speeds. The dependence is, most likely, due to the role the moving particles play in
heat transfer processes. Strong MgO radiation and production of dense MgO smoke clouds
were observed in all the experiments, including those with the slow propagation flames.
Therefore, Dreizin and Hoffman concluded that the MgO produced in the vapor phase is not
the primary source of the MgO coating found on the burned particle surfaces. An alternative
mechanism of forming the oxide coating suggested that the oxide coating was formed via the
formation of a metal vapor solution, followed by a phase separation occurring within the
burning particles.
Goroshin and Lee (1999) performed dust cloud experiments in the KC-135 aircraft on
aluminum with concentrations as high as 1200 g/m3 and particle sizes of about 18 μ m
diameter. The experiments were conducted with semi-open tubes with free expansion and
overboard venting of the combustion products. These Bunsen-type dust flows were used to
measure laminar burning velocities and quenching distances for variations in dust and
oxygen concentrations. They also developed a theoretical model to calculate the quenching
distance and compare theoretical results with experimental results.
2.3 Flammability Limits and Flame Quenching
A fuel is usually considered to be flammable if external ignition results in the
formation of a flame, which can propagate through the mixture. It has been found empirically
38
that flame propagation in hydrocarbon-air mixtures is quenched if its temperature is lowered
to about 1000-2000 C and the propagation velocity at the moment of extinction has a finite
value of a few centimeters per second. The existence of flammability limits is a result of heat
loss to the surroundings. When a certain relationship exists between the heat loss rate and the
heat release rate is satisfied within the flame front, the flame ceases to propagate and dies out.
The first attempt to analyze the problem of flammability limits theoretically was
made by Zeldovich (1944). He related the occurrence of flammability limits to the
phenomenon of heat transfer from the preheat zone or chemical reaction zone to the
surrounding walls and formulated appropriate equations. He also showed that all flammable
mixtures should have limit compositions, below which flame propagation is impossible, due
to heat loss from the flame to its surroundings. The influence of diffusion, particularly in
cases where the Lewis number differs from unity was studied (Zeldovich and Barenblatt
1959). Latter, Spalding (1957) proposed simplified, one-dimensional models for the
extinction mechanism based on the thermal theory of heat from the flame to its surroundings.
Among the various other proposed theories have been several based on various
assumptions. These include flame quenching due to the effects of convection (Hertzberg,
1980), chemical kinetics (Macek, 1966), flame stretch as a result of the existence of velocity
gradients (Lewis and von Elbe, 1961), preferential diffusion of one of the reactants of the
flame (Bregeon, 1978), and the action of factors bringing about instability (Kydd and Foss,
1964). It follows that the mechanism of flame quenching is probably a combination of more
than one factor, depending upon the particular flame.
39
2.3.1 Gas fuel flame quenching
The study of flame quenching was initiated by Humphrey David in 1985, when he
became interested in devising ways to prevent explosions in coal mines. He was able to
design the well-known Davy safety lamp, which depends on the principle that explosions in
methane will not pass through small apertures or tubes.
Payman and Wheeler (1923) conducted an experiment on the propagation of
methane-air and coal gas-air flame through tubes with a small diameter. They carried out
their experiments by recording the flame speed with the tube diameter for various mixtures to
demonstrate the cooling effect of the walls as the tube diameter was decreased. They noted
the effect of flame speed on the ability of the flame to pass through tubes or holes of small
diameter, and they established the fact that flames with higher propagation velocities are
more difficult to quench.
Holm (1932) provided a quantitative measure of flame quenching via the burner
method. The principle of this technique is to determine whether a flame stabilized on a
burner will flash back or be quenched when the flow of the combustible mixture is
interrupted. Methane-air and coal gas-air experiments were carried out at various
concentrations. It was concluded that while the thermal properties of the gaseous mixture
certainly affected the flame quenching, the thermal conductivity of the wall material was
comparatively unimportant. It follows the conclusion that flame extinction occurs, due to the
cooling effect of the unburned gas in contact with its external surface.
Berland and Potter (1956) designed a variable width rectangular channel burner to
study the effect of oxygen and inert diluent concentrations on propane-air quenching. They
carried out experiments on propane-air quenching as a function of fuel-air ratio and pressure,
40
and also explored the effect of the type of fuel on flame quenching, using eight fuels of three
groups-saturated, unsaturated, and aromatic hydrocarbons, as well as hydrogen. For lean
mixtures, they found that the quenching distance increased as the carbon chain is lengthened
or branched, while for rich mixtures, the quenching distance depends on the molecular
weight of the fuel with a calculated exponent of -0.3 and -0.5, for equivalence ratios equal to
1.5 and 1.7, respectively. An expression was established relating the pressure exponent of the
quenching distance to the overall reaction order, the pressure dependence of flame
temperature, and the flame activation energy.
Berland and Potter (1956) also studied the effect of inert gas on the quenching
distance by replacing helium with argon in the propane-oxygen mixture. Such replacement
would affect the thermal conductivity and diffusion of the mixture differently. Thus, its effect
on any flame property could be used to distinguish between the thermal and diffusion
mechanism. The experimental data confirmed that the well-accepted thermal mechanism was
responsible for flame quenching. They also empirically established a relationship between
the burning velocity, the boundary velocity gradient, and the quenching distance.
Another method, known as the flanged electrode method to determine the quenching
distance, was carried out by Lewis and von Elbe (1961). In this method, two spark electrodes
were flanged with glass plates and a series of experiments were conducted to obtain the
relationship between the minimum ignition energy versus the distance separating the
electrodes. When the electrode spacing approached to within a critical distance the curve
took a rather sharp vertical turn, meaning that the glass plate suppressed the development of a
self-sustaining flame. This distance was considered to be the quenching distance. Using this
method, the quenching distances for methane, propane and hydrogen-air mixtures were
41
measured as a function of pressure, fuel, and diluent concentrations. Quenching of methane
and propane with oxygen and nitrogen mixtures flames were also measured in cylindrical
tubes and between parallel plates by determining the limits of flammability for downward
propagation at atmospheric pressure.
Ballal and Lefebvre (1975, 1977) conducted experiments to study the influence of
pressure, velocity, turbulence intensity, turbulence scale, and mixture composition on
minimum ignition energy and quenching distance in flowing gaseous mixtures, including
methane and propane fuels. They also replaced the nitrogen in air with various inert gases,
such as carbon dioxide, helium, or argon in some experiments. It was observed that the
minimum ignition energy required for different inert gas substitutes descended by order—
carbon dioxide, helium, nitrogen, and argon—or both methane and propane fuels. It was
found the minimum ignition energy decreased with increased pressure for both fuels. A slight
increase in oxygen content was found to achieve substantial reduction in minimum ignition
energy. Their results also showed that an increase in flow velocity necessitated a higher value
of minimum ignition energy, but the velocity per se had little or no effect on minimum
ignition energy except in so far as it changed the level of turbulence intensity.
They also derived equations for quenching distances at low turbulence regime and
high turbulence regime,
dq =
dq =
Ak / (c p ρ )
(S
L
− 0.16u ' )
Bk / (c p ρ )
(S
T
− 0.63u '
)
(2.20)
(2.21)
42
where SL is the laminar flame burning velocity, ST is the turbulent flame burning velocity, u '
is the turbulence intensity, A and B are constants of 10 which was determined by experiment.
By assuming the minimum ignition energy as the energy required to heat a spherical volume
of gas with diameter d q , the cubic relationship between the minimum ignition energy and
quenching distance was confirmed. This result is at variance with the ignition theory
developed by Lewis and von Elbe (1961), which predicts a square relationship.
Williams (1985) developed a one-dimensional model of flame quenching to
determine the limit conditions of flame existence and the characteristic points on the
temperature curve for the laminar flame. The temperature profile is shown in Figure 2.8.
Williams assumed the following: 1). Lewis number equals one. 2). Chemical reaction rate
equals 0 for x<0 and x>b. 3). Thermal conductivity and specific heat are constants. Then he
derived an expression of non-dimensional temperature as a function of non-dimensional
coordinates.
Figure 2.8. Temperature profile in a flame (Williams, 1985).
43
The above research deals mainly with the thermal aspects of heat loss from the
reaction zone to a cold wall. Research relating to a chemical quenching effect is
comparatively poor. Sloane and Schoene (1983) estimated heterogeneous radical
recombination on the surface by numerical and experimental approaches. They demonstrated
that the removal of the active radical from the reaction zone by diffusion to the surface
significantly affects flame stability during ignition and quenching. More recently, the effects
of thermal and chemical surface-flame interactions on flame quenching were examined by
Kim et al. (2006). They designed and prepared the cold wall plates with different materials,
and both inert and reactive plates were prepared to distinguish the effects of thermal
quenching and chemical quenching on methane-air flames. The experiments were also
carried out with different plate temperatures to estimate the effect on quenching. Their results
indicated there are three distinct regimes in the quenching behavior. For low surface
temperatures of the plates ranging from 100-350 C, the quenching is mainly dominated by
the heat loss to the plates and is independent of surface properties. When the surface
temperature of the plates increases beyond 400 C, the flame quenching is controlled by
radical removal to surface defects. When the plate temperature increases beyond 600 C, the
homogeneous chemical reactions overcome the effects of radical removal and the flame
becomes resistant to quenching.
Recently, research interest on quenching distance was extended to transient laminar
flames common in internal combustion engines. Because of the unsteady conditions, it
becomes difficult to measure head-on or side-wall quenching distances aggravated by the
small spatial scales of the phenomenon. Bellenoue et al. (2003) developed a method to
provide accurate measurements of head-on quenching distances using direct visualization of
44
the flame front. They used a direct photography method with high spatial resolution to record
the instantaneous image of a flame emission near the single wall surface of an obstacle
situated in a closed vessel. The quenching distance was measured as the thickness of the dark
space between the wall surface and the head of the luminous zone. During side-wall and near
the side-wall quenching surface, the shape of the emission zone was determined to be similar
to the curved plate. This two-dimensional character of flame-wall interaction in the
measurement area produced a high contrast image of the flame front and allowed an easier
detection of the quenching effect.
Sotton et al. (2005) studied the quenching of methane-air mixture in a direct way.
They proposed a formulation adapted to transient flame conditions, based on the chemical
reaction rate. The formation was experimentally validated for stoichiometric methane-air
mixtures. Although the formation predicts the tendency of a quenching distance quite well, it
requires an experimental coefficient to obtain quantitative correlation.
Boust et al. (2007) improved the head-on model to a thermal formulation of singlewall (both head-on and sidewall) flame quenching. The model describes a relationship
between the quenching distance and the heat flux, taking into account the effects of pressure
and mixture composition. The model was also validated in the case of head-on quenching as
well as sidewall quenching in agreement with experimental data. A major asset of this
formulation lies in the absence of empirical coefficients, which makes it a predictive tool to
improve near-wall calculations as far as flame-wall interaction is concerned.
45
2.3.2 Particle-gas cloud flame quenching
A limited number of studies have been reported on particle-gas cloud flame ignition
propagation and quenching in the absence of electric fields, e.g., Yu et al. (1983) utilize an
electric field as discussed later. This is mainly due to the experimental difficulties in the
generation of a uniform laminar dust suspension with a well-controlled reproducible
concentration, as well as due to the fact that different dusts and particle sizes of dust can
significantly influence the combustion mechanism. The limited experimental results known
in the literature are often apparatus dependent and contradict each other.
Ballal and Lefebvre (1978) studied the ignition and flame quenching of quiescent
multi-droplet fuel mists, including iso-octane, kerosene, diesel oil, light fuel oil, and heavy
fuel oil. They used a light scattering technique to measure the Sauter mean diameter of a fuel
spray or mist. Their experiments confirmed that both quenching distance and minimum
ignition energy are strongly dependent on droplet size, air density, equivalence ratio, and
volatility. A model was proposed for the ignition based on the assumption that chemical
reaction rates are infinitely fast, and the sole criterion for successful ignition is the generation,
by the spark, of an adequate concentration of fuel vapor in the ignition zone. They assumed
the process of ignition occur in the following manner. Passage of the spark creates a small,
roughly spherical, volume of gas whose temperature is high enough to initiate rapid
evaporation of the fuel drops contained within the volume. Assuming the chemical reaction is
fast enough compared to droplet evaporation and flame propagation, the spark kernel will
grow in size when the heat release rate of evaporation exceeds the heat loss rate by thermal
conduction at the surface of the flame volume. Based on these assumptions, they derived a
quenching distance equation for stagnant mixture as
46
[
]
d q = D ρ f ρ aφ ln (1 + Bst ) 2
1
(2.22)
where d q is the quenching distance, D is the mean droplet diameter, ρ f and ρ a are the
density of fuel and air, respectively, φ is the equivalence ratio, and the Bst is the mass
transfer number at stoichiometric ratio.
Ballal and Lefebvre (1979) extended the model of quenching distance in quiescent
mixtures to include both quiescent and flowing mixtures. The basis for the proposed model
for the ignition of flowing, heterogeneous liquid fuel-air mixtures is that, over a range of
velocity, pressure, fuel/air ratio, and mean droplet size, the ignition process occurs
independently of chemical reaction kinetic and is evaporation controlled. Also they assumed
that radiation effects are negligibly small, and that any effects arising from flows internal to
the fuel drops are insignificant for the relatively small drop sizes. Then, they derived the
quadratic equation for the quenching distance d q :
dq =
where
a=
b + b 2 − 4ac
2a
(2.23)
(
ρ a ka
φ log(1 + Bst ) 1 + 0.25 Re D 0.5
ρ f D2
)
b = 0.08c p ρ au '
c = ka
where k a is the thermal conductivity of air, c p is the specific heat at constant pressure, and
u ' is the root-mean-square value of fluctuating velocity. Equation (2.23) provides a general
47
relationship between quenching distance, fuel volatility, mean droplet size, and the physical
properties of the mixture. It is equally valid for both stagnant and flowing mixtures.
Ballal (1980) carried out the experiments to study the influence of particle size, dust
concentration, pressure, mass transfer number, and oxygen/nitrogen ratio on quenching
distance and minimum ignition energy of dust clouds of solid fuels, including aluminum,
magnesium, titanium, and carbon. His experiments showed that the mean particle size has a
strong influence on quenching distance, minimum ignition energy, and optimum spark
duration. The quenching distance and minimum ignition energy will increase along with the
increased particle diameter. For the same diameter of the same particle, the minimum
ignition energy and quenching distance will decrease with increased equivalence ratio. The
mass transfer number also plays an important role in ignition of solid fuels. The minimum
ignition energy and quenching distances decrease with increased mass transfer numbers
(different types of particles).
The results showed a strong similarity between the mechanism of ignition and
quenching of dust clouds of solid fuels and the liquid droplet/air mixtures. As reviewed
previously, “Glassman’s criterion” can be used to classify the metal combustion into gas
(vapor phase of metal) phase combustion and heterogeneous reaction at the surface of a metal
particle. For solid fuels such as aluminum and magnesium, which have higher boiling points
of the metal oxide than the metal, they generally can be treated as gas phase combustion—
very similar to liquid fuels. Based on the strong similarity, Ballal (1980) suggested that Eq.
(2.23) for quenching distance of liquid fuels is also applicable for solid fuels such as
aluminum and magnesium. For the quiescent dust mixtures, the equation for solid fuels can
be written as
48
ρf
⎡C3
⎤
d q = D32 ⎢ 3
⎥
⎣ C1 ρ aφ ln (1 + Bst ) ⎦
0.5
(2.24)
C1 = D20 D32
C3 = D30 D32
where D20 , D30 , D32 are surface mean diameter, volume mean diameter, and Sauter mean
diameter, respectively.
Ballal (1983a) further studied the theoretical model for the quenching distance and
minimum ignition energy. In addition to diffusion terms, chemical reaction and radiation heat
loss were also considered. He described the criterion for successful ignition as the time
required for evaporation and burning must be less than or equal to the time required for the
cold surrounding mixture to quench the spark kernel. Then he derived a modified quenching
distance equation as
d q = (8α )
0.5
−1
⎧⎡
9qC12εσTp4 ⎫⎪
C33 ρ f D322
12.5α ⎤
⎪
+
⎥ −
⎬
⎨⎢
2
S L2 ⎦⎥
c p ρ f C33 fD32 ΔTst ⎪
⎪⎩⎣⎢ 8C1 f (k / c p )φ ln(1 + Bst )
⎭
−0.5
(2.25)
where f is the swelling factor, k is thermal conductivity, q is the fuel/air ratio, ε is the
particle emissivity, σ is the Stefan-Boltzmann constant, S L is the laminar burning velocity,
Tp is the particle temperature, and ΔTst is the temperature rise at stoichiometric ratio.
To overcome the upward buoyant motion of burned gas flames and the downward
settling of the fuel particles relative to the flame front induced by gravitational force, Ballal
(1983b) conducted experiments in a zero-gravity environment. A vertical tube was filled with
a uniform quiescent dust cloud and the flame was initiated by spark ignition while the tube
fell freely. In this way, the influence of buoyant motion and the settling of fuel drops on the
49
propagation flame were then virtually eliminated. The burning velocities through quiescent
dust clouds of carbon, coal, aluminum, and magnesium were measured by recording the
flame propagation history inside the tube. It was found that burning velocities were
influenced by particle size, dust concentration, pressure, mass transfer number, and
oxygen/nitrogen ratio. It was also observed that the radiation heat loss from fine dust
particles can significantly reduce the rate of flame propagation through the dust cloud. Ballal
found that the radiation heat loss can reduce the burning velocity by 25% for the dust cloud
of 10 μ m aluminum particles. A theoretical model was also derived for flame thickness and
burning velocity and the predictions were consistent with experimental results.
Jarosinski et al. (1986) studied flame quenching in mixtures of cornstarch, aluminum
and coal dust with air. The experiments were carried out in a vertical tube 0.19 m inside
diameter and 1.8 m long with quenching plates held in the middle. The quenching distance
was measured for a flame propagating upward from the open to the closed end of the tube.
The minimum quenching distances were 5.5 mm for cornstarch at a dust concentration of 800
g/m3, 3.4 times the theoretical stoichiometric mixture, 10.4 mm for aluminum at a dust
concentration of 850 g/m3 which is 2.8 times the theoretical stoichiometric mixture, and 25
mm for fine coal with less than 5 μ m diameter at a dust concentration of 590 g/m3, 4.7 times
the theoretical stoichiometric mixture. They also used the quenching distance to identify the
controlling processes in laminar pre-mixed coal dust-air clouds, because there is a rigorous
relationship between flame thickness and quenching distance. The laminar flame thickness is
typically half the quenching distance (Jarosinski, 1984) for gas phase flames. From the point
of view of process similarities in the flames, the same relationship should be valid for both
gas and dust-air flames. It follows from this the flame thickness which corresponding to the
50
minimum values of quenching distance for cornstarch with air is approximately 2.8 mm and
for aluminum with air is approximately 5.2 mm.
Goroshin et al. (1996) studied the quenching distance of laminar flames in aluminum
dust clouds. They modified the pulse-jet technique (Jarosinski et al., 1986) to produce more
uniform and better controlled, stable dust concentration. Quenching distances were measured
in aluminum dust flames over a wide range of dust concentrations and at different oxygen
concentrations. Substitution of the inert diluent (nitrogen) with helium was also used to study
the influence of gas molecular transport properties on quenching distances. The laminarized
dust flow ascending in a vertical tube (0.05 m diameter, 1.2 m length) was ignited at the open
tube end. Constant pressure flames propagated downwards and a set of thin, evenly spaced
steel plates was installed in the upper part of the tube to determine the flame quenching
distance. Three different stages of flame propagation were observed: laminar, oscillating, and
turbulent accelerating flames. It was found that the quenching distance is a very weak
function of dust concentration in rich mixtures. The minimum quenching distance was found
to be about 5 mm in air and increased to 15 mm in mixtures of 11% O2. The substitution of
nitrogen for helium in air increased the minimum quenching distance from 5 to 7 mm.
Goroshin et al also developed a simple analytical model for a quasi one-dimensional
dust flame with heat losses, assuming that the particle burning rate in the flame front is
controlled by the rate of oxygen diffusion to the particle surface or towards a gas flame zone
that envelopes each particle close to its surface. It is noteworthy this assumption is in
disagreement with Ballal’s (1983). Then, Goroshin et al derived the governing equations for
gas flame and particle heating:
51
Vu ρ gu cg
Vu ms cs
dTg
dx
= λu
d 2Tg
dx
2
+ WF
ρ
ρ gu
− α gu (Tg − Tgu )
ρg
ρg
dTs λu
= (4πr 2 )(Ts − Tg )
dx
ru
(2.26)
(2.27)
where Vu is the burning velocity (index u indicates the value is related to the unburned
mixture); ρ gu , cg , Tg are density, specific heat, and temperature of the gas, respectively; cs ,
Ts , ms are specific heat, temperature, and mass of the solid particles; respectively. α is the
heat exchange coefficient between gas and the channel wall (for flat channel it can be written
as α = 2 Nu λ d 2 ), and WF is a heat source term.
The above research is based on the pulse-jet dispersion technique in which the
uniformity of dust concentration is dubious. Another disadvantage of gas dispersion is the
effect of turbulence on dust dispersion and flame propagation is the limiting concentration
that can be achieved. Kim (1989) used an electric particulate suspension (EPS) technique to
generated uniform suspension. He overcame the turbulence effect and reached higher particle
concentrations. The EPS technique is used in this research and will be introduced in detail in
Chapter 3. Table 2.2 gives a comparison of minimum quenching distances, lean flammability,
and particle concentrations for aluminum-air mixture. There are differences in Table 2.2,
especially the data from Jarosinski (1986). The differences are likely due to the turbulence
flow effect on flame propagation.
2.4 Summary
A literature review of investigations of particle and metal combustion shows that the
mechanism of particle ignition, and combustion and flame quenching are not well understood
52
and that experimental results are dependent upon the apparatus and limiting concentrations.
The EPS technique implements an electrostatic technique and offers an alternative method
for studies of quenching combustible dust mixture. Particle concentrations using EPS are
limited by gravitational force by gravity conditions, but can be extended to rich fuel mixtures
(this research) in microgravity generating highly uniform particulate clouds.
Table 2.2. Comparison of quenching distance and flammability
Ballal
Jarosinski et al.
Kim
Goroshin et al.
(1983a)
(1986)
(1989)
(1996)
Diameter ( μ m)
10
7.5-9.5
11.8
5.4
Minimum
quenching
distance d q (mm)
4
10.4
3.5
5
1.4
2.8
3.7
flat plateau
160
400
90
150
Equivalence
Ratio φ at
quenching
Lean
flammability limit
(g/m3)
53
CHAPTER 3. EXPERIMENTAL METHOD
The EPS method provides a unique electrostatic technique for dispersing and
controlling particle suspension (particle clouds) in vacuum and pressurized environments,
both in normal and zero gravity. It supports steady-state (flowing) and bath (confined)
processing of powders in normal and zero gravity. One advantage of the EPS method is that
it is compact and portable, and it does not require a significant sample size of powder
compared to traditional flow dispersion methods. It is also possible to generate highly
uniform particle suspensions at high particle concentrations in microgravity. In this chapter,
the experiment method, apparatus design, and experiment preparation are discussed.
3.1 Basic Experimental Setup
An electric particulate suspension can be generated by the application of applied
electric fields applied between parallel plates with readily available low power input (mWÆ
W) high voltage (kV) and low current ( μ AÆmA) power supplies. The basic setup for a
single EPS cell is shown in Figure 3.1. This new EPS setup provides independent suspension
(High Voltage 1) and spark (High Voltage 2) voltages from two variable 25 kV power
supplies. This independent high voltage arrangement offers a precise control of the particle
suspension and the ignition processes. At the beginning of the ignition tests, power supply 1
provides high voltage to suspend particles inside the EPS cell while power supply 2
independently charges the ignition capacitors. When a uniform particle suspension has been
generated inside a test cell, the high voltage sparking switch is closed, transferring the energy
stored in capacitors to the needle electrode and igniting the particle mixture. Acoustic
54
vibration is optionally added with a shaker table to aid in breakup of cohesive powders such
as pulverized coal and aluminum, which have small diameters (less than 10 μ m).
50 MOHM
HIGH
VOLTAGE 1
S2
NEEDLE ELECTRODE
50 MOHM
PYREX RETAINER
55 pF
HIGH
VOLTAGE 2
POWER
AMP.
FUNCTION
GENERATOR
He-Ne LASER
scan
TEFLON
VIBRATION EXCITER
SUSPENSION
Figure 3.1. Basic EPS setup.
For ground-based studies at 1 g, an automated vertical scanner (not shown in Figure
3.1) uses a He-Ne laser and a laser power meter (Metrologic model 45-540) to measure the
particle concentration profile, providing the information on cloud uniformity and
stratification in gravity (Greene, 2004). During the scan, the laser and laser power meter
remain stationary with the vertical movement of the test cell controlled by a stepper motor. A
LabVIEW program was developed to monitor initial laser intensity and varied laser intensity
during the suspension and then calculate the particle concentration based on the BeerLambert law:
N =−
ln(
It
Ii
AC lγ
)
(3.1)
55
where I t is the transmitted beam intensity, I i is the initial beam intensity, l is the beam
length passing through the suspension (the test cell diameter for our case), γ is the light
extinction coefficient, and AC is the mean integrated cross section for extinction (average
particle projected area).
Because of its size, the accurate single beam laser moving scan method was restricted
to normal gravity to measure the particle suspension concentration profiles. A more compact
measurement system was needed for use in drop tower microgravity experiments (Greene,
2004; Xu et al., 2008). The high speed laser-photodiode scanner system was designed for
EPS test cells during a 2.2 second drop in microgravity (Colver et al., 2008). Although the
photodiode scanning system was designed for drop tower microgravity experiments, it was
also used to measure particle concentration for ground-based experiments in this study.
The photodiode scanning system is shown in Figure 3.2 comprising the laser beam,
optics (expand beam) and photodiode scanner for detecting the laser intensity at different
heights in the EPS cell using Eq. (3.1). The 32-channel photodiode scanner schematic is
shown in Figure 3.3. The laser beam is expanded into a sheet ~5 cm (~2 in) in height using
successive concave/convex cylindrical lenses. Two photodiode arrays pick up the laser
intensity simultaneously and provide the average laser intensities at different heights.
The quenching distance was determined by trial ignitions (yes/no ignition tests) by
varying the amount of combustible powders in the EPS test cell and sparking the system with
the same ignition energy. The ignition energy is calculated by the product of the spark
voltage and capacitance. The separation of the test cell (quenching distance) remains the
same until the boundary (yes/no ignition boundary) particle concentration is found by trial
56
Photodiode array
EPS CELL
He-Ne expanded
Laser Beam
Multiplexer
High Voltage
Computer + Data Input
Figure 3.2. Laser photodiode scanning system (Greene, 2004).
(Manufactured Printed Circuit Board)
Analog Multiplexer
AI 0
Op-Amp
Gain
Control
Op-Amp
GND
GND
5V
Analog Multiplexer
AI 1
PhotoDiode PhotoDiode
Array
Array
Gain
Control
D0
D1
D2
D3
12V
Figure 3.3. Laser photodiode scanner (Greene, 2004).
57
ignition tests. Then, the test cell is changed to another separation (either by changing the test
cell with different separation or using test cells with adjustable separations) to repeat the trail
ignition tests. By repeating this procedure, a complete quenching distance versus particle
concentration can be obtained. The test cells used in this study have different separations of
20, 17, 14, 11, 8, and 5mm. The peak voltage is 25 kV and the capacitances remain 0.04 μ F,
so the maximum spark energy can reach 12 J.
3.2 Particle Preparation
For this research, different types of conductive and semi-conductive particles have
been tested: aluminum (Ai), magnesium (Mg), titanium (Ti), copper (Cu), iron (Fe) and glass
beads (Gl). Aluminum particles from ALCOA are sifted to different size ranges. Titanium
and iron particles are provided by AEE (Atlantic Equipment Engineers). Magnesium
particles are provided by Alfa Aesar. Copper particles are provided by U.S. Bronze. All
particles are assumed to be atomized regular powders. A particle size distribution was made
using a HIAC/ROYCO model 4300 analyzer. The particle size sensor (model HRLD 150)
ranges from 10-100 μ m. All particles were sifted to sieve size ranges in a sonic sifter (ATM
model L3P). The sieved particle size ranges are summarized in Table 3.1. Among these
particles, 15-20, 20-25 and 25-30 μ m Al, 63-74 μ m Cu and 53-63 μ m Gl were used in the
drop tower microgravity experiments. Additional particles were used in ground-base
experiments. The size distributions and micro-photos are shown in Figures 3.4-3.15. Microphotographs suggest that the copper particles are very close to a sphere, but aluminum
particles with smaller diameters are not exactly spherical. Other particles also show some
irregularity.
58
Table 3.1. Particle size range
Size Range ( μ m)
Particle
Copper (Cu)
30-38
Aluminum (Al)
44-53
15-20
Glass (Gl)
53-63
63-74
20-25
25-30
25-30
53-63
Magnesium (Mg)
< 44
Iron (Fe)
<44
Titanium (Ti)
< 44
0.5
Al (15-20)
Frequency
0.4
0.3
0.2
0.1
0
0
10
20
30
40
Particle diameter (um)
Figure 3.4. Size distribution and micro-photo of Al (15-20 μ m).
0.8
Al (20-25)
Frequency
0.6
0.4
0.2
0
0
10
20
30
Particle diameter (um)
40
Figure 3.5. Size distribution and micro-photo of Al (20-25 μ m).
59
0.4
Al (25-30)
Frequncy
0.3
0.2
0.1
0
10
20
30
40
50
Particle diameter (um)
Figure 3.6. Size distribution and micro-photo of Al (25-30 μ m).
0.6
Cu (30-38)
Frequency
0.4
0.2
0
10
20
30
40
50
60
Particle diameter (um)
Figure 3.7. Size distribution and micro-photo of Cu (30-38 μ m).
0.3
Cu (44-53)
Frequency
0.2
0.1
0
20
40
Particle diameter
60
Figure 3.8. Size distribution and micro-photo of Cu (44-53 μ m).
60
0.5
Cu (53-63)
Frequency
0.4
0.3
0.2
0.1
0
15
28.75
42.5
56.25
70
Particle diameter (um)
Figure 3.9. Size distribution and micro-photo of Cu (53-63 μ m).
0.3
Cu (63-74)
Frequency
0.2
0.1
0
40
50
60
70
80
Particle diameter
Figure 3.10. Size distribution and micro-photo of Cu (63-74 μ m).
0.5
Glass (25-30)
Frequency
0.4
0.3
0.2
0.1
0
0
10
20
30
Particle diameter (um)
40
50
Figure 3.11. Size distribution and micro-photo of Gl (25-30 μ m).
61
0.4
Glass (53-63)
Frequency
0.3
0.2
0.1
0
40
50
60
70
80
Particle diameter (um)
Figure 3.12. Size distribution and micro-photo of Gl (53-63 μ m).
0.4
Fe (<44)
Frequency
0.3
0.2
0.1
0
0
20
40
60
Particle diameter (um)
Figure 3.13. Size distribution and micro-photo of Fe (0-44 μ m).
0.3
Frequency
Mg (<44)
0.2
0.1
0
0
10
20
30
40
Particle diameter (um)
50
60
Figure 3.14. Size distribution and micro-photo of Mg (0-44 μ m).
62
0.5
Ti (<44)
Frequncy
0.4
0.3
0.2
0.1
0
0
10
20
30
40
50
Particle diameter (um)
Figure 3.15. Size distribution and micro-photo of Ti (0-44 μ m).
The various particle diameters can be calculated based on the particle size distribution,
depending upon the forms of weighing factors. The arithmetic mean diameter, d l , is the
average diameter based on the number density function of the sample. The surface mean
diameter, d S , is the diameter of a hypothetical particle having the same average surface area
as that of the given sample. The volume mean diameter, dV , is the diameter of a hypothetical
particle having the same averaged volume as that of the given sample. Sauter mean diameter,
d32 , is the diameter of a hypothetical particle having the same averaged specific surface area
per unit volume as that of the given sample. The diameters are defined by
∞
dl
∫ bf
=
∫ f
0
∞
d
2
S
∞
0
∫
∞
∞
d
(3.2)
b 2 f N (b)db
0
3
V
(b)db
N (b) db
0
∫
=
N
∫ b f (b)db
=
∫ f (b)db
0
3
N
∞
0
(3.3)
f N (b)db
N
(3.4)
63
d 32
∫
=
∫
∞
0
∞
0
b3 f N (b)db
(3.5)
b 2 f N (b)db
where b is the particle size range and f N (b) is the number density function for the particle
size range. In this study, the number density function is defined as the percentage of particles
inside certain particle size ranges. In the average size calculation, the midpoint of the size
range is used for b . The different particle diameters are summarized in Tables 3.2 and 3.3.
Table 3.2. Particle diameter table-part1
Cu
Cu
Cu
Cu
Gl
(30-38)
(44-53)
(53-63)
(63-74)
(25-30)
Gl
(53-63)
d l ( μ m)
35.41
48.06
55.37
63.95
26.56
58.47
d S ( μ m)
36.06
48.96
55.61
64.31
27.88
58.97
dV ( μ m)
36.70
49.74
55.83
64.65
29.01
59.43
d32 ( μ m)
38.02
51.34
56.26
65.34
31.40
60.38
Table 3.3. Particle diameter table-part2
Al
Al
Al
Mg
Ti
(15-20)
(20-25)
(25-30)
(0-44)
(0-44)
Fe
(0-44)
d l ( μ m)
16.62
22.78
28.65
30.35
24.58
27.06
d S ( μ m)
17.32
23.07
29.42
32.76
26.35
29.05
dV ( μ m)
18.01
23.28
30.22
34.95
28.04
30.95
d32 ( μ m)
19.47
23.72
31.87
39.78
31.76
35.16
These different diameters are used in this study for different purposes. For example,
in the determination of the light extinction coefficient, dV is used when the information of
number of particles in the test cell is needed; d S is used to calculate the extinction coefficient
64
when the integrated projection cross area is of most interest. During flame propagation and
quenching, the surface area per unit volume may be important. In this case, Sauter’s averaged
diameter d32 is used.
3.3 Light Extinction Coefficient
The light extinction coefficient, γ , can be determined by theoretical calculation and
experiment. For a sphere of arbitrary diameter and relative index, Mie theory can be applied
to calculate the extinction and scattering light by small particles. In this study, both
theoretical calculations and experiments are performed to determine the light extinction
coefficient and cross check with each other.
3.3.1 Theoretical calculation of extinction coefficient
The theoretical calculations performed in this study originate from the method
detailed by Bohren and Huffman (1983). In calculation of light extinction and scattering, the
dimensionless particle size parameter, x, is often used to combine the effect of the diameter
and the wavelength of incident light:
x=
πD
λ
(3.6)
where D is the diameter of the particle, and λ is the wavelength of incident light.
The light extinction coefficient γ is calculated by
γ=
2 ∞
∑ (2n + 1)Re{an + bn }
x 2 n =1
(3.7)
where an and bn are complex functions of size parameter and relative index of particle. They
can be written in the form of Riccati-Bessel functions:
65
mψ n (mx)ψ n' ( x) − ψ n ( x)ψ n' (mx)
mψ n (mx)ξ n' ( x) − ξ n ( x)ψ n' (mx)
(3.8)
ψ n (mx)ψ n' ( x) − mψ n ( x)ψ n' (mx)
bn =
ψ n (mx)ξ n' ( x) − mξ n ( x)ψ n' (mx)
(3.9)
an =
where m is the relative index of particle, ψ n ( x) , ψ n' ( x) , ξ n ( x) , and ξ n' ( x) are Riccati-Bessel
functions in forms of Bessel functions of first, second and third kind. More detailed forms
can be found in Bohren and Huffman (1983).
Note that Eq. (3.7) is an infinite series, so we have to determine the appropriate
number of terms. The final term was suggested as the integer closest to x + 4 x1 3 + 2 with Eq.
(3.7) rewritten as
γ=
2 N
∑ (2n + 1)Re{an + bn }
x 2 n =1
(3.10)
A Matlab program was written to implement the Eq. (3.10) for calculation of light
extinction coefficient. The laser used in this study has a wavelength of 670 nm. The relative
indices for particles (aluminum, glass beads, and copper) used in this laser scan study are 1.7,
1.6 and 0.326, respectively. The relative indices are taken from Xu (2000). The relative index
of refraction normally includes real and imaginary parts. However, this study only considers
the real part of the relative index. The effect of the imaginary part on the light extinction
coefficient has been examined. It is found that different imaginary parts change the
amplitudes of the coefficient at a lower size parameter. Calculations were performed from 0100 for the size parameter. The results are shown in Figure 3.16.
It is found that the largest extinction coefficient occurs when the particle and the light
wavelength have a similar size. The light extinction coefficient approaches the limiting value
66
Extinction Coefficient
4
Aluminum
Glass Bead
Copper
3
2
1
0
0
20
40
60
80
100
Size Parameter
Figure 3.16. Calculated light extinction coefficients curve.
of 2 as the size parameter increases. For larger particle sizes (extinction coefficient = 2), all
the geometrically incident light flux to be scattered by the usual processes of scattering
including reflection and refraction or partially absorbed within the sphere are considered as
well as an equal flux from the surrounding beam to be scattered by the normal Fraunhofer
diffraction.
In this study, the size parameters of particles are larger than 100. For example, for 2025 μ m Al, 53-63 μ m Cu and 53-63 μ m Gl, the size parameters are 139, 358 and 387,
respectively. Consequently, this study assumes the light extinction coefficient as 2 for
particle concentration calculations.
67
3.3.2 Experimental determination of extinction coefficient
The light extinction coefficient can be determined independently using EPS. The
experiments are carried out with the same EPS setup used for combustion testing in Figure
3.1. The separation distance of the test cell is chosen as 20 mm (any distance can be used).
The laser and the laser power meter are aligned at the middle height of the test cell, i.e., 10
mm from the bottom plate of the test cell. The high voltage power supply is increased to
generate the particle suspension until the laser power meter reading remains the same. The
stable laser intensity reading suggests a uniform and complete suspension of particles inside
the test cell. Both laser intensity before suspension (Ii) and after suspension (It) in Eq. (3.1)
are recorded. Assuming a complete suspension of particles, the number density inside the test
cell is calculated, based on the diameter and loading mass of particles. Then the light
extinction coefficient can be calculated, based on the Beer-Lambert law Eq. (3.1). The
particles used for the extinction coefficient experiment are aluminum (25-30 μ m), copper
(53-63 μ m) and glass beads (53-63 μ m).
The iron particles (102 μ m) reported here are taken from collaborative research with
McGill University. The unique EPS technique provides a uniform particle suspension with
known particle concentration in conjunction with a particle concentration measurement
system developed by McGill. The original purpose of this collaborative effort is to calibrate
their particle concentration device. The calculated extinction coefficients are in good
agreement with theoretical predictions.
The results are shown in Figure 3.17, presented as Ln(Ii/It) vs. Ac*l*N with the slope
equal to the light extinction coefficient. The experimental results are in good agreement with
68
the theoretical prediction value of 2 at low particle concentration. When the number density
in the test cell reaches a critical level the data deviate (less than) from the predicted value 2.
The disagreement at the high loading mass is possibly due to the limitation that the EPS is
suspending the mass in gravity. The maximum mass is determined by electric field intensity
E. This maximum-of-mass concentration is discussed in Chapter 3.4.3 and Chapter 4.
4.0
3.5
Aluminum
Copper
Iron
Glass Bead
Value=2
3.0
Ln(Ii/It)
2.5
2.0
1.5
1.0
0.5
0.0
0.5
1.0
1.5
2.0
AC*l*N
Figure 3.17. Experimental light extinction coefficients.
3.4 Experiments for Microgravity Study
The EPS experiment for microgravity has been taken from concept to prototype in
this research (Colver, 2007). A specially-designed EPS system was extensively tested in the
2.2-second drop tower (79 ft 1 in) at NASA Glenn Research Center (NGRC). The closed
system test cells were preloaded with different powder/oxidant ratios providing 5 quenching
69
data points in a single drop to determine quenching distance. With this method, complete
curves of ignition-quenching distance for powders (e.g., aluminum) and oxidants (e.g., air)
can be obtained. The specially-designed EPS combustion system included 5 test cells so that
5 ignition data points can be provided per drop. Each closed cell was comprised of two
parallel plate electrodes separated by Pyrex glass retainers (80 mm ID), varying in height
from 10 to 20 mm (the quenching distance).
3.4.1 Drop tower microgravity test
Experiments are dropped under normal atmospheric conditions from a height of 79 ft
1 in. using the 2.2 second drop tower facility at NASA Glenn Research Center (NGRC)
located in Cleveland, OH. All instruments and the experimental setup are mounted on an
assembly called the A-frame provided by NASA. Air drag force effect is minimized, using a
surrounding drag shield. The experiment (A-Frame) drops approximately 7-1/2 in. to the
bottom of the drag shield during a 2.2 second drop. This arrangement limits the accelerations
of the A-Frame to approximately 10-4 g. The complete assembly is hoisted to the top of the
drop tower where it is mated with the facility’s thin wire release mechanism. The entire
assembly remains supported by cables prior to the drop. Once the package is in place,
additional electrical connections are made with the experiment. These connections trigger the
onboard systems at the drop. After setup preparations (video and electronic signals
initialized), spacers are removed from the bottom of the drag shield, and the exterior drag
shield doors are closed. Next the overhead hoist cables are disconnected. At this point, the
entire package, which weighing up to 1075 lb, is suspended by a length of thin hardened steel
wire. The wire is cut instantaneously releasing the drag shield and the A-Frame of the
70
experiment rig smoothly into free fall. The release causes the opening of electrical
connections to signal the onboard data systems that the drop has been initiated.
A drop is terminated when the drag shield assembly impacts an air bag at the bottom
of the drop tower. At the time of impact, the experimental package has traversed the
available vertical distance within the drag shield and rests on the floor of the drag shield. The
deceleration levels at impact have peak values of 15 to 30 g. These relatively low impact
levels, as well as precautions taken during experimental design, permit the use of many offthe-shelf electronic items including video cameras, low-power lasers, light bulbs, and data
acquisition and control systems.
All electrical power for the A-Frame is provided by onboard 28 V dc battery packs.
Real-time videos are transmitted to remote video recorders via a fiber optic cable dropped
with the experiment. Onboard data acquisition and control systems also record data supplied
by instruments, such as pressure transducers, thermocouples, and flow meters.
3.4.2 EPS 5-tiered ignition system and A-Frame layout
The original EPS design for quenching distance experiments is a single test cell for
each ignition. The design is simple and the experiment rig is compact and rugged. To
facilitate the collection of data, a specially designed EPS 5-tiered system (Figure 3.18) is
developed, permitting up to 5 ignition measurements during a single drop (Colver, 2007;
Colver et al., 2008). The 5-tiered experiment is designed for quick disassembly and easy
cleaning/loading of powder samples after a single end plate is removed. The frame is
structured from three 20-inch long threaded rods that can accommodate up to 5 floating
(freely sliding) EPS cells. The ground side (positive) electrode of each EPS cell is a 10-inch
71
diameter stainless steel plate acting also as a floating support member. The complete rig
weighs 24 kg (53 lb) when fully assembled with 5 EPS cells. A banana plug terminates each
high voltage lead connecting the (negative) electrode plate of the cell.
Figure 3.18. EPS 5-tiered experiment rig.
To seal each test cell against high pressure combustion (calculated at 8 atm), 8 cm id
black O-rings are fabricated by lathe and placed on the top and bottom of each Pyrex retainer
(Greene, 2004). Each EPS cell assembly (except the bottom-most cell) slides freely along the
3 threaded rods and is held under a pre-load compression force by heavy duty springs on
each threaded rod. A single top compression screw transmits the same force to each EPS cell
along the center axis via large ball bearings acting as point force contacts. To help protect
against Pyrex glass breakage during the 15-30 g impact (after cell initial cell compression has
been set), individual nuts on each threaded rod are backed up against the floating stainless
72
ground plate to compensate for the accumulated weight effect of the upper tiers on lower
cells. In the event of a retainer glass breakage, the 6 stainless plates and combustion chamber
form enclosures at each tier level to contain glass fragments.
An adjustable screw needle point electrode is positioned at the center of each
grounded (positive) plate and exposed 1 to 5 mm to trigger a spark kernel in the suspension.
The threaded needles can be accessed inside stainless cylindrical seats holding each ball
bearing. The needle points are shaped by hand using a fine hobby file and Dremmel tool, and
positioned in the (positive) plate electrode utilizing the pitch of the threads (4-40). The
schematic of the 5-tiered system and instrument is shown in Figure 3.19.
Teflon
v
z
4"x6" Rectangular
Window (x2)
p
(-) HV Electrode
S
Pyrex Retainer
x
S
x
S
Expanded
Laser Beam
x
S
Photodiode Array
32 sensors
+
x
S
t
y
Figure 3.19. Schematic of EPS 5-tiered system.
Once assembled, the 5-tiered rig and 10-inch diameter stainless plates fit snugly
inside the combustion chamber located on the A-Frame. Inside the combustion chamber, the
73
rig is secured by 3 industrial strength springs (one on each threaded rod), slightly compressed
against the lid of the combustion chamber. The springs eliminate rattling and bouncing of the
rig during the drop impact. In the present study a large hole in the lid of the combustion
chamber is open to the atmosphere providing feed through for high voltage leads. Pressurized
powder combustion is also possible by pressurizing individual EPS test cells before loading
or by sealing the lid and pressurizing the combustion chamber, allowing test gases to enter
test cells, for example, through pin-holes using a scavenging gas process.
The high voltage (HV) suitcase (EPS power supply) is designed with dual adjustableprogrammable 0-25 kV power supplies for independent control of the EPS suspension
voltage and spark breakdown voltage (and energy). In previous ground-based studies, only a
single power supply was used for both suspension and spark. To protect the sensitive
onboard circuitry (microcontroller, photodiodes, op-amps, multiplexer, CCD camera) from
EMI (electromagnetic interference), each of 5 HV coaxial (shielded) leads running from the
HV relays is connected only to a single EPS cell and terminated with a ferrite bead. This
reduces ground loop currents and provides high frequency damping. A single lead supplies
both the EPS cell for suspension and sparking and enters the combustion chamber through an
opening in the combustion chamber lid.
The HV circuit shown in Figure 3.20 used in the 5-tiered rig follows the design of
dual power supplies shown in Figure 3.1 providing independent voltage control for the
suspension and the spark (Colver, 2007; Colver et al., 2008). The 5 high voltage relays for
sparking ignition are triggered (18-30 VDC solenoids) by the DDACS (droppable data
acquisition and control system) through the onboard 24-28 VDC (20~30 A) battery pack. At
74
25 kV, maximum spark energy of 12.5 J (=½CV2) can be stored in each capacitor bank
connected to an EPS cell in Figure 3.20.
to PDM J20B
Spark Trigger +28 VDC
S4-b
100 MOhm
100 MOhm
R2
R1
0.04 uF
C2
15 pF
C1
EPS TEST CELL
to PDM J20B
Spark Trigger +28 VDC
Teflon encased or connector
100 MOhm
S4-a
100 MOhm
R1
R2
D1
0.04 uF
D3
C1
[-]
7 pins
EPS TEST CELL
C2
[-]
MANUAL OVERIDE SWITCHES
S2
D3
15 pF
S3
7 pins
S1
[+]
0-> - 25 kV DC variable
Power Supply
[+]
to PDM J20D
HV Power Supplies +28VDC
0-> - 25 kV DC
variable
Power Supply
Figure 3.20. Dual HV circuit.
The actual energy delivered to the spark is less than the stored value from losses in
the circuit and ignition kernel. Manual knob adjustments of LCD displays at the front panel
of the suitcase are used to set voltages for the suspension and the spark before the drop (the
HV power supplies can also be programming with a ramping voltage etc.). The HV diodes
provide current, limiting protection when directly connecting two HV power supplies or from
spark induced oscillations or voltage surges in the event of an internal arc failure in the HV
relay. The 100 MΩ resistors limit the HV supply current to 0.25 mA when shorted by the
spark. These also serve to isolate the 5 sparking circuits from each other through the RC
(C=0.03→0.04 μF) time constant (~4 S) during spark discharge-recharging in the event that a
75
relay sticks in a closed position; however, this never occurred. Circuit isolation is provided
by the high voltage SPDT relays and ensures that each EPS cell receives a full charge from
its own capacitance bank for the spark.
The schematic layout of the experiment on the A-Frame is shown in Figure 3.21
(Greene, 2004; Colver, 2008). The actual picture of A-Frame is shown in Figure 3.22. The
major components by item number are: (1) 5-tiered EPS rig and combustion chamber (main
experiment and protective housing); (3, 4, 5) high voltage suitcase (power supplies, relays,
capacitor banks)—provides high voltage and charge storage for EPS suspension and sparking,
respectively); (10, 11, 13) laser-photodiode optics system (measures particle concentration
profile in middle EPS cell); (2, 9) CCD camera and mirror (for videos); (6) power
distribution module (PDM—distributes various camera & mirror (for videos); (6) power
distribution module (PDM—distributes various voltages to circuits); (7) DDACS-Tattletale
computer (experimental computer sequencing and data acquisition—TTBasic language). As
previously noted, only the middle EPS cell (Figure 3.19) can be viewed by the laser beam
and photodiode array sensor through 4x6 in2 windows located on opposite sides of the
combustion chamber. Similarly, the CCD camera can only view the middle EPS cell through
a circular window and mirror arrangement (2).
The electrical networking connections are shown as follows: (a) photo diode array
scan control, (b) power to the photo diode arrays, (c) bundle of 5 high voltage wires and
ground leads to individual EPS cells and 5 ground leads to combustion chamber, (d) power
on-off control to HV supplies, (e) the control wire for the high voltage relays, (f) the control
wire for the camera, (g) the main power distribution wiring, (h) +12 VDC supply to laser;
and (i) is DDACS to PDM feeds.
76
b
a
i
7
6
1
f
10
11
13
2
9
h
8
c
d
3
4
5
e
g
12
12
Figure 3.21. Schematic of A-Frame layout.
DDACS
Combustion
Chamber
PDM
Laser
Scanner
Laser +
Optics
CCD Video
Camera
Mirror
HV
Suitcase
A-Frame
Figure 3.22. Actual picture of A-Frame layout.
77
3.4.3 Drop tower control and data sequencing
To prepare a drop, each of the 5 test cells is fitted with a Pyrex retainer (8 cm id) of
specified height (the quenching distance) and loaded with a pre-weighed amount of
combustible powder to set the fuel/air ratio. Ambient air is allowed to fill each test cell. A
selection of sizes and kinds of powder samples, prepared at ISU, are packaged and placed in
multi-layer of foam and finally secured in a suitcase and transported to NGRC. After
numerous designs and trials, small 5 cm squares of paper, folded and stapled into triangles
prove to be a reliable method for storing and recovering the powder samples by sweeping
with a small, soft artist’s brush. The mass and size ranges of the samples are aluminum (73667 mg; 15-20, 20-25, 25-30 μ m), glass (1789 mg; 53-63 μ m), and copper (6818-7044 mg;
63-74 μ m). These values are chosen to include an excess of powder for rich mixture
combustion testing at the very high concentrations anticipated at 0 g. The maximum-of-mass
concentrations produced in microgravity for a given electric field intensity can then be
compared with maximum suspensions at normal gravity. Such data are important to the
development of theory and modeling of suspension dynamics with/without gravity.
The event sequencing for a drop includes startup of the high voltage power suitcase,
presetting voltages for the suspensions and charging capacitance banks, recording laser
photodiode scans of particle concentration before (1 g) and during the drop (0 g) for the
middle EPS test cell, sparking the powder ignition, and shutdown of the high voltage power
suitcase. The detailed sequence is discussed next
Before the EPS 5-tier system is fitted into the drag shield, weighed out amounts of
powder are placed into individual test cells and the voltages for suspension and ignition are
78
set to specific values. After the EPS system is fitted into the drag shield, a TTBasic program
is downloaded into the on-board DDACS and executed. The laser and the photodiode scanner
are enabled first. The laser intensities at different heights are scanned and results are saved in
DDACS for future retrieval. This scan is performed 3 times to determine the average
readings. The laser intensity data at this step give the laser profile without the suspension (i.e.,
initial laser intensity). After the initial scan is completed, the HV power for suspension is
triggered by the controlled PDM (power distribution module) to suspend the powders inside
the test cells. The laser intensities are scanned 3 times and saved in DDACS. The stored laser
intensity data in this step provide a laser profile of the suspension at 1 g. When this step is
completed, the program will notify the performer that the system is ready for drop.
When the drag shield is released from the top of the drop tower, a banana cable on the
top of the A-frame will be detached to inform the DDACS is informed that the drop is
initiated by listening to a certain port. A loop sequence in the TTBasic program is designed
to listen to the status change at this port, so that the next step can be executed correctly.
When the drop is initiated, laser intensities at different heights are scanned 3 times and saved
in DDACS. The laser intensity data at this step give the laser profile with a suspension (0 g).
After this final scan, the ignition relay is triggered to ignite the suspended powders. Finally,
the HV power for both suspension and ignition is disabled. A detailed control flow chart is
shown in Figure 3.23.
DDACS is a microcomputer system developed by NASA for drop tower microgravity
experiment control and low-speed data acquisition. The DDACS is based on an Onset
Computer Tattletale model 4A, single-board computer and an in house-designed daughterboard that provides several ancillary functions not supplied by the Tattletale. There are
79
Start
No
B=3?
Yes
Select CH0
Suspension Scan With Microgravity
B=B+1
Suspension Scan With Gravity
A=1
Select CH15
Sample, Convert
and Store 2
Arrays Outputs
A=A+1
Initial Scan Without Suspension
Sample, Convert
and Store 2
Arrays Outputs
C=1
No
Enable Laser, Scanner
Select CH0
Yes
Drop?
Sample, Convert
and Store 2
Arrays Outputs
Select CH15
Sample, Convert
and Store 2
Arrays Outputs
Select CH0
Trigger Sparks
Sample, Convert
and Store 2
Arrays Outputs
Select CH15
Sample, Convert
and Store 2
Arrays Outputs
C=C+1
Yes
C=3?
No
B=1
No
A=3?
Yes
Enable HV Power
Disable HV,
Spark Relay
Stop
Control Program Flowchart
Figure 3.23. Drop tower control flowchart.
different types of units available. The one used in this study is equipped with a 12-bit A/D
converter, a 28 KB configurable onboard memory, 8 analog input channels, and 16 digital
output channels. The analog channels take 0-5 V single-ended inputs. The default sample rate
us 100 Hz (8 channels) and sample rates are up to 1400 Hz (1 channel). These 16 digital
channels are grouped into 4 ports (A, B, C, and D) to control the PDM (power distribution
module) and other instruments. The PDM provides a variety of voltages (5, 12, and 28V) and
is controlled by DDACS. Among the 28 KB onboard memory, 16 KB is assigned to program
memory and 8KB is assigned for data memory. To make use of these available equipments,
researchers have to make their own cables and hardware to incorporate with the apparatus.
The detailed drop tower DAQ (data acquisition) diagram used in this study is shown in
Figure 3.24.
80
Multi1
J20
D0
A0
I/O A
D1
A1
D2
D3
A2
D4
D5
I/O B
DDACS
A4
I/O C
A6
I/O D
A7
D6
D7
D8
D9
D10
D11
D12
D13
Multi4
HV Spark
Relay
+28V
HV Power
Multi3
D
C
B
A
+5V
D0
D1
D2
D3
Laser Intensity
scanner
A5
D14
D15
+28V
PDM
J21
A3
Multi2
D
C
B
A
+12V
Laser
EN
out1
out2
Digital
Analog
Figure 3.24. Drop tower DAQ diagram.
The sampling time for a single photodiode sensor is programmed at 30 ms (plus
execution time) giving a total scan time of 0.48 s for a 16-photodiode array (two arrays are
sampled simultaneously). This limits to 3 the number of full scans (one concentration profile)
possible during the 1.6-1.8 seconds of microgravity. Sparking of the 5 EPS test cells is
triggered simultaneously, 1.6 -1.8 seconds into the drop by the DDACS-Tattletale computer
immediately after the scan is finished, allowing sufficient time for a steady-state suspension
to be developed while in microgravity. The particle collision time constants for gravity
adjustment are typically a few ms, which is sufficiently less than the time of 1.6 -1.8 seconds
before sparking.
A total of 9 scans are utilized for the initial laser scans, suspension laser scans at 1 g,
and suspension laser scan at 0 g. For each single photodiode sensor, a 12-bit data is saved in
81
DDACS, requiring 2 bytes of storage space in the onboard data memory. The maximum
required data storage space during a drop is about 576 bytes, much less than the storage
space available on the DDACS.
Both EPS and sparking voltages are manually set initially with knobs located on the
front of the HV suitcase by using the LCD readouts. Under computer control, the
concentration profiles are measured and stored by the laser-photodiode scanner before the
drop (1 g) and during the drop (0 g). Videos of the suspension development, and sparking
and ignition events are recorded for the middle tier EPS test cells starting before and
throughout the drop by the onboard CCD (30 frames/sec) camera. The camera is positioned
using a mirror to “see” the light scattered from the particulate cloud from an expanded 5 cm
(height) red laser beam that traverses the center of the EPS cell. Photos of EPS suspensions
in microgravity are presented later in this dissertation.
The 3 mW He-Ne laser, when expanded optically through the concave/convex
cylindrical lenses, produces a parallel ray sheet of light ~ 5 cm in height (Colver, 2008). The
lenses are spaced according to their focal lengths. The two photodiode arrays (Photonic
Detectors Inc. PDB-C216) has 16 photodiodes (32 in total) which, when placed end to end,
intercept at 5.0 cm (1.57 mm between photodiode centers) of the expanded beam. This means
between 5 and 13 active photodiodes can “see” the laser sheet, depending upon the height of
the Pyrex cylinder retainer used in a particular quenching test (0.8, 1.1, 1.5, 2.0 cm Pyrex
heights available). The reduced number of readings at smaller quenching distances amount to
a loss in spatial resolution and ones ability to reproduce the curvature of the concentration
profile.
82
The video signal is transmitted via a long fiber optics cable from the camera back to
the remote VCR and TV monitor located on the top floor of the drop tower. Following a drop,
the rig and test cells are removed from the combustion chamber, opened and examined to
determine if burning occurred (yes/no for ignition). Still photos are taken of the EPS test cells
(still sealed inside their transparent Pyrex retainers). Some interesting “stringer and web”
formations are present from the combustion and photographed again when opened. Photos of
these powder formations are presented later in the report.
3.5. Summary
The experimental methods, basic EPS mechanisms, and test cell setups are introduced
in this chapter. The EPS method is simple, utilizing electrostatic forces to suspend powders
inside combustion test cells. This method avoids the non-uniformities caused by the flow jet
of the traditional pneumatic flow dispersion methods. A scanned movable EPS system
utilizes a point laser for ground-based studies (1 g), while a laser sheet scan is utilized in
microgravity (0 g).
Particle diameter size distributions and photos used in testing are included. For the
maximum suspension experiments, a variety of particles are reviewed, including Al, Cu, Fe,
Mg, etc. For quenching experiments, the combustible powders are Al and Mg. Microscope
photographs show that aluminum particles are irregular spheres. Copper particles are very
close to spherical, but include irregular spheres. The glass bead particles shown are highly
spherical in shape for both diameters investigated. Smaller particles such as titanium,
magnesium, and iron, are highly irregular. For theoretical purposes, all particles are treated as
spheres.
83
For the purpose of measuring particle suspension concentrations, the light extinction
coefficients are determined independently by theoretical and experimental methods. At low
particle suspension concentrations, theoretical and experimental results match well with each
other. At high particle suspension concentrations, experimental results deviate from predicted
theoretical values, possibly due to the limitation of maximum suspension produced by
electrostatic forces.
A specially designed 5-tier EPS system is used for the drop tower test providing the
acquisition of multiple data during a single drop. Videos and laser-photodiode scans are
taken of the middle EPS test cells of the rig before and during the drop to observe suspension
development and for the measurement of particle concentration profile, respectively. Each
test cell is opened, photographed, and examined, following a drop to determine if ignition
was achieved for the given quenching distance.
84
CHAPTER 4. PARTICLE SUSPENSION CONCENTRATION
Achieving uniform dust concentration has been identified as the key problem to be
resolved in dust flame studies. With the EPS method, cloud uniformity can be improved
under certain conditions. Gravity limits EPS in the following ways: (1) increased cloud
stratification; (2) lowered fuel/air ratios attainable for rich mixtures and reduced quenching
distances (for given electric field intensity); and (3) shifting the ignition-quenching curve
from the requirement of increased electric field intensities (increasing particle-oxidant
relative velocity). Item (1) is an important focus area for use of EPS as a ‘benchmark’ design
for flame propagation in dusty gas mixtures and provides NASA with a fire safety standard
for testing gas-powder mixtures. Item (2) is important for achieving large fuel oxidant ratios,
which has remained elusive in normal gravity testing. Item (3) identifies a new reaction
mechanism in combustion chemistry that, to date, is virtually unexplored. The approach to
item (3) is to isolate the particle-gas relative velocity effect using the dual power supply
system of Figure 3.1, wherein the suspension and sparking voltages are independently
controlled.
The following methods can be used to determine particle concentration with EPS:
1. Measure the weight (mass) of power placed in the cell and convert to number
density, using particle material density and volume; divide by volume of cell –
this is the simplest method, but must confirm 100% of the powder is dispersed
for a given electric field intensity. This technique is used for both onboard (0
g) and our ground-based laboratory (1 g) experiments.
85
2. Measure laser attenuation applied to the Beer-Lambert law to give the particle
number density—best for low density suspensions and must correct extinction
coefficient for scattering by small particles and finite apertures of photocells
(lens focusing is preferred but was not convenient here for our drop tests).
This technique is used for onboard (0 g) and in laboratory (1 g) experiments.
3. Calibrate maximum mass suspended (number density) against current flow in
external EPS circuit or suspension electric field intensity. This technique is
used here in ground-based experiments (1 g) to determine maximum
suspension as follows. Starting with a known mass of powder in the cell and
increase the E field and observe leveling off of external current at maximum
particle concentration. Calculate the maximum particle concentration from the
known mass suspended and cell volume at critical values of current and
electric field intensities.
In this chapter, the particle concentration measurement with method (2) is introduced
at normal and microgravity. The maximum concentration determined by method (3) will be
introduced in the next chapter.
4.1 Particle Concentration Measurement in Normal Gravity
In this research, two EPS setups are designed to measure the particle concentration
profile along the height. The first design (Figure 3.1) uses an automated vertical scanner with
a He-Ne laser and a laser power meter (Greene, 2004; Colver et al., 2004). Three different
particle materials (aluminum, copper and glass beads) are used for the measurement of
suspension concentration profiles with this method (Xu et al., 2008).
86
The particle concentration profile shown in Figure 4.1 is the particle concentration
profile of aluminum (size range 20-25 μ m). The average surface diameter of 24.72 is used
to calculate the particle concentration from the Beer-Lambert law. The loading mass of the
particles is 0.2 gm, separation between two EPS plates is 3.9 cm, and the diameter of the
glass ring is 8 cm. The electric field intensity varies from 400 to 700kV/m.
4
400kV/m
500kV/m
600kV/m
700kV/m
Height (cm)
3
2
1
0
0
10000
20000
30000
40000
50000
3
Particle Number Density (#/cm )
Figure 4.1. Concentration profile of 0.2 gm Al (20-25 μ m) with moving laser scan - 1 g.
The aluminum concentration profile shows uniformity in the middle of the glass
container, but non-uniformity at the position close to the plates. At different electric
intensities, the concentration profiles show a similar trend—flat and uniform concentration at
the middle and stratification at the two EPS plates. When the electric field intensity, E, varies
from a lower value to larger values, the suspended particle concentration is correspondingly
87
increased. The concentration profiles are expected to be the same (or close to each other)
when the electric field intensity reaches a certain high value, i.e., the particles inside the
chamber are all suspended uniformly.
The particle concentration profile shown in Figure 4.2 is the particle concentration
profile of glass beads (size range 53-63 μ m). The average surface diameter of 58.97 is used
to calculate the particle concentration from the Beer-Lambert law. The loading mass of the
particles is 0.2 gm, separation between two EPS plates is 3.9 cm, and the ID of the glass ring
is 8 cm. The electric field intensity varies from 400 to 800kV/m.
4
Height (cm)
3
400kV/m
500kV/m
600kV/m
700kV/m
800kV/m
2
1
0
0
5000
10000
15000
20000
3
Particle Number Density (#/cm )
Figure 4.2. Concentration profile of 0.2 gm Gl (53-63 μ m) with moving laser scan - 1 g.
The glass bead concentration profile shows similar trends with aluminum—
uniformity in the middle of the glass container, but non-uniformity at the position close to the
88
plates. When the electric field intensity, E, varies from a lower value to larger values, the
suspended particle concentration increases correspondingly. The concentration profiles
approach overlapping each other when the electric field intensity reaches a critical value
(500kV/m for this case of glass beads) for the maximum-of-mass in suspension. This means
that all loaded particles are fully suspended inside the chamber. Some stratification is still
apparent, but is small compared to aluminum (Figure 4.1).
The particle concentration profile shown in Figure 4.3 is copper powder (size range
53-63 μ m). The average surface diameter, 55.61 μ m, is used to calculate the particle
concentration from the Beer-Lambert law. The loading mass of particles is 1.0 gram,
separation between two EPS plates is 3.9 cm, and the diameter of the glass ring is 8 cm. The
electric field intensity varies from 400kV/m to 1100kV/m.
4
400kV/m
500kV/m
600kV/m
700kV/m
800kV/m
900kV/m
1000kV/m
1100kV/m
Height (cm)
3
2
1
0
0
5000
10000
3
15000
Particle Number Density (#/cm )
Figure 4.3. Concentration profile of 1.0 gm Cu (53-63 μ m) with moving laser scan - 1 g.
89
The copper concentration profile shows similar trends to aluminum powders and
glass beads, i.e., uniformity at near the center and non-uniformity at the position close to the
plates, especially at the bottom plate where stratification is apparent. As observed with all
other suspensions, when the electric field intensity, E, in increased the suspended particle
concentration correspondingly increases. Another characteristic of copper suspensions is the
requirement of a higher electric field intensity compared to aluminum and glass beads. The
increase is due to the weight of the particles (high density of copper) compared to aluminum
and glass requirements to overcome gravity.
The concentration profiles of copper approach each other when the electric field
intensity reaches a critical value of about 800kV/m. The stratification is more obvious at the
higher electric intensities as compared to aluminum and glass beads, due to the high density
of the copper particle. When copper particles collide with each other, the charges will be
exchanged so that some particles are neutralized or have a reduced charge and tend to collect
at the lower electrode. A second reason is the particle shielding effect from charge gradients
in suspension (Gauss’ Law).
The moving laser scanning method is well suited for the measurement of
concentration profiles. However, it is not suited for the microgravity experiment because of
its large size and slow scan speed (2.2 seconds available for the microgravity test). An
alternative photodiode EPS (Figure 3.2) has been especially designed for microgravity
(Colver, 2007). Compared to the moving laser scan system, the laser array scan system is fast,
but has a lower resolution in the number of readings, due to the spacing of individual
photodiodes (Colver et al., 2008). Particles used in the scanning EPS system have also been
tested with the photodiode EPS laser sheet array system. To make the best use of the total 32
90
photodiode cells in ground based tests, a second set of convex/concave lenses located
between the glass test ring and the laser photodiode array are applied to expand the laser
sheet height from 2 cm to about 5 cm. Therefore, detection of the laser intensities is at all
photodiode cells.
The particle concentration profile shown in Figure 4.4 shows the particle
concentration profile of glass bead (size range 53-63 μ m). The loading mass of particles is
0.2 gm, the separation distance between two EPS plates is 2 cm, and the inside diameter of
the glass ring is 8 cm. The electric field intensity varies from 400 to 1000kV/m.
Height (cm)
2
400kV/m
500kV/m
600kV/m
700kV/m
800kV/m
900kV/m
1000kV/m
1
0
0
5000
10000
15000
20000
3
Particle Number Density (#/cm )
Figure 4.4. Concentration profile of 0.2 gm Gl (53-63 μ m) with laser array scan - 1 g.
Compared to the particle concentration profiles from a moving laser scan at the same
loading mass, the concentration profile from the laser array scan gives similar profiles and
91
comparable values at the same electric field intensity. The maximum electric field intensities
tested are 900 and 1000kV/m. The particle concentrations at these two high electric
intensities nearly overlap the lower electric intensities (700 and 800kV/m), suggesting that all
loading particles are suspended inside the chamber by electrostatic forces, i.e., the lower
intensities are the critical values for maximum-of-mass suspension.
The copper particle concentration profiles (size range 53-63 μ m) are presented in
Figures 4.5-4.9. The loading mass varies from 0.1 to 0.5 gm. The separation between two
EPS plates is 2 cm, and the diameter of the glass ring is 8 cm. The electric field intensity
varies from 400 to 1200kV/m.
Height (cm)
2
400kV/m
500kV/m
600kV/m
700kV/m
800kV/m
1000kV/m
1
0
500
1000
1500
2000
2500
3
Particle Number Density (#/cm )
Figure 4.5. Concentration profile of 0.1 gm Cu (53-63 μ m) with laser array scan - 1 g.
92
Height (cm)
2
400kV/m
480kV/m
560kV/m
600kV/m
680kV/m
760kV/m
800kV/m
880kV/m
920kV/m
1000kV/m
1
0
0
1000
2000
3000
4000
5000
6000
3
Particle Number Density (#/cm )
Figure 4.6. Concentration profile of 0.2 gm Cu (53-63 μ m) with laser array scan - 1 g.
Height (cm)
2
400kV/m
500kV/m
600kV/m
700kV/m
800kV/m
900kV/m
1000kV/m
1100kV/m
1200kV/m
1
0
1000
2000
3000
4000
5000
6000
7000
3
Particle Number Density (#/cm )
Figure 4.7. Concentration profile of 0.3 gm Cu (53-63 μ m) with laser array scan - 1 g.
93
Height (cm)
2
400kV/m
500kV/m
600kV/m
700kV/m
800kV/m
900kV/m
1000kV/m
1100kV/m
1
0
1000
2000
3000
4000
5000
6000
7000
3
Particle Number Density (#/cm )
Figure 4.8. Concentration profile of 0.4 gm Cu (53-63 μ m) with laser array scan - 1 g.
Height (cm)
2
400kV/m
500kV/m
600kV/m
700kV/m
800kV/m
900kV/m
1000kV/m
1100kV/m
1
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
3
Particle Number Density (#/cm )
Figure 4.9. Concentration profile of 0.5 gm Cu (53-63 μ m) with laser array scan - 1 g.
94
A 100% suspension of particles is reached when the electric field intensity reaches
650kV/m for a 0.1 gm loading mass. But the minimum electric field intensity for complete
suspension is 760kV/m for a 0.2 gm loading mass. The minimum electric field intensity for
complete suspension is about 900kV/m for a 0.5 gm loading mass. It can be concluded that
the required electric field intensity for complete suspension increases with an increase in the
loading mass of particles.
Another direct way to determine the minimum electric field intensity for complete
suspension is to integrate the particle number density along the cell height and then calculate
the average particle number density. Assuming a known mass of powder in test cells
suspends uniformly inside the test cells, the number density can be calculated from the
known loading mass and the volume of the test cell. By comparing the calculated number
density with the measured average number density by the lease attenuation method, we can
determine if the loaded powders suspend completely or not. One example of this method is
shown in Figure 4.10. The result in Figure 4.10 is directly an integration result from Figures
4.5-4.9. The average number density at a given electric field intensity is calculated by
N ave =
∫
H
0
N ⋅ dh
V
(4.1)
where N ave is the average number density inside a test cell, N is the number density at
different height h, and V is the volume of the test cell. The average measured number density
can be compared to the calculated theoretical number density, based on the loading mass,
particle diameter, and particle density.
Assuming a spherical particle, the theoretical number density can be calculated by
95
N cal =
6m
πρV dV3V
(4.2)
where N cal is the calculated theoretical number density inside the test cell, m is the loading
mass of the powder, ρV is the density of the particle, dV is the volume mean diameter of the
particle, and V is the volume of the test cell. By comparing the theoretical number density
with the measured average number density, it can be determined if a maximum-of-mass
6000
3
Average Number Density (#/cm )
concentration has been formed for the given electric field intensity.
5000
4000
0.1 gm
0.2 gm
0.3 gm
0.4 gm
0.5 gm
3000
2000
1000
0
400
600
800
1000
1200
Electric Instensity (kV/m)
Figure 4.10. Maximum number density of Cu (53-63 μ m) at given E - 1 g.
A comparison of calculated number density and measured average number density for
different amounts of powder loaded in the EPS cell is listed in Table 4.1. The calculated
number density and measured average number density are in close agreement. The largest
96
difference between the two of them occurs at 0.3 gm loading powder, about 13% variance.
The other comparisons are pretty close, less than 5% variance. The comparison also shows
the laser attenuation method works well for measuring particle concentrations.
Table 4.1. Comparison of N ave and N cal
Loading mass (gram)
N ave (#/cm^3)
N cal (#/cm^3)
0.1
~1286
1226
0.2
~2376
2448
0.3
~3191
3672
0.4
~4756
4896
0.5
~6060
6120
The average number density curves taken at different loading masses in Figure 4.10
show similar trends. The average number density increases with the increased electric field
intensity, reaching a saturation stage following a critical electric field intensity, which
indicates a maximum-of-mass of the powder. By increasing the powder mass introduced in
the cell, the required minimum electric field intensity for complete suspension is increased as
well for a complete suspension of all particles. For example, the minimum electric field
intensity is about 600 kV/m for 0.1 gm of powder, but the minimum electric field intensity
for complete suspension increases to about 820 kV/m for 0.5 gm powder. The change of
minimum electric field intensity for different loadings of copper powder is shown in Figure
4.11. The minimum electric field intensity for complete suspension increases almost linearly
with the increase of loaded mass.
97
850
Minimum Electric Intensity
Electric Intensity (kV/m)
800
750
700
650
600
0.1
0.2
0.3
0.4
0.5
Loading Mass (gm)
Figure 4.11. Minimum E of Cu (53-63 μ m) vs. mass loading - 1 g.
For these studies at 1 g, the measurement of particle concentration profile using laser
attenuation method is confirmed here, using integration for a low mass loading and low
electric field intensity. However, an issue using laser attenuation at large particle
concentrations is that light scattering affects the light extinction coefficient as demonstrated
in Figure 3.17. Furthermore, very high powder concentrations can block virtually 100 % of
the incident laser light. Consequently another method based on the electric field intensity has
been developed for the loading mass under these conditions. The so-called “Excess Electric
Field Intensity” is introduced in Chapter 5. Both the laser attenuation and electric field
intensity methods are utilized in microgravity tests.
98
4.2 Particle Concentration Measurement in Microgravity
The achievement of increased as well as more uniform particle concentrations in
combustion tests of powders in microgravity (compared to normal gravity) are important
goals of the present study. In fact, higher powder concentrations with symmetrical profile
distributions are observed for all sizes and types of powders tested at NGRC (aluminum,
glass, and copper). During the drop tower tests, the EPS suspension voltage is adjusted to a
preset value and remains constant before and during the drop. Because of the limitation of the
drag shield, only the middle EPS test cell can be accessible by the laser sheet. Three scans
are made to measure the laser voltage readings to calculate the particle concentrations at
normal gravity and microgravity.
One example of an on-board laser voltage scan raw data of aluminum (20-25 μ m) is
presented in Figure 4.12. The data were taken at 3.5 kV suspension voltage for a 667 mg
loading mass of aluminum. The height of the test cell is 11mm, including ~ 1 mm thickness
from the rubber rings located on both top and bottom of the Pyrex retainers. The electric field
intensity is about 292 kV/m. As noted in the previous chapter, the total height of the
photodiode array is about 50 mm high. Only some of the photodiode cells “see” the laser
sheet, while other cells are blocked. The measurement procedure follows an initial scan
(without particles) and two subsequent scans (with particles) at normal and microgravity. It
appears that the dark field laser voltage (to be subtracted) is about 1.8 V and the parallel laser
sheet profile follows a Gaussian distribution.
Based on the laser scan voltages at different positions and the Beer-Lambert law
introduced in the previous chapter (Eq. (3.1)), the particle suspension concentration profile
99
can be determined. The height of the photodiode array is converted to the height of the test
cell and the resolution of position is about 1.57 mm per data point. The particle concentration
profile corresponding to Figure 4.12 is presented in Figure 4.13. It is obvious that the particle
concentration profiles at both normal and microgravity are seen to be non-uniform (constant
value), but symmetrical (without stratification) for the given electric field intensity (292
kV/m) and loading powders (667 mg). And, it is obvious that a reduced gravitational force
allows an increase in the amount of suspended particles in the test cell (example up to 2-6
times more) compared to normal gravity (Colver, 2007; Xu et al., 2008; Colver et al., 2008).
50
Initial Scan
Normal Gravity
Microgravity
height (mm)
40
30
20
10
0
1.8
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
Laser Intensity (V)
Figure 4.12. Laser intensity profile of 667 mg Al (20-25 μ m) at E=292 kV/m.
100
12
292 kV/m at 1 g
292 kV/m at 0 g
10
Height (mm)
8
6
4
2
0
5000
10000
15000
3
Particle Number Density (#/cm )
Figure 4.13. Particle concentrations of 667 mg Al (20-25 μ m) in 11 mm test cell.
16
14
Height (mm)
12
10
8
533 kV/m at 1 g
533 kV/m at 0 g
600 kV/m at 1 g
600 kV/m at 0 g
6
4
2
0
-2
20000
25000
30000
35000
40000
45000
50000
55000
60000
3
Particle Number Density (#/cm )
Figure 4.14. Particle concentrations of 500 mg Al (20-25 μ m) in 14 mm test cell.
101
The effect of different electric field intensities is also tested during the microgravity
experiments (Colver, 2007; Xu et al., 2008). A comparison of the particle suspension
concentration profiles is presented in Figure 4.14. The powder is aluminum (20-25 μ m) with
a (smaller) loading mass of 500 mg and test cell retainer height of 14 mm. The experiments
are run at voltages of 8 and 9 kV, respectively. The total electrode separation is 15 mm,
including the rubber O-rings giving electric field intensities of 533 and 600 kV/m
respectively. The effect of reduced gravity is again seen to increase the amount of powder
suspended. In this case, reduced gravity produces about twice the particle concentration,
compared to normal gravity, for both suspension voltages. A uniform aluminum suspension
is also achieved at the electric field intensity of 600 kV/m for microgravity. Also compared
to the suspension at 292 kV/m presented in Figure 4.13, more particles are suspended at a
higher electric field intensity, despite the fact that more powders is placed inside the test cells
at a lower electric field intensity. This means that 100% of the powder is not in suspension
and is deposited on the bottom of the test cells or as sometimes observed on the container
walls.
Another example of an aluminum suspension is presented in Figure 4.15 for a larger
aluminum particle of diameter range 25-30 μ m. A 400 mg aluminum sample is loaded
inside a test cell of 20 mm height. The suspension voltage is 11 kV, providing an electric
field intensity of about 524 kV/m. The reduced gravity effect is also obvious: the particle
concentration at microgravity is about 1.5 times larger than the concentration in normal
gravity. However, a uniform particle suspension is not achieved, although a symmetric
profile is observed for both 1 and 0 g.
102
20
524 kV/m at 1 g
524 kV/m at 0 g
Height (mm)
15
10
5
0
3000
3500
4000
4500
5000
3
Particle Number Density (#/cm )
Figure 4.15. Particle concentrations of 400 mg Al (25-30 μ m) in 20 mm test cell.
The particle concentration profiles of glass (53-63 μ m) particle is also examined and
presented in Figure 4.16 comparing 0 g and 1 g suspensions. A total of 1789 mg powder was
placed inside a test cell of 20 mm height. The experiments were performed at different
voltages of 3, 4.3, and 6 kV. At E=143 kV/m, only small quantities of particles are suspended
in both normal and microgravity in contrast to expectations and the recording video. One
possible reason is particle deposits on the retainer Pyrex glass, affecting the laser beam
intensity. At higher electric intensities, an increase in the concentration in microgravity is
clear. For example, at E=205 kV/m, the concentration of powder suspended during
microgravity increases by a factor of 4 over normal gravity, while at E=286 kV/m the
increase is a factor of 3.
103
20
143 kV/m at 1 g
143 kV/m at 0 g
205 kV/m at 1 g
205 kV/m at 0 g
286 kV/m at 1 g
286 kV/m at 0g
Height (mm)
15
10
5
0
0
2000
4000
6000
8000
10000
12000
3
Particle Number Density (#/cm )
Figure 4.16. Particle concentrations of 1789 mg Gl (53-63 μ m) in 20 mm test cell.
A stratification effect is apparent at the lower electric field intensity (E=143, 205
kV/m) in normal gravity. As the electric field intensity increases, stratification diminishes,
producing more uniform concentration profiles, but with increased profile data scatter (e.g.,
E=286 kV/m). The scatter is related to a new suspension instability phenomena occurring in
microgravity. Successive pictures of glass powder suspension in 20 mm test cell during the
drop are presented in Figure 4.17.
A new EPS cloud instability phenomenon for glass bead particles (E=286 kV/m,
H=20 mm, m=1789 mg, V=6 kV) observed in microgravity is shown in Figure 4.17, captured
by the onboard video camera running at 30 frames/sec (Colver, 2007; Xu et al., 2008). The
drop release time is used as a starting (zero) reference condition (e.g., -0.3 seconds before
drop). The expanded laser sheet beam in Figure 4.17 appears as red scattered light from the
104
suspension with brighter regions indicating increased particle concentration (e.g., compare
before and during drop). Four successive time frames at – 0.14, 0.0, 0.54 and 1.5 seconds
show the transition from a visually uniform low concentration suspension (1 g) to a high
concentration with the development of periodic cell structures and then column structures in
microgravity. The instability also explains the concentration scatter in Figure 4.16 at high
electric field intensity in 0 g. Similar instabilities have been observed in normal gravity only
at very high electric field intensity necessary to overwhelm gravitation forces.
-0.14 Seconds (1 g)
Uniform (low) concentration
0.00 Seconds (Drop) (0 g)
Uniform (increasing) concentration
0.54 Seconds (0 g)
1.50 Seconds (0 g)
Periodic cell structure (high) concentration
Column structure (high) concentration
Figure 4.17. Microgravity videos of Gl (53-63 μ m).
105
Due to alignment problems with the laser beam and a malfunction of one analog input
channel of the TattleTale onboard computer affecting the photodiode array measurements,
concentration profiles are not available for copper powder. Example pictures of the transition
recorded by onboard video camera are shown in Figure 4.18. The copper powder (63-74 μ m)
is shown inside a 20 mm height of test cell at a suspension voltage of 8 kV. A large loading
mass of 6818 mg is tested to show any effect of gravity. The suspensions produced are
visually uniform.
Successive frames (-0.02, 0, 0.02, 0.07, 0.44, and 0.58 s) in Figure 4.18 show the
transition from normal gravity to microgravity. It is seen that additional powder is suspended
when the drop begins, as the laser sheet becomes brighter from the additional particles.
Shortly into the drop, the particle suspension continues developing (0.02, 0.07, and 0.44 s)
even blocking the laser sheet so only a black area is visible on the right side of the frame.
This explains the laser attenuation method has upper concentration limitations. Additional
suspension growth can be observed up to about 0.58 s, when the beam becomes substantially
blocked. This sequence clearly demonstrates that reduced gravity results in greatly increased
concentrations in particle suspensions.
106
-0.02 Seconds (1 g)
Low concentration
0.00 Seconds (Drop) (0 g)
Increasing concentration
0.02 Seconds (0 g)
Increasing concentration
0.07 Seconds (0 g)
Increasing concentration
0.44 Seconds (0 g)
0.58 Seconds (0 g)
Increasing concentration
Increasing (max) concentration
Figure 4.18. Microgravity videos of Copper (63-74 μ m).
107
4.3 Particle Concentration Stratification Criteria
The following attributes of suspension profiles are used to help identify desirable and
undesirable conditions for combustion testing (Colver, 2007): The ideal uniform suspension
(desirable) is one in which the value of the particle concentration profile is a constant value
across the test chamber (top to bottom). Stratification is the (undesirable) asymmetry in the
particle concentration profile arising from the presence of gravity.
Certain stratification criteria can be applied to analyze the uniformity of the
suspended particle cloud: (1) Small particle force ratio Fg / FE (gravitational force and
electrostatic force) reduces the segregation between charge-neutral particles and particles
with charge. (2) Small mean free path ratio λp/ λw (particle-particle and particle-wall) reduces
the numbers of charge-neutral particles produced by collisions compared to charged particles
produced at walls. (3) Large value in coefficient of restitution (~1) between particles and
electrode walls (top and bottom) supports the generation of a constant velocity profile.
The first of above criteria states that stratification is reduced when gravity forces Fg
acting on charge-neutral particles are small compared to electrostatic forces FE acting on
charged particles. The calculation equation can be written as (Colver, 2007; Colver et al.,
2008):
Fg
FE
=
ρ s dg
π 2εE 2
(4.3)
where ρ s is the density of the solid particle, d is the particle diameter, ε is the fluid
permittivity, g is the acceleration of gravity, and E is the electric field intensity.
The second criterion states that stratification is reduced if collision processes produce
fewer numbers of charge-neutral particles compared to charged particles. The relative
108
collision probability is determined as the ratio of the mean free path for a particle-wall λw
collision being significantly smaller than the mean free path for a particle-particle collision λp.
The calculation equation can be written as
λw
= nπ d 2 h
λp
(4.4)
where n is the suspension number density, d is the particle diameter, and h is the plate
electrode separation distance.
Equations (4.3) and (4.4) can be combined into a single criteria test for stratification if
it is assumed that the two criteria behave independently. The combined calculation equation
can be written as (Colver et al., 2008):
⎡ Fg λ w ⎤
⎛ Fg ⎞
⎟⎟
F⎢ ,
⎥ = f ⎜⎜
F
F
λ
⎝ E⎠
⎣⎢ E p ⎥⎦
⎛ λw
g⎜
⎜ λp
⎝
⎞
⎟
⎟
⎠
(4.5)
the functions f(x) and g(x) in Eq. (4.5) are not determined.
The particle suspension concentration under normal and microgravity are presented in
the following. All results can be analyzed with the stratification criteria above. The
stratification criteria can also be used to predict the particle concentration stratification. If the
calculation results for Eqs. (4.3) and (4.4) are much less than 1, the stratification is reduced.
Check the stratification criteria with particle concentration profile measured by light
attenuation method. First, the calculation example is the aluminum powder concentration,
presented in Figure 4.1. The material density is about 2.7 g/cm3 with particle diameter d 32 of
23.72 μ m and the permittivity for air is about 8.85 × 10 −12 F/m. The height of the test cell is
about 39 mm. The number density varies with the electric field intensity E before it reaches
the minimum electric field intensity for complete suspension. The actual number density can
109
be determined by integrating the particle concentration profile presented before. The average
number density of aluminum particle suspension at 400 and 700 kV/m are 2792 #/cm3 and
25015 #/cm3, respectively. For criteria Eq. (4.3), the calculations are 0.043 and 0.014 for 400
and 700 kV/m electric field intensity. Both satisfy the criteria 1. However, the result for Eq.
(4.4) is 0.17 for 400 kV/m, while the result for 700 kV/m is 1.55. This does not satisfy the
criteria 2. From the calculations, it can be determined that the aluminum suspension
stratification is present at the electric field intensity of 700 kV/m, but not present at the
electric field intensity of 400 kV/m. This conclusion matches the particle concentration
profile presented in Figure 4.1.
Another numerical example is copper particles with a material density of 8.92 g/cm3.
The height of the test cell is still 39 mm. The calculation is completed to compare the results
for electric field intensity of 400 and 900 kV/m, respectively. The copper particle
concentration profiles are presented in Figure 4.3. The average number density of suspended
copper is 135 #/cm3, and 6055 #/cm3 for 400 and 900 kV/m, respectively. Both results (0.39
and 0.077) from Eq. (4.3) are far less than 1, which satisfy the criteria 1. The calculation
result of Eq. (4.4) for 900 kV/m is 2.4, which doesn’t satisfy criteria 2. On the other hand, the
calculation result for 400 kV/m is only 0.053, less than 1. So, it can be predicted that the
stratification should be present at 900 kV/m electric field intensity, but not present at 400
kV/m electric field intensity. This conclusion is also confirmed by the particle concentration
profile presented in Figure 4.3.
These two numerical examples show the proposed stratification criteria are useful to
determine and predict if stratification occurs. However, the stratification criteria cannot
predict all conditions successfully. One reason is the criteria Eq. (4.4) requires information
110
on the average suspension particle concentration, often unknown or difficult to measure. A
common assumption is that all particles will be suspended inside the test cell so the average
number density can be calculated—unfortunately this assumption is only true for some cases.
From the previous analysis of minimum electric field intensity requirements for a complete
suspension, for example, Figures.4.10 and 4.11, it is obvious that critical electric field
intensity can only sustain a limiting mass of particles. It is desirable to know the amount of
power in suspension at specified electric field intensity.
It is noted that the light attenuation method can provide the required information on
particle suspension concentrations so the average number density can be integrated, based on
using the particle concentration profile, for example, Figures. 4.5 to 4.10. Unfortunately, the
light attenuation method does not work well at high mass loadings as was observed
previously.
In a summary, the proposed stratification criteria provide a qualitative prediction of
the suspension stratification, but may not be able to predict the stratifications due to the
unknown particle suspension number density. It is desirable to establish the relationship
between the electric field intensity and maximum loading mass of particle for suspension,
which will be introduced in Chapter 5, using the external current measurement method.
4.4 Summary
In this chapter, EPS generated measurements of particle concentration profiles have
been reported for a variety of particles using the light attenuation method in both normal and
microgravity environments. In general, particle concentrations show uniform profiles at
relatively low electric field intensity at low particle concentrations. With the addition of
111
particles, stratification is observed at 1 g, which is undesirable for uniform clouds in
combustion testing.
A curve of average number density (i.e., total powder suspended) versus electric field
intensity can be obtained by integrating the particle concentration profiles. The average
number density increases with the increased electric field intensity at small field intensities.
With increasing field intensity, the average number density reaches a plateau that indicates
the maximum-of-mass condition for a suspension is achieved. From a family of such number
density curves, one can determine the relationship between the minimum electric field
intensity and the (maximum possible) loading mass.
As predicted, higher particle concentrations are possible using EPS in microgravity.
Particle concentration profiles observed in microgravity are symmetric rather than stratified
due to the absence of gravity. New phenomena in the form of column structure suspension
are observed in drop tower videos, indicating suspension instability. Videos of drop tower
experiments also show penetration of the laser sheet is blocked entirely by the heavy particle
suspension, making the light attenuation method ineffective.
Tentative suspension stratification criteria are proposed. However, the criterion is
useful only if the suspension number density is known.
112
CHAPTER 5. MAXIMUM SUSPENSION CONCENTRATION
Two questions arise pertaining to the maximum possible suspension concentration of
powder dispersions for a given electric field intensity using EPS (Colver, 2007): (1) Will
100% of the powder be suspended whatever the sample size and E field? (2) What happens to
excess powder if not all of the powders is suspended? It might seem intuitively correct that
any amount of powder can simply be suspended or “floated” in microgravity, whatever the
electric field intensity. However, this is not seen in the drop tests. Videos taken over drop test
times of about 2 seconds do not support the idea of a freely floating suspension for either fine
glass, aluminum powders, or heavier copper spheres during EPS formation with weak
electric fields. Rather, the videos show a significant increase in the amount of powder
suspended with increasing electric field intensity for all of the powders tested (copper, glass,
aluminum).
It can be argued the cohesive nature of powders from van der Waals and electronic
forces act to retain powders on walls unless driven into suspension by the electric field; i.e.,
the initial condition of the powder on walls or floating determines its subsequent behavior in
microgravity. Furthermore, the application of an electric field at 1 g during startup—prior to
each drop—as well as during a suspension produces a current flow through the bulk (bed)
powder with resultant particle-particle forces in the packed bed that continues after the
suspension is generated (Colver, 1980). Similar powder adhesion forces are observed in
electrostatic precipitators (back ionization phenomena) and in packed and electro-fluidized
beds that retains powder strings and coating on electrodes (White, 1963; Colver, 2000).
113
From ground-based studies it is known that EPS suspensions are generated by electric
fields of sufficient strength to lift particles against gravity, other attractions, and adhesion
and/or cohesion forces. Each value of the electric field intensity presents an upper limit to the
particle concentration. In an attempt to correlate maximum concentration in both normal and
microgravity, a concept of excess electric field intensity is introduced. The excess electric
field intensity is the additional electric field intensity to overcome gravity and induced forces
in forming suspension.
Three different methods of determining the maximum particle suspension
concentration were introduced in Chapter 4. The light attenuation method has the advantage
of giving detailed information about the distribution of suspension particle concentrations at
different locations, while limited by the corrected light extinction coefficient and light
scattering by particles. As noted above, light attenuation cannot be used if there is total
blockage from heavy particle suspensions. An alternative method is to measure the external
current induced by the collision between charged particles with the upper plate. Assuming
the collision relaxation time is the same, the induced current correlates directly to the
concentration of particles through particle collisions with electrodes. At a given electric field
intensity, the flat plateau of the current profile suggests the suspension of certain particle
reaches the maximum concentration value.
In this chapter, the results of maximum particle concentration at a given electric field
intensity using the external current measurements are initially presented. Next the concept of
excess electric field intensity is introduced and will be used to correct the data in
microgravity in Chapter 6. Finally, a correlation for the excess electric field intensity will be
discussed.
114
5.1 Maximum Suspension Concentration by Current Measurement in
Normal Gravity
This method utilizes the current measured in the external circuit of an EPS cell, which
avoids the problematic issues arising from laser scattering by particles at high concentrations
(Colver, 2007). In the current method, a guard-ring electrode is substituted as part of the
ground side electrode in Figure 3.1. The diameter of the guard-ring electrode exposed to
charged particles is about 3.5 cm. Subsequently, a current is generated in the external circuit
corresponding to the steady-state particle concentration inside the EPS test cells (Colver,
1976). When the powder loading increases, there is a corresponding current increase as well.
When a maximum-of-mass suspension is reached, the generated current will be stead-steady
in time.
Different particles are tested using this method varying the electrode separation
distances at 1.1, 1.4, 1.7 and 2 cm. A family of curves of current and loading mass at
specified electric intensities are obtained giving a relationship between the electric field
intensity for a maximum suspension, or alternatively the maximum suspended mass.
The current profiles shown in Figure. 5.1 are for aluminum (20-25 μ m) powder at 1.1, 1.4,
1.7, and 2.0 cm separations. All curves of current (at constant electric field intensity) show
similar trends of a near-linear increase in current at a lower loading mass followed by
constant current with increased loading mass. This suggests that a maximum suspension is
reached at the constant current condition. These maximum current levels increase with
increased electric field intensities because more powders can be suspended with increased
electric field intensity. The maximum current increases from 3.4 μ A to 8.5 μ A at 1.1 cm
115
separation. The maximum current value of 9.7 μ A is obtained at the 1.7 cm separation and
20 kV suspension voltage, i.e., an electric field intensity of 1081 kV/m. By combining these
current profiles at different separations and suspension voltages, a relationship between
maximum suspension particle mass and electric field intensity can be established (Colver,
2007). It is noteworthy that by using aluminum particles (25-30 μ m), instability in the
particle suspension occurs before a steady state current is reached (not shown here).
9
13 kV
11 kV
9 kV
8
12 kV
10 kV
8
7
6
16 kV
14 kV
12 kV
15 kV
13 kV
6
5
4
4
2
3
H=11 mm
Current (μA)
0.0
0.1
0.2
0.3
20 kV
18 kV
16 kV
19 kV
17 kV
15 kV
0.4
0.5
7
6
5
4
3
2
1
H= 17 mm
0.3
0.4
0.5
0.6
H=14 mm
2
0
0.7
1
0.1
10
9
8
7
6
5
4
3
2
1
0.80.2
0.2
21 kV
19 kV
17 kV
0.3
0.4
0.5
0.6
0.7
20 kV
18 kV
16 kV
H=2.0 mm
0.3
0.4
0.5
0.6
0.7
Loading Mass (gm)
Figure 5.1. Induced currents of Al (20-25 μ m) at given voltages and heights - 1 g.
0.8
116
A series of current profiles for copper particles is show in Figures 5.2 – 5.4. Three
different diameters of copper powders are tested to examine the effect of particle diameter on
electrical charging and suspension. For the copper powder (30-38 μ m) shown in Figure 5.2,
the maximum current is about 0.64 μ A at 1.1, 1.4, and 1.7 cm, and 0.47 μ A at 2.0 cm. For
the larger copper powder with diameter 44-53 μ m shown in Figure 5.3, the maximum
current is about 0.59 μ A at 1.1 cm, 13 kV (an electric field intensity of 1040 kV/m). For a
2.0 cm electrode separation distance, the maximum current is about 4.6 μ A at 20kV (an
electric field intensity of 930 kV/m).
0.7
13 kV
11 kV
9 kV
Current (μA)
0.6
12 kV
10 kV
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.2
0.6
H=11 mm
0.4
19 kV
17 kV
15 kV
0.6
16 kV
14 kV
12 kV
0.6
0.8
1.0
1.2
18 kV
16 kV
14 kV
0.1
0.5
0.4
0.5
0.4
15 kV
13 kV
11 kV
H=14 mm
0.4
0.6
20 kV
18 kV
16 kV
0.8
1.0
1.2
1.4
19 kV
17 kV
0.3
0.3
0.2
0.2
0.1
H=17 mm
0.4 0.6 0.8 1.0 1.2 1.4 1.6
0.1
H=20 mm
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Loading Mass (gm)
Figure 5.2. Induced currents of Cu (30-38 μ m) at given voltages and heights - 1g.
117
Current (μA)
0.6
13 kV
11 kV
9 kV
0.6
12 kV
10 kV
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.2
0.6
0.5
H=11 mm
0.4
0.6
19 kV
17 kV
14 kV
18 kV
16 kV
13 kV
0.8
16 kV
14 kV
12 kV
15 kV
13 kV
11 kV
H=14 mm
0.1
1.0
1.2
0.5
0.4
0.4
0.6
20 kV
18 kV
16 kV
0.8
1.0
1.2
1.4
19 kV
17 kV
0.4
0.3
0.3
0.2
0.2
0.1
H=17 mm
0.4
0.6
0.8
1.0
1.2
1.4
H=20 mm
0.1
1.6
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0
Loading Mass (gm)
Figure 5.3. Induced currents of Cu (44-53 μ m) at given voltages and heights - 1g.
Note there is a peak value in the current for both the 1.7 and 2.0 cm separation in
Figure5.3, which is not observed for the 1.1 cm and 1.4 cm electrode plate separations. A
possible reason is that when more particles are put into the test cell, a layer of particle is
deposited at the bottom electrode, so the suspended particles collide with stationary particles
rather than a metal electrode. This affects the particle suspension, then the induced current
drops slightly. This current drop may not be shown in other tests, due to test resolution.
For the large diameter copper powder (63-74 μ m) shown in Figure 5.4, the
maximum current is about 0.64 μ A at 1.1cm, 13 kV (an electric field intensity of 1040
118
kV/m). A similar current drop is also observed. A comparison of current profiles for copper
powder shows that the difference between various diameters is not that significant at given
conditions. This suggests that particle diameter may not be a dominant factor for achieving a
maximum particle suspension concentration.
0.7
13 kV
11 kV
9 kV
0.6
0.7
12 kV
10 kV
0.6
16 kV
14 kV
12 kV
15 kV
13 kV
11 kV
0.5
0.5
0.4
0.4
0.3
Current (μA)
0.3
0.2
0.2
0.1
0.2
0.6
0.1
H=11 mm
0.4
19 kV
17 kV
15 kV
0.6
0.8
1.0
1.2
18 kV
16 kV
14 kV
0.4
0.5
0.5
H= 14 mm
0.6
20 kV
18 kV
16 kV
0.8
1.0
1.2
1.4
19 kV
17 kV
0.4
0.4
0.3
0.3
0.2
0.2
H=17 mm
0.1
0.4
0.6
0.8
1.0
1.2
H=20 mm
0.1
1.4
1.6
0.6
0.8
1.0
1.2
1.4
1.6
1.8
Loading Mass (gm)
Figure 5.4. Induced currents of Cu (63-74 μ m) at given voltages and heights - 1g.
All copper particle current profiles increase at lower loading mass and then reach a
stable value at the maximum-of-mass conditions. All current profiles show similar trends as
follows: (1) the current increases with the increased suspension voltage for the same loading
119
mass, which suggests that the oscillating velocity of the particle changes with an increase in
the electric field (Eq. (2.6)); and (2) at a given electric field intensity, the current increases
with the increase of loading mass, suggesting the maximum of mass condition has not been
reached. The increase in current is a result of increased particle suspension concentration
prior to reaching the saturation condition for current. The effect of oscillating velocity of
particles at this condition is negligible because the electric field intensity remains the same.
Glass particles of different diameters (25-30 μ m and 53-63 μ m) are also tested in
normal gravity. The results are presented in Figures 5.5 and 5.6.
2.5
13 kV
11 kV
9 kV
2.5
12 kV
10 kV
1.5
Current (μA)
1.5
1.0
1.0
H=11 mm
0.5
0.1
2.0
15 kV
13 kV
11 kV
2.0
2.0
2.5
16 kV
14 kV
12 kV
0.2
19 kV
17 kV
15 kV
0.3
0.4
0.5
0.5
0.6
0.7
0.2
2.0
18 kV
16 kV
14 kV
H=14 mm
20 kV
18 kV
16 kV
0.4
0.6
0.8
19 kV
17 kV
1.5
1.5
1.0
1.0
0.5
H=17 mm
0.2
0.4
0.6
0.8
H=20 mm
0.5
1.0
0.2
0.4
0.6
0.8
1.0
Loading Mass (gm)
Figure 5.5. Induced currents of Gl (25-30 μ m) at given voltages and heights – 1 g.
120
1.4
13 kV
11 kV
9 kV
1.2
12 kV
10 kV
16 kV
14 kV
12 kV
1.4
15 kV
13 kV
11 kV
1.2
1.0
1.0
0.8
0.8
Current (μA)
0.6
0.6
0.4
H=11 mm
0.2
0.1
1.4
1.2
0.2
19 kV
17 kV
15 kV
0.3
0.4
0.5
H=14 mm
0.4
0.6
0.2
20 kV
18 kV
16 kV
1.2
18 kV
16 kV
14 kV
0.3
0.4
0.5
0.6
0.7
0.8
19 kV
17 kV
1.0
1.0
0.8
0.8
0.6
0.6
0.4
0.2
0.4
0.6
H=20 mm
0.4
H=17 mm
0.2
0.8
0.2
0.4
0.6
0.8
1.0
Loading Mass (gm)
Figure 5.6 Induced currents of Gl (53-63 μ m) at given voltages and heights - 1 g.
For the smaller diameter shown in Figure 5.5, the maximum current is about 2.5 μ A
at 1.1cm, 13 kV (an electric field intensity of 1040 kV/m). At 2.0 cm separation, the
maximum current is about 2.05 μ A at an electric field intensity of 930 kV/m (20 kV). For
the larger diameter shown in Figure 5.6, the maximum current is about 1.47 μ A at 1.1 cm,
13 kV (an electric field intensity of 1040 kV/m). At 2.0 cm separation, the maximum current
is about 1.25 μ A at electric field intensity of 930 kV/m (20 kV). Unlike copper powder, the
glass particles show a clear dependence on particle diameter.
121
The results of other powders—magnesium, titanium and iron— are presented in
Figures 5.7-5.9, respectively. All three particles have a size range between 0-44 μ m. The
detailed size distributions were presented in a previous chapter. In Figure 5.7, the current
profile of magnesium shows the maximum current of 3.9 μ A at 1.1 cm, 13 kV (an electric
field intensity of 1040 kV/m) without a current peak.
Current (μA)
3.5
13 kV
11 kV
9 kV
12 kV
10 kV
3.5
3.0
3.0
2.5
2.5
2.0
2.0
1.5
1.5
1.0
0.1
3.5
3.0
19 kV
17 kV
15 kV
0.2
15 kV
13 kV
11 kV
1.0
H=11 mm
0.5
16 kV
14 kV
12 kV
0.3
3.0
18 kV
16 kV
14 kV
2.5
2.5
H=14 mm
0.1
20 kV
18 kV
16 kV
0.2
0.3
0.4
19 kV
17 kV
2.0
2.0
1.5
1.5
1.0
1.0
H=20 mm
H=17 mm
0.5
0.1
0.2
0.3
0.4
0.1
0.2
0.3
0.4
0.5
Loading Mass (gm)
Figure 5.7. Induced currents of Mg (0-44 μ m) at given voltages and heights - 1 g.
122
The current profile of titanium, in Figure 5.8, shows a maximum current of 5.1 μ A at
1.1 cm, 13 kV (an electric field intensity of 1040 kV/m), with a clear current drop at an
electrode plate separation distance of 1.7 cm.
5
5
H=11 mm
H=14 mm
4
4
3
3
Current (μA)
2
2
13 kV
11 kV
9 kV
1
5
0.1
0.2
12 kV
10 kv
0.3
16 kV
14 kV
12 kV
1
0.4
0.1
4
H=17 mm
4
0.2
0.3
15 kV
13 kV
11 kV
0.4
0.5
0.6
H=20 mm
3
3
2
2
19 kV
17 kV
15 kV
1
0.1
0.2
0.3
0.4
18 kV
16 kV
14 kV
0.5
20 kV
18 kV
16 kV
1
0.6
0.1
0.2
0.3
0.4
19 kV
17 kV
0.5
0.6
0.7
Loading Mass (gm)
Figure 5.8. Induced currents of Ti (0-44 μ m) at given voltages and heights - 1 g.
The current profile of iron in Figure 5.9 has a maximum current of 3.6 μ A at 1.1 cm,
13 kV (an electric field strength of 1040 kV/m), approaching a constant value without a peak
current. The three powders have similar maximum currents at the same electrode spacing and
applied voltage.
123
H=11 mm
H=14 mm
3
3
2
Current (μA)
2
13 kV
11 kV
9 kV
1
0.2
12 kV
10 kV
16 kV
14 kV
12 kV
1
0.4
0.6
0.8
0.2
3
H=17mm
0.4
15 kV
13 kV
11 kV
0.6
0.8
1.0
H=20 mm
3
2
2
19 kV
17 kV
15 kV
0.4
0.6
0.8
18 kV
16 kV
14 kV
1.0
20 kV
18 kV
16 kV
1
1.2
0.4
0.6
0.8
19 kV
17 kV
1.0
1.2
Loading Mass (gm)
Figure 5.9. Induced currents of Fe (0-44 μ m) at given voltages and heights - 1 g.
5.2 Excess Electric Field Intensity
In an attempt to correlate maximum concentrations in both normal gravity and
microgravity fields, the concept of “excess electric field intensity” has been introduced
(Colver, 2007; Colver et al., 2008). The excess field is defined as the additional electric field
intensity to overcome gravity and induce electrical attraction forces into forming a
suspension. The gravitational, particle-particle, and particle-wall forces are included in the
concept of excess electric field intensity.
124
To simplify the analysis, the force schematic of a single particle suspended in electric
field is shown in Figure 2.1. A force balance for this single particle can be written by
∑F
P
= FE − Fg − Fad
(5.1)
where FE is the electrostatic force, which will drive the particle against gravitational and
other induced forces to levitate the particle in the air. Fg is the gravitational force. Fad is the
sum of the other induced particle-particle and particle-wall, which will keep particles from
levitating.
The electrostatic force, FE , due to the interaction of induced charge of particle with
the external electric field, can be defined as follows:
FE = 4πεa 2 E 2 k1
(5.2)
where ε is the permittivity of the medium (air for our case), which has the magnitude of
ε = 136π × 10 −9 ( F / m)
a is the radius of a particle, E is the far field electric field intensity, and k1 is a constant
accounting for geometric surface effects (partially exposed particle in bed). It was
experimentally determined as 1.37 for isolated spherical particles by Colver (1976).
For a particle inside an electric field, the induced Maxwell type of induced surface
charge q surface can be written in the form of
q surface = 4πεa 2 Ek 2
(5.3)
where k2 is a constant accounting for the geometric surface effect on the particle charge. It
was determined as 1.64 for an isolated spherical particle by Colver (1976).
125
To normalize both normal gravity and microgravity conditions, one defines the excess
electric field intensity to incorporate the gravity effect to handle both normal and
microgravity data. Equation (5.1) can be rewritten as
∑F = q
surface
ΔE excess − Fad
(5.4)
where the excess electric field intensity ΔEexcess is defined, including image forces and
Maxwell type charging for particles at the surface through added surface effect constants k1
and k2, respectively, as (Colver, 2007; Colver et al., 2008)
ΔEexcess =
FE − Fg
q surface
=
ρ pd p g
k1
E−
k2
6k 2εE
(5.5)
For a transient lift-off condition of the particle, the total force will be larger than zero,
so the criterion to form a suspension in normal or microgravity follows Eq. (5.5) as
ΔEexcess −
Fad
q surface
≥0
(5.6)
The additional force, Fad , accounts for packed bed forces is only significant for
smaller particle diameters. For large particles, Fad can be ignored in comparison to gravity
forces, so that particle lift-off criterion could become ΔEexcess ≥ 0 . This implies that larger
particles can float off during microgravity, while fine particles are held by cohesion/adhesion
forces as layers on walls supporting our observations from drop tests (Colver, 2007).
To evaluate the effect of the excess electric field intensity, calculate the ratio of
excess electric field intensity to the corresponding electric field intensity as
α=
k1 ρ p d p g
−
k 2 6 k 2 εE 2
(5.7)
126
Assuming the additional force Fad is negligible, Eq. (5.7) can be used to predict the
minimum nominal electric field intensity for particle lift-off. An example calculation for
Excess Electric Field Intensity Ratio
copper particles is presented in Figure 5.10 for normal gravity.
1
0
-1
-2
Cu 30-38 μm
Cu 44-53 μm
Cu 63-74 μm
-3
-4
-5
-6
0
200
400
600
800
1000
Electric Field Intensity (kV/m)
Figure 5.10. Excess Electric Field Intensity Ratio α - 1 g.
Three different diameters of copper particles were used to examine the effect of
diameter on excess electric field intensity. The electric field intensity varies from 100 to 1000
kV/m. The curves in Figure 5.10 become constant, ~ 0.8, when the electric filed intensity
becomes large, ~1000 kV/m for all diameters of copper particles. The ratio decreases with
decreasing electric field intensity, while the difference between curves was more apparent.
Applying the particle lift-off criterion shown in Eq. (5.6), the electric field intensity at
ratio=0 represents the particle lift-off electric field intensity. From the excess electric field
127
intensity ratio profiles, it is obvious that larger particles require more electric field intensity.
The predicted values are compared with experimental particle lift-off electric field intensity
in Table 5.1.
Table 5.1. Comparison of minimum particle lift-off electric field intensity
Minimum E
Cu (44-53 μ m)
Cu (63-74 μ m)
Cu (30-38 μ m)
Experimental Emin (kV/m)
Predicted Emin (kV/m)
256-384
246-440
260-472
210
245
279
The experimental data in Table 5.1 are carried out in external current measurement
tests. A laser beam is passes through the middle of the EPS test cell during the experiment.
The electric field is increased until the particle lifts off and is made visible by the beam.
From Table 5.1 it can be concluded that the predicted minimum particle lift-off electric field
intensity is within the range of experimental results, except for the smaller copper powders
where adhesion force, Fad , is no longer negligible for fine powders.
Since the excess electric field intensity is independent of the value of gravity, it might
be expected to correlate the maximum particle concentration for a given electric field
intensity, i.e., as in normal gravity. In some of the drop tower experiments, large quantities of
aluminum powder were tested to generate very high values of particle concentration for rich
limit ignition tests. The onboard photodiode data was used to calculate the suspension
concentrations prior to and during the drop. Using aluminum (20-25 μ m) drop tower data
(presented in Figures 4.13 and 4.14) from the laser-photodiode in both normal and
128
microgravity together with Eq. (3.1) to calculate the powder concentration, a plot of the
3
Particle Number Density *#/cm )
particle concentration against ΔEexcess is shown in Figure 5.11.
40000
Micro-gravity
Normal-gravity
Linear Fit
30000
20000
95% confidence interval of fit
10000
0
200
300
400
500
600
Excess Electric Intensity (kV/m)
Figure 5.11. Maximum Particle Concentration with Excess Electric Field Intensity.
Both normal gravity and microgravity particle concentration values fall along a
straight line, which suggests the concept of excess electric field intensity properly accounts
for gravity in this case. A linear regression with 95% confidence interval gives
N = 94.34 * ΔE excess − 1.68 × 10 4 ± 2.78 × 10 3 (95% confidence)
(5.8)
where N is the particle suspension number density (#/cm^3) and ΔEexcess is the excess electric
field intensity (kV/m). The correlation coefficient of this linear regression R2=0.98. The
details for limiting the significant figures and curve fitting confidence intervals are discussed
in appendix.
129
5.3 Maximum Particle Concentration Correlation
All current profiles presented above show a similar trend: the current increases almost
linearly at a lower loading mass (given electric field intensity) then reaches a peak current
corresponding to the maximum particle concentration. It can be concluded that the
suspension experiences two stages: a rising stage—a steady state stage. Using linear
regression for the rising stage of the current profile and setting the steady stage peak current
as a target value, the maximum particle mass concentration at a given suspension voltage and
test cell separation can be obtained. The detailed procedure is presented elsewhere (Colver,
2007). To normalize the results, the electric field intensity is used to account for the effect of
suspension voltage and test cell separation. Also to normalize the effect of loading mass of
particle, test cell diameter and the test cell separation, the particle mass concentration is used.
The results of maximum particle mass concentration versus electric field intensity are
presented in Figures 5.12-5.20 for various particles. A linear regression with 95% confidence
interval for each type of particle is presented in each graph as well. The plot of the maximum
aluminum particle mass concentration with respect to electric field intensity is presented in
Figure 5.12. A linear regression gives a good fit to the experimental data with a correlation
coefficient of 0.84. Other regressions are found in a similar way.
The linear regressions for three copper powders in Figures 5.13-5.15 are linear with a
correlation coefficient of 0.96 for the smallest diameter (30-38 μ m). The remaining two
copper powders of larger diameter show increased scatter with correlation coefficients of
0.56 and 0.53, respectively. For the glass bead powders (Figures 5.16 and 5.17), the smaller
diameter glass bead powder has an improved correlation coefficient (R2=0.74) compared to
130
the larger diameter (R2=0.47). Three other metal powders (Figures 5.18-5.20) also show
linear trends and good correlations, except iron. It has correlation coefficient of 0.49.
Efforts have been made to correlate the maximum suspended mass for a variety of
particles with electric field intensity or excess electric intensity. However, this has not proven
successful. Only linear relations with 95% confidence intervals for different types of particles
are listed below in Table 5.2. The details for limiting significant figures and curve fit
3
Particle Mass Concentraion (gm/m )
confidence interval calculations are discussed in appendix.
6500
Experiment Data
Linear Fit
6000
5500
5000
95% confidence interval of fit
4500
700
750
800
850
900
950
1000
1050
1100
Electric Field Intensity (kV/m)
Figure 5.12. Max particle concentration at given E for Al (20-25 μ m) - 1 g.
3
Particle Mass Concentration (gm/m )
131
15000
Experiment Data
Linear Fit
14000
13000
12000
11000
95% confidence interval of fit
10000
700
800
900
1000
Electric Field Intensity (kV/m)
3
Particle Mass Concentration (gm/m )
Figure 5.13. Max particle concentration at given E for Cu (30-38 μ m) - 1 g.
15000
Experiment Data
Linear Fit
14000
13000
12000
95% confidence interval of fit
11000
700
800
900
1000
Electric Field Intensity (kV/m)
Figure 5.14. Max particle concentration at given E for Cu (44-53 μ m) - 1 g.
3
Particle Mass Concentration (gm/m )
132
Experiment Data
Linear Fit
13000
12000
95% confidence interval of fit
11000
700
800
900
1000
Electric Field Intensity (kV/m)
3
Particle Mass Concentration (gm/m )
Figure 5.15. Max particle concentration at given E for Cu (63-74 μ m) - 1 g.
8000
Experiment Data
Linear Fit
7000
95% confidence interval of fit
6000
5000
700
800
900
1000
Electric Field Intensity (kV/m)
Figure 5.16. Max particle concentration at given E for Gl (25-30 μ m) - 1 g.
3
Particle Mass Concentration (gm/m )
133
8500
Experiment Data
Linear Fit
8000
7500
95% confidence interval of fit
7000
6500
700
800
900
1000
Electric Field Intensity (kV/m)
3
Particle Mass Concentration (gm/m )
Figure 5.17. Max particle concentration at given E for Gl (53-63 μ m) - 1 g.
Experiment Data
Linear Fit
4000
3000
95% confidence
interval of fit
2000
700
800
900
1000
Electric Field Intensity (kV/m)
Figure 5.18. Max particle concentration at given E for Mg (0-44 μ m) - 1 g.
3
Particle Mass Concentraion (gm/m )
134
5000
Experiment Data
Linear Fit
4500
4000
95% confidence interval of fit
3500
3000
700
800
900
1000
Electric Field Intensity (kV/m)
3
Particle Mass Concentration (gm/m )
Figure 5.19. Max particle concentration at given E for Ti (0-44 μ m) - 1 g.
10000
Experiment Data
Linear Fit
9000
8000
95% confidence interval of fit
7000
6000
5000
700
800
900
Electric Field Intensity (kV/m)
1000
Figure 5.20. Max particle concentration at given E for Fe (0-44 μ m) - 1 g.
135
Table 5.2. Linear Regression Equations (% confidence interval)
R2
Particle
Linear Regression
Al (20-25 μ m)
C = 4.67 * E + 1436.00 ± 93.69 (95%)
0.84
Cu (30-38 μ m)
C = 11.65 * E + 2344.20 ± 107.25 (95%)
0.96
Cu (44-53 μ m)
C = 5.72 * E + 7649.70 ± 234.66 (95%)
0.56
Cu (63-74 μ m)
C = 4.23 * E + 8404.40 ± 184.58 (95%)
0.53
Gl (25-30 μ m)
C = 6.42 * E + 1286.20 ± 177.13 (95%)
0.74
Gl (53-63 μ m)
C = 2.67 * E + 5165.30 ± 131.49 (95%)
0.47
Mg (0-44 μ m)
C = 5.20 * E − 1351.10 ± 147.05 (95%)
0.73
Ti (0-44 μ m)
C = 4.16 * E + 349.59 ± 73.99 (95%)
0.87
Fe (0-44 μ m)
C = 8.47 * E + 891.36 ± 403.37 (95%)
0.49
The normalized forms of maximum particle suspension concentration expressed as
mass concentration (g/m3) and electric field intensity (V/m) are shown in Figure 5.21 (Colver,
2007). It is apparent the photodiode drop tower data are deficit in predicting the maximum of
mass concentration by a factor of 5 (or greater) below ΔEexcess ≈ 106 V/m compared to the
ground-based current method. It is not clear which one is the better evaluation. In the drop
tower experiments, the particle concentration is calculated by the laser intensity ratio.
However, the laser intensity reading from the photodiode could be affected significantly by
the heavy suspension cloud, i.e., the assumed value of 2 for the light extinction coefficient
could be different. In the ground-based current test, it is sometimes observed that column
structures develop so the cloud is not uniform. There is also a chance some particles stick to
the bottom electrode or to the glass ring during suspension formation. Consequently, the
136
maximum particle mass concentration from ground-based current method may be overpredicted.
Maximum Mass Concentration
MAXIMUM-OF-MASS CONCENTRATION
(gm/m^3)
8000
20-25 um Spherical
Aluminum Powder
7000
6000
5000
4000
current (1.2 cm)
3000
current (2.2 cm)
2000
photodiode (1.2, 1.6 cm)
1000
0
0
1 .10
5
2 .10
5
3 .10
5
4 .10
5
5 .10
5
6 .10
EXCESS ELECTRIC FIELD
5
7 .10
5
8 .10
5
9 .10
5
1 .10
6
(V/m)
Figure 5.21. Summary of correlations with excess electric field intensity.
5.4 Other Observed Phenomena in EPS
Two pictures showing interesting phenomena for 20-25 μ m glass powder are
presented in Figure 5.22. The glass “post” formed after the suspension appears a couple of
times for some unknown reasons, which could be due to the electrostatic property of glass
powder. Similar stringer and filament formations using glass powders have been measured
and photographed in normal gravity by Colver (1980) and in microgravity (Colver, 2007).
137
Another interesting phenomenon observed in glass powder is the hexagon formation
following the collapse of the EPS. Particles remain at the edge of the Pyrex glass retainer at
the plate’s electrode center forming a hexagonal pattern; whereas, a few particles are found
between the center and the glass ring. This phenomenon occurred only one time and has not
been observed for other particles. The reason for this hexagon shape is unknown.
(a) Glass “post” formation after suspension
(b) Glass hexagon after suspension
Figure 5.22. Interesting phenomena of glass powder after suspension - 1 g.
138
5.5 Summary
A new external current method for determining the maximum suspension particle is
introduced using EPS. A variety of particles are tested show a similar two-stage development:
an initial increase stage of current followed by a steady stage. The rising stage lends itself to
a linear regression analysis. It is found that different electrode separation distances can be
normalized by a single linear regression equation for some particles.
The concept of excess electric field intensity is utilized to correlate the normal gravity
and microgravity. It is also used to predict the minimum electric field intensity for particle
lift-off. The prediction is comparable to experimental results. By using the drop tower
photodiode data, normal gravity concentrations and microgravity concentrations are
successfully correlated, which suggests the concept of excess electric field intensity is
applicable in handling gravity field effects on EPS suspensions. Attempts at correlating
maximum particle suspension concentrations were not successful. However, a list of
individual linear regression equations found for different materials (glass, copper, aluminum,
magnesium, titanium, iron) and particle diameters are given. These particles are observed to
follow a linear regression.
139
CHAPTER 6. ALUMINUM PARTICLE IGNITION AND QUENCHING
Quenching distance is an important parameter regarding fire safety and laminar flame
theory. It has a rigorous relationship with the laminar flame thickness important for laminar
flame theory, but more difficult to measure experimentally. Because of the broad application
on rocket engines, combustible metal particle clouds have drawn a special interest by
researchers. Only limited data are available on flammability limits, spark ignition energy, and
quenching distance, particularly for rich fuel-air mixtures, as a result of the difficulty of
generating uniform particle clouds. In this chapter, the combustion results of aluminum and
air mixtures are presented, including quenching of aluminum/air mixtures and ignition of
aluminum/copper/air mixtures.
6.1 Aluminum Quenching
Ground based studies using admixtures of coal, nitrogen, and oxygen were previously
studied by Kim (1989). One issue in previous EPS designs in this research is the sealing of
the test chamber. At high concentration, leakage of aluminum powder required that
quenching experiments be performed at relatively low concentrations. Recently designed
sealed test cells produced the ability to achieve high concentrations of aluminum.
The aluminum quenching experiments were carried out in both normal gravity and
microgravity. In the drop tower microgravity experiments, the fuel-air ratios were extended
to rich mixtures for combustion and ignition at fuel-rich side. The flammability experiments
were by trial either ignition or non-ignition. A new 5-tier combustion rig was designed to
perform the experiments more efficiently in both normal and microgravity environments.
140
In drop tower tests, pre-weighed amounts of aluminum powder were placed into each
of the 5-tier test cells (Figure 3.18) set at a fixed quenching distance (electrode separation
distance) by the Pyrex retainers. A predetermined voltage was next applied to generate a
particle cloud inside all test cells simultaneously (Figure 3.20). A second independent high
voltage power supply was set to charge the 5 sets of capacitor banks. Five high voltage
solenoid switches, when triggered, connected the high voltage capacitor banks to the
electrodes of each test cell, causing a spark. This ignition occurred 1.5-1.8 seconds after the
drop. The (apparent) ignition energy is calculated based on the capacitance and preset voltage.
By changing the aluminum-air mixture ratio from lean to fuel-rich side, an ignition
map can be produced at the specified quenching distance. In this way the boundary between
ignition and non-ignition mixtures can be determined. Repeating the procedure can produce a
complete map (and relationships) between quenching distance and aluminum-air ratio.
6.1.1 Aluminum quenching at normal gravity
An example of an EPS generated quenching curve is shown in Figure 6.1, using the
5-tier apparatus in normal gravity for aluminum-air mixtures (sieved size 10-15 μ m,
d32=19.47 μ m). The data show a well-defined ignition-quenching flammability curve
extending into a rich fuel-air limit. A linear regression equation with 95% confidence for the
quenching distance dq is given by
d q = 4.81 × 10 −5 * C al2 − 0.11 * C al + 70.56 ± 0.52 (95% confidence)
(6.1)
where mal is the particle mass concentration of aluminum suspension in a unit of g/m3, dq is
the quenching distance in a unit of mm. This best curve fit is restricted to concentrations less
than about 1500 g/m3, since high concentrations are difficult to produce in normal gravity.
141
The existence of the upper particle limit is mainly due to the gravitational force restriction.
Under these conditions, the suspension can become unstable, with column structures
observed. Two other sizes of aluminum have been tested in normal gravity. They will be
presented along with quenching data in microgravity. These sizes are d32=23.72 μ m (sieved
size 20-25 μ m) and d32=31.87 μ m (sieved size 25-30 μ m). In microgravity experiments,
only the middle test cell in the 5-tiered rig is monitored by a video camera (Figure 3.19) and
is accessible to optical measurements. Ignition versus non-ignition is simply judged by
examining the burned products after the drop.
No Flame
Flame
Polynomial Fit
Quenching Distance (mm)
20
15
95% confidence interval of fit
10
5
500
1000
1500
3
Particle Concentration (g/m )
Figure 6.1. Quenching curve for Al (d32=19.47 μ m) - 1 g.
142
6.1.2 Aluminum quenching in microgravity
The normal gravity quenching data and microgravity quenching data are presented
together in Figure 6.2 for aluminum (15-20 μ m). For the microgravity experiments, three
different quenching distances were tested, 13, 17, and 22 mm. Only the ignitions at 17 mm
were tested twice, the others were only tested once (Colver, 2007). The results prove
interesting.
22
Quenching Distance (mm)
20
18
16
14
12
10
No Ignition - 1 g
Ignition - 1 g
No Ignition - 0 g
Ignition - 0 g
8
6
4
0
1000
2000
3000
4000
5000
6000
3
Particle Mass Concentration (g/m )
Figure 6.2. Quenching curve for Al (d32=19.47 μ m) - 1 g and 0 g.
At least one microgravity test was ignited falling inside the flammability limit based
on the ground test, i.e., the first ignition result at a quenching distance of 17 mm. However,
not all microgravity tests inside the flammability tests meeting this criterion were ignited.
One possible explanation is that ignition is a random event and a single test does not
143
guarantee a successful ignition, even inside the flammability limit. The high humidity is also
a factor during drop tests, which can affect suspension formation and ignition (Colver, 2007).
High humidity affects the surface conductivity of the Pyrex retainer and the tendency for
particles to deposit on the glass. Some of the drop tower experiments were performed during
August in Cleveland, OH with relative humidity up to 70%.
The ignition development pictures of aluminum (15-20 μ m, 500 mg Al sample at 17
mm separation) are shown in Figure 6.3. They are recorded by a video camera during the
drop tower test working at 30 frames/sec. The ignition is bright at the beginning from the
ignition of particles, which suggests a higher combustion heat release. The total burn time for
the constant volume combustion process during microgravity is approximately 0.5 seconds.
The post combustion products suggest that not all aluminum particles are burned during the
combustion process for rich mixtures.
The quenching data for aluminum (20-25 μ m) are presented in Figure 6.4. Unlike
the smaller diameter for aluminum presented in Figure 6.2, most of the aluminum mixtures
are successfully ignited. A noteworthy difference is the ignition test at 17 mm quenching
distance in which the experiments were performed 6 times. Tests with over 50% successful
ignitions are considered a successful at a particular concentration. Concentrations at 13 and
22 mm quenching distances were only performed once and most tests were ignited
successfully as determined by an examination of the combustion products. Aluminum
concentration of 1117 g/m3 at 13 mm and a concentration of 1202 g/m3 were not ignited,
even though they fell inside the flammable limit. Again, this can be explained statistically as
the result of a single test.
144
Before ignition
Bright ignition
Bright combustion
Burn
Burnout
Burnout
Figure 6.3. Ignition development in drop tower for Al (d32=19.47 μ m).
145
Quenching Distance (mm)
22
20
18
16
14
12
No Ignition - 1 g
Ignition - 1 g
No Ignition - 0 g
No Ignition - 0 g
10
8
6
0
1000
2000
3000
4000
5000
6000
3
Particle Mass Concentration (g/m )
Figure 6.4. Quenching curves Al (d32=23.72 μ m) - 1 g and 0 g.
The quenching data for aluminum (25-30 μ m) are presented in Figure 6.5. For these
runs there were no data available at the 13 mm quenching distance, due to the limited drop
tower experiment time. The experiments were performed twice at a 17 mm separation and 3
at 22 mm separation. At 17 mm quenching distance, the isolated non-ignition point is likely
due to the probability.
146
Quenching Distance (mm)
22
20
18
16
14
12
No Ignition - 1 g
Ignition - 1 g
No Ignition - 0 g
Ignition - 0 g
10
8
6
0
1000
2000
3000
4000
5000
6000
3
Particle Mass Concentration (g/m )
Figure 6.5. Quenching curves for Al (d32=31.87 μ m) - 1 g and 0 g.
Figure 6.6 shows a video burn sequence spark → ignition → burnout in microgravity
of 20-25 μ m aluminum-air mixture (500 mg Al sample at 17 mm separation) (Colver, 2007;
Colver et al., 2008)). The spark is initiated just before 1.57 seconds giving the suspension
sufficient time to form the high concentration in microgravity. The ignition of powder
follows at 1.60 seconds burning until about 2.0 seconds. From the first frame (1.57 seconds)
there is visual evidence of a breakdown occurring at the wall (not at the needle electrode),
accounting for the burn propagation from right to left in the suspension. The total burn time
for the constant volume combustion process during microgravity is approximately 0.50 ±
0.015 seconds. The initial spark does not occur happen at the center needle electrode as
expected but at the wall, which suggests the deposition of aluminum particles on the wall due
to a humidity effect.
147
No Ignition
Wall spark
Wall ignition
Ignition
Burn
Burn
Burnout
Burnout
Figure 6.6. Ignition development in drop tower for Al (d32=23.72 μ m).
148
It is noteworthy that all diameters of aluminum failed to ignite at concentrations 1120
g/m3 in contrast to ground-based EPS experiments. One possible explanation is the effect of
humidity reducing the actual concentration to outside the lean flammability limit, while it is
assumed that all particles are suspended. This question cannot be answered since cameras (or
laser) are not checking all 5 test cells.
Figure 6.7 shows another video burn sequence spark → ignition → burnout in
microgravity of 25-30 μ m aluminum-air mixture (500 mg Al sample at 17 mm separation).
From the video frames, it is evident that the spark is initiated at the wall rather than at the
center point electrode. This is probably due to the particle buildup at the wall because of
humidity heavy loading mass of particles. A wall spark is not desirable, but acceptable. This
ignition process takes place “locally” with only the particles at the right side of the cell
ignited. In this case the flame does not propagate across the chamber but burns out locally.
This could be explained that the rich fuel ratio requires additional energy, which is lost to the
wall while the remainder of the suspension is receives insufficient energy in the form of
radiative heating for ignition.
149
No Ignition
Wall spark
Wall ignition
Ignition
Burn
Burnout
Burnout
Burnout
Figure 6.7. Ignition development in drop tower for Al (d32=31.87 μ m).
150
6.1.3 Aluminum concentration correction at microgravity
The quenching results of aluminum in microgravity are presented in the previous
section. The data assume 100% of the sample powders are suspended inside the test cell by
electric field intensity. Unfortunately, this assumption can be faulty, due to the upper limits
of suspension concentration at given electric field intensity. Consequently, some of the (very)
rich fuel data must be lower mass concentration values. The assumed mass concentrations are
tested against the linear regression equation in Figure 5.12—the maximum permitted mass
concentration for a given excess E-field value and quenching distance—and the smaller of
the two values of concentration is then plotted in corrected quenching points. The suspension
mass corrections below, using excess electric field intensities combined with the maximumof-mass suspension from circuit current measurement follows that of Colver (2007).
For the 20-25 μ m aluminum, the maximum permitted suspension concentration for a
given electric field intensity based on external current tests in Table 5.2 is
C al = 4.672 * E + 1437
(6.2)
where Cal is in unit of g/m3 and E is in a unit of kV/m. The quenching data are for
microgravity, but the maximum concentration correlations are for normal gravity. The
concept of excess field intensity (Eq. (5.5)) will be used to correlate the two environments (0
and 1g).
During the drop tower microgravity experiments, the suspension voltages at different
separations were set to the values listed in Table 6.1. The electric field intensity and
corresponding excess electric filed intensity can be calculated. Using Eq. (6.2), the
corresponding maximum mass concentrations are determined for each given electric field.
151
The corrected maximum of mass concentration value replaces the overestimated mass
concentration in the test cell. For example, at a concentration of 4681 g/m3 (assumes 100%
of loaded powder is suspended for quenching distance 17 mm), the corrected maximum
suspension mass is 3370 g/m3, which is considerable reduced.
Table 6.1. Max mass concentration for aluminum (20-25 μ m)
Height (mm)
E (kV/m)
ΔEexcess (kV/m)
Max C al (g/m3)
13
629
512
3779
17
529
423
3370
22
546
438
3437
The corrected quenching data for 20-25 μ m aluminum are presented in Figure 6.8. It is clear
that the mass concentration range is corrected to a narrower range. There are no linear
regression equations available for 15-20 μ m and 25-30 μ m aluminums. However, the linear
equations of copper of different diameters suggest that particle diameter is not a strong factor
for the suspension. If the linear equation for 20-25 μ m aluminum is used, the corrected
results for the other diameters are as shown in Figures 6.9 and Figure 6.10. As in Figure 6.8,
the rich fuel concentrations are now corrected to the maximum permitted concentrations for a
given excess electric field intensity. It is noted that the excess electric field intensity is related
to the particle diameter, so the excess electric field intensity for different diameters will vary.
Consequently, the maximum mass concentrations at a given electric field strength are
different as well. The effect of particle diameter on excess electric field intensity for the same
material is shown in Figure 5.10. The effect becomes significant at the lower electric field
152
intensities, i.e., below 300 kV/m. In these experiments, the electric field intensities are higher
than 500 kV/m, such that the diameter change does not significantly affect the excess field.
intensity. The data in Figures 6.9 and 6.10 are calculated based on Eq. (5.5) and Eq. (6.2).
The detailed calculations are not presented here.
22
Quenchign Distance (mm)
20
18
16
14
12
No Ignition - 1 g
Ignition - 1 g
No Ignition - 0 g
Ignition - 0 g
No Ignition - 0 g (corrected)
Ignition - 0 g (corrected)
10
8
6
4
0
1000
2000
3000
4000
5000
3
Particle Mass Concentration (g/m )
6000
Figure 6.8. Corrected microgravity quenching data for Al (d32=23.72 μ m).
153
22
Quenching Distance (mm)
20
18
16
14
12
No Ignition - 1 g
Ignition - 1 g
No Ignition - 0 g
Ignition - 0 g
No Ignition - 0 g (corrected)
Ignition - 0 g (corrected)
10
8
6
4
0
1000
2000
3000
4000
5000
6000
3
Particle Mass Concentration (g/m )
Figure 6.9. Corrected microgravity quenching data for Al (d32=19.47 μ m).
Quenching Distance (mm)
22
20
18
16
No Ignition - 1 g
Ignition - 1 g
No Ignition - 0 g
Ignition - 0 g
No Ignition - 0 g (corrected)
Ignition - 0 g (corrected)
14
12
10
8
6
0
1000
2000
3000
4000
5000
6000
7000
3
Particle Mass Concentration (g/m )
Figure 6.10. Corrected microgravity quenching data for Al (d32=31.87 μ m).
154
6.1.4 Quenching distance comparison
Ballal (1980) carried out experiments to study the influence of particle size, dust
concentration, pressure, mass transfer number, and oxygen/nitrogen ratio on quenching
distance and minimum ignition energy of dust clouds of solid fuels, including aluminum,
magnesium, titanium, and carbon. His results show a strong similarity between the
mechanism of ignition and quenching of dust clouds of solid fuels and the liquid droplet/air
mixtures. As reviewed previously, “Glassman’s criterion” can be used to classify the metal
combustion into gas (vapor phase of metal) phase combustion and heterogeneous reaction at
the surface of the metal particle. For solid fuels such as aluminum and magnesium, which
have a higher boiling point of the metal oxide than of the metal itself, the reactions generally
can be treated as gas phase combustion similar to liquid fuels. Based on the similarity, Ballal
(1980) suggested that the correlation equation for the quenching distance of liquid fuels is
also applicable for solid fuels, such as aluminum and magnesium. For quiescent dust
mixtures, the equation for solid fuels can be written as
ρf
⎡C3
⎤
d q = D32 ⎢ 3
⎥
⎣ C1 ρ aφ ln (1 + Bst ) ⎦
0.5
(6.3)
C1 = D20 D32
C3 = D30 D32
where D20 , D30 , D32 are surface mean diameter, volume mean diameter, and Sauter mean
diameter, respectively, for mono-dispersed or poly-dispersed particles. φ is the equivalent
ratio (from mass concentration in air). ρ f and ρ a are the fuel and air densities, respectively.
Bst is the mass transfer number at the stoichiometirc ratio. The evaluation of the mass transfer
155
number depends on the way in which the solid fuel burns. For aluminum, the boiling
temperature of the metal Al (2740 K) is less than that of the oxide Al2O3 (3253 K), so that the
particles burns as a vapor phase diffusion flame similar to a liquid droplet with the mass
transfer number quoted by Ballal as
B=
[q H + C (T − T )]
[L + C (T − T )]
st
p ,a
p ,s
g
b
b
(6.4)
s
where H is the heat of combustion for the reaction, L is the latent heat of vaporization of the
solid, qst is the stoichiometric fuel/air ratio, C p ,s and C p ,a are the constant pressure specific
heats for the solid fuel and gas, respectively, and Tg, Tb, and Ts are the temperatures,
respectively, of the gas phase, solids boiling point, and the particle surface.
A comparison of the EPS experiment quenching distance curve and calculations from
Eq. (6.3) is presented in Figure 6.11 (Colver, 2007). The lean ignition on the flammability
curve is taken from the 5-tier EPS ground-based experiments. Equation (6.3) is used for the
quenching distance dq normalized to a single data point, using calculated mass concentration
values to evaluate the equivalence ratio and quenching distance. The trend of the predicted
curve is in good agreement with the ground-based data for 20-25 μ m aluminum (d32=23.72
μ m).
156
IGNITION: 20-25 µm Aluminum-Air Mixtures
QUENCHING DISTANCE (mm)
25
20
15
Theory
Ballal [ 1980 ]
10
Ignition - 1g
No Ignition - 1g
Ignition Normal
Gravity
5
Ballal Eqn
0
0
200
400
600
800
1000
1200
1400
1600
1800
POWDER CONCENTION (g/m^3)
Figure 6.11. Comparison of experimental data with predicted curve - 1 g.
6.1.5 Microgravity combustion products
A newfound phenomenon observed during combustion in microgravity tests with
aluminum-air mixtures is the formation of “stringer and web” structures in combustion
products (Colver, 2007; Colver et al., 2008). Figure 6.12 shows the formation of the
stringer/spider web structure after drop with or without ignition. Figure 6.12 (a) shows a
product of combustion of aluminum after drop. The suspension voltage is 13 kV, the spark
voltage is 25 kV. The products stick to each other to form a fiber type rather than particles.
Due to technical limitation, there is no analysis of the composition of the product. The spider
“stringer” type structures in Figure 6.12 (b) are up to 2 cm in length connecting opposite
positive (lower) and negative (upper) electrodes. The experiment condition is 8 kV
157
suspension voltage and 25 kV spark voltage. Figure 6.12 (c) shows the “spider web-like”
formations with tunnels and pathways across the test cell. It is not know if the web structures
are formed before or after impact, possibly surviving 15-30 g impact. It is postulated that the
chain-like structures are a result of charged and/or polarized particles that agglomerate in the
residual (capacitance) electric fields between the electrodes following shut down. It is
noteworthy that powder columns to 1 cm height and diameter have been “grown” in normal
gravity in the form of stalactites under favorable laboratory conditions utilizing high voltages
and semi-insulating fine glass particles (Colver, 1980). Figure 6.12 (d) shows the growth of
glass particle stringers on Pyrex walls for high humidity conditions at NGRC (> 60 %). EPS
works best in low humidity (< 40 %) or in the absence of walls altogether. Glass becomes
increasingly conductive with humidity, due to the formation of mobile surface ions (Holland,
1966).
In another microgravity study (without E-fields), Mulholland et. al. (2001) studied
gas phase formations of agglomerates and gel-like structures from silane and acetylene
flames. They observed fused spherule structures as small as 150 nm for silene, while
acetylene produced super-agglomerates of 50-100 μm in the plumes of flames. These super
agglomerates appear to be somewhat similar in structures to stringer formations observed
with our EPS experiments. They proposed that the formation of the filaments is due to the
combination of high combustion temperature for silane combustion, together with the lower
melting point of the SiO2 particulate. The same theory can be applied to aluminum
combustion as well. However, the reason for the formation of a stringer/spider web still
remains unknown.
158
(a) Post microgravity stringer product for
667 mg aluminum (25-30 μ m)
(b) Formation of stringer for 400 mg
aluminum (25-30 μ m)
(c) Formation of “spider web” for 100 mg
aluminum (25-30 μ m)
(d) Growth of stringer for 1789 mg glass
bead (53-63 μ m)
Figure 6.12. New found formation of stringer/web - 0 g.
Figure 6.13 show microphotographs of aluminum-air combustion products collected
in microgravity. The reference size is 20-25 μ m unburned aluminum in Figure 6.13 (a). The
smallest scale is 10 μ m. The sample shown in Figure 6.13 (b) from stringers (c.f. Figure
6.12 (b)) shows some indication of agglomeration. Figure 6.13 (c) is “fluff” material similar
to that found in Figure 6.12(b) collected outside the EPS test cell as a result of cell leakage
during combustion. There is no clear distinction between particle sizes and agglomerates
between the figures. Combustion of aluminum metal produces hollow shell-like structures
with low density compared to the parent aluminum. It is possible that the lighter products
159
will more readily form the stringers and webs in Figure 6.12 via electrostatic and other
particle-particle forces. Solids in flames are known to be charged by four mechanisms:
thermionic, impact, photo, and field emission (Lawton and Weinberg 1969).
(a) Unburned
(c) Fluff product
(b) Stringer product
Figure 6.13. Microphotos of aluminum combustion product.
6.2 Effect of Inert Particles on Aluminum Apparent Ignition Energy
Besides the quenching distance, the ignition energy is also an important parameter
regarding powder ignition. The apparent ignition energy is calculated, based on the
capacitance bank and preset voltage. Quenching and extinction of flames can be considered
due to the extraction of heat by the environment and inert particles as studied by Yu (1983)
with EPS. The spark ignition characteristics of particle clouds of spherical copper (inert)
particles and spherical aluminum (combustible) particles were investigated using the EPS
method, by varying the aluminum (combustible) particle number density, copper (inert)
particle diameter, number density and ignition spark energy. This investigation was only
performed at normal gravity.
160
The experiment is similar to the 5-tier EPS setup in Figure 3.18, but only 2 test cells
were used to save time during assembling and disassembling of the rig. Specified amounts of
aluminum and copper particles were weighed into each of the 2 test cells for each test. To
ensure that an electrical breakdown occurred at a lower ignition voltage (lower spark energy),
test cells of small (11mm) electrode separation distances were used. A needle electrode at the
center of each test cell delivered the spark pulse to ignite the mixture in each test cell.
The ignition spark energy was controlled by either changing the stored capacitance or
the voltage. The capacitance bank was set at 30 nF at first, then the spark voltage was
changed from a higher (20 kV for example) value to lower (15 kV) values to determine the
boundary between ignition and non-ignition. A minimum voltage is required for breakdown.
Subsequently, the capacitance bank was changed to 20 nF to maintain the electrical
breakdown at a lower spark energy (but a sufficiently high sparking voltage). The minimum
capacitance was 10 nF. During the experiments, the minimum spark voltage for successful
electrical breakdown for the 11mm separation test cell was found to be approximately 8-9 kV.
Consequently, the possible apparent spark energy range was varied from 0.32 to 9.4 J.
6.2.1 Aluminum apparent ignition energy without inert particles
An example of the apparent ignition spark energy plotted against the aluminum
(d32=19.47 μ m) mass concentration is shown in Figure 6.14. It is clear there is a boundary
between ignition and non-ignition, and this boundary voltage (energy) represents the
minimum ignition voltage (energy). It is also evident that increased minimum ignition energy
is required for a leaner fuel/air ratio. The particle/air mixture is not ignitable at 90 g/m3 for an
aluminum mass concentration. The required ignition energy decreases with an increase in
161
aluminum concentration. During the experiments, clouds with mass concentrations higher
than 723 g/m3 were successfully ignited with 0.98 to 0.32 J (14 to 8 kV). Because of the
difficulty of generating electrical breakdown for a spark voltage lower than 8 kV, the ignition
testing of aluminum was not possible below a mass concentration of 723 g/m3. It is expected
that the required ignition energy will continue dropping to a minimum aluminum mass
concentration.
5
Spark Energy (J)
4
No Ignition
Ignition
Boundary
3
2
1
0
0
100
200
300
400
500
600
700
800
3
Particle Mass Concentration (g/m )
Figure 6.14. Apparent ignition energy of aluminum (15-20 μ m) - 1 g
Ignition energy plots for other diameters of aluminums (d32=23.72, and d32=31.87
μ m, respectively) are presented in Figures 6.15 and 6.16. Both graphs show a similar trends
with Figure 6.14 in which the required minimum ignition energy decreases with an increase
of the aluminum mass concentration. The aluminum particles were not able to be ignited at
162
concentrations of 90 and 127 g/m3. The maximum powder mass concentrations are 814 and
1086 g/m3, respectively, for 20-25 and 25-30 μ m aluminum. The minimum ignition energy
occurs at the maximum powder mass concentration: 0.34 J for 20-25 μ m aluminum and
0.453 J for 25-30 μ m aluminum.
It is also evident that additional ignition energy is required for larger diameter
aluminum compared to smaller diameter aluminum at the same concentration. For example,
at a mass concentration of 633 g/m3, the minimum ignition energies for 15-20, 20-25 and 2530 μ m aluminum are 0.453, 0.536 and 0.783 J, respectively. The conclusion is also
consistent with the aluminum quenching test, which shows a larger quenching distance for
larger diameters of aluminum.
5
Spark Energy (J)
4
No Ignition
Ignition
Boundary
3
2
1
0
0
100
200
300
400
500
600
700
800
3
Particle Mass Concentration (g/m )
Figure 6.15. Apparent ignition energy of aluminum (20-25 μ m) - 1 g.
900
163
5
Spark Energy (J)
4
No Ignition
Ignition
Boundary
3
2
1
0
0
200
400
600
800
1000
1200
3
Particle Mass Concentration (g/m )
Figure 6.16. Apparent ignition energy of aluminum (25-30 μ m) - 1 g.
It is also evident that smaller diameters of aluminum have a broader ignition range
limit compared to larger diameters of aluminum. The lean flammability limit for 15-20 μ m
aluminum is about 127 g/m3, while the lean flammability limit for 20 -25 μ m and 25-30
μ m is about 163 g/m3. The difference between minimum ignition energy and ignition limit
for different diameters can be explained by the different surface/volume ratios. The greater
the surface/volume ratio (smaller diameter) the faster is the heat generation compared to the
heat loss.
Kim (1986) also investigated the ignition energy of aluminum for diameters of 20-25
and 25-30 μ m. From his experiments, the lean flammability limit was 70 and 90 g/m3 for
20-25 and 25-30 μ m aluminum particles respectively, which is smaller than the values of
164
163 g/m3 shown above in this study. The discrepancy might be due to the different sizes of
test cells used.
6.2.2 Aluminum apparent ignition with inert particles (copper)
In this study, 15-20 μ m aluminum powder was taken as the combustible particle
with copper particles chosen as an inert medium to extract heat from the combustion process.
Three diameters of copper particles were tested. The ignitability of aluminum-copper-air
mixtures was tested by changing the spark energy, the particle number density of inert
particles, and the aluminum/air ratio. Copper particles 0.2 and 0.5 gm were tested in
aluminum mass concentrations (181, 362 and 543 g/m3). The suspension voltage was held
constant at 15 kV to generate a steady-state suspension and prevent the generation of corona
discharge.
An example of aluminum/copper/air mixture ignition energy plotted against mass
concentration is shown in Figure 6.17. The inert particles used in this test are 30-38 μ m
copper (d32=38.02 μ m). The ignition energy of aluminum/copper/air mixture shows similar
trends in that more energy is required at leaner concentrations of aluminum. It is evident that
0.5 gm of copper particle requires more energy to ignite compared to 0.2 gm of copper
particle. This is easy to understand because additional inert particles extract more heat from
the ignition-burning process. The additional ignition energy at different aluminum mass
concentrations is nearly constant for the same mass loading of copper, which suggests that
fuel/air ratio may not affect the additional ignition energy due to inert particles. For 0.2 gm of
copper particles, the total ignition energy is about 2.105, 1.44 and 1.05 J at aluminum mass
concentration of 180, 362 and 543 g/m3, respectively. Similarly, the added ignition energy
165
difference ( Δ Ei) is 0.448, 0.54 and 0.5 J, respectively. For 0.5 gm of copper particles, the
additional ignition energy requirement is increased: 1.77 J at 180 g/m3 aluminum
concentration, 1.82 J at 362 g/m3, and 1.85 J at 543 g/m3. The total ignition energy is 3.425,
2.725 and 2.405 J, respectively.
Spark(without Cu)
Spark(0.2 gm Cu)
Spark(0.5 gm Cu)
No Ignition(0.2 gm Cu)
Ignition(0.2 gm Cu)
No Ignition(0.5 gm Cu)
Ignition(0.5 gm Cu)
Spark Energy (J)
4
3
2
1
100
200
300
400
500
600
700
3
Particle Mass Concentration (g/m )
Figure 6.17 Apparent ignition energy of Al (15-20 μ m) with Cu (30-38 μ m) - 1 g.
Figure 6.18 shows the ignition energy for aluminum/air mixture with an addition of
44-53 μ m copper (d32=51.34 μ m). It shows a similar trend with the ignition energy graph
of 30-38 μ m copper. The difference is the additional ignition energy due to inert copper
particles. For 0.2 gm of copper particles, the additional ignition energy difference ( Δ Ei) is
0.37 J, 0.31 J and 0.35 J, respectively. For 0.5 gram of copper particles, increased additional
166
ignition energy is required: 1.23 J at 180 g/m3 aluminum concentration, 1.2 J at 362 g/m3, and
1.27 J at 543 g/m3.
Spark (without Cu)
Spark (0.2 gm Cu)
Spark (0.5 gm Cu)
No Ignition (0.2 gm Cu)
Ignition (0.2 gm Cu)
No Ignition (0.5 gm Cu)
Ignition (0.5 gm Cu)
Spark Energy (J)
4
3
2
1
100
200
300
400
500
600
700
800
3
Particle Mass Concentration (g/m )
Figure 6.18. Apparent ignition energy of Al (15-20 μ m) with Cu (44-53 μ m) - 1 g.
Figure 6.19 shows the ignition graph for 64-73 μ m copper (d32=65.34 μ m). The
graph also shows different additional ignition energies requirements for the same loading
masses compared to the smaller diameter copper particles. For 0.2 gm of copper particles, the
additional ignition energy difference ( Δ Ei) is 0.31, 0.2, and 0.23 J at aluminum
concentrations of 180, 362, and 543 g/m3, respectively. For 0.5 gm of copper particles,
increased additional ignition energy is required: 0.9 J at 180 g/m3 aluminum concentration,
0.81 J at 362 g/m3, and 0.81 J at 543 g/m3.
167
Spark (without Cu)
Spark (0.2 gm Cu)
Spark (0.5 gm Cu)
No Ignition (0.2 gm Cu)
Ignition (0.2 gm Cu)
No Ignition (0.5 gm Cu)
Ignition (0.5 gm Cu)
Spark Energy (J)
3
2
1
100
200
300
400
500
600
700
800
3
Particle Mass Concentration (g/m )
Figure 6.19. Apparent ignition energy of Al (15-20 μ m) with Cu (63-74 μ m) - 1 g.
It is interesting to note that at the same loading mass, the larger copper particles
extract less additional energy than the smaller copper particles. This may indicate that not
only the diameter, but the number density, can affect the ignition. In the next correlation
section, the combination of number density and diameter of inert particles will be considered.
6.2.3 Apparent ignition energy correlation
It can be concluded that the combination of number density and diameter of inert
particles play important roles on the ignition energy (Yu, 1983). Therefore, the apparent
additional ignition energy can be written as
ΔEi = f ( N ) * g ( D)
(6.5)
168
To form a correlation, the first step is to plot the additional apparent ignition energy
Δ Ei for ignition against the number density of copper particles N using a logarithmic scale.
Alternatively, one can plot the Log ( Δ Ei) against Log (N) directly.
Figure 6.20 shows the plot of Log ( Δ Ei) against Log (N) for aluminum concentrations
of 180.9, 361.7, and 542.6 g/m3. The apparent additional ignition energy gives linear fits for
0.2 gm of copper as shown at the lower left corner region of the plot. Similarly, the apparent
additional ignition energy has linear fits for 0.5 gm of copper as shown at the top right corner
region of the figure.
0.3
0.2
m(Cu)=0.5 gm
0.1
3
CAl=180.9 g/m
Slope=0.23
3
CAl=180.9 g/m
0.0
LOG (Δ Ei) (J)
-0.1
Slope=0.58
3
CAl=180.9 g/m
-0.2
-0.3
Slope=0.46
3
CAl=180.9 g/m
-0.4
Slope=0.40
3
CAl=180.9 g/m
-0.5
m(Cu)=0.2 gm
-0.6
Slope=0.48
3
CAl=180.9 g/m
-0.7
-0.8
Slope=0.49
-0.9
3.3
3.4
3.5
3.6
3.7
3.8
3.9
4.0
4.1
4.2
4.3
3
LOG(N) (#/cm )
Figure 6.20. Log ( Δ Ei) Vs. Log (N) - 1 g.
4.4
4.5
4.6
4.7
169
For the same loading mass, the difference of N is the diameter effect of copper
particles. At aluminum concentration of180.9 g/m3, the slope is about 0.23 for 0.2 gm of
copper and 0.4 for the 0.5 gm of copper, respectively. At aluminum concentration of 362
g/m3, the slope is 0.58 for 0.2 gm of copper and 0.48 for 0.5 gm of copper, respectively. At
aluminum concentrations of 543 g/m3, the lines are nearly parallel to each other compared to
other aluminum concentrations. The slope is 0.46 for 0.2 gm of copper, and 0.49 for 0.5 gm
of copper, respectively.
Based on energy plotted against the number density, it appears there is a single
exponential relationship between the additional ignition energy and the number density. The
exponential number is 0.43, if averaging the slopes at different aluminum concentrations and
copper loading masses.
To further examine the relationship between additional ignition energy and the
diameter of inert copper particles, a graph of Log ( Δ Ei) against Log (D) is plotted, following
a similar procedure. Such a plot of Log ( Δ Ei) against Log (D) is shown in Figure 6.21. At
aluminum concentration of 181 g/m3, a linear regression line for each loading mass is
obtained, having slopes 0.69 and 1.19 for 0.2 and 0.5 gm of copper particles, respectively. At
362 g/m3 aluminum concentration, the slopes are nearly parallel, which is expected and
desirable. The slope is 1.74 and 1.43 for 0.2 and 0.5 gm of copper particles. For 543 g/m3
aluminum concentration, the slopes are closer compared to the previous concentrations. The
slope is 1.36 for 0.2 gm of copper and 1.46 fro 0.5 gm of copper particles.
The graphs of additional ignition energy against particle diameter also suggest the
data can be reduced to a single exponential type relation. The exponential value is 1.3, if
averaging the slopes at different aluminum concentrations and copper loading masses.
170
0.3
3
m(Cu)=0.5 gm
0.2
LOG (Δ Ei) (J)
0.1
CAl=180.9 g/m
Slope=1.19
3
CAl=361.7 g/m
Slope=1.43
3
CAl=542.6 g/m
0.0
-0.1
Slope=1.46
3
CAl=180.9 g/m
-0.2
m(Cu)=0.2 gm
-0.3
Slope=0.69
3
CAl=361.7 g/m
-0.4
Slope=1.74
3
CAl=542.6 g/m
-0.5
-0.6
Slope=1.36
-0.7
1.6
1.7
1.8
1.9
LOG (D) (μm)
Figure 6.21. Log ( Δ Ei) Vs. Log (D) - 1 g.
From the correlations above, it is evident that the additional apparent ignition energy
may have exponential relations with diameter and number density of inert particles. By
combining the relationships obtained for number density and diameter of the inert particle, it
is possible to build a new function to correlate the additional apparent ignition energy with
the inert particle. Based on the results shown in Figures 6.20 and 6.21, it can be concluded
that the apparent additional ignition energy can be written as a function of new parameters N
and D in the form of N 0.43 D1.3 . Then Eq. (6.5) can be rewritten as
ΔEi = f ( N 0.43 D1.3 )
(6.6)
Dewitte et al. (1964) developed a theory for the flame inhibition by solid inert
particles based on the kinetic theory of gases. They predicted that the actual kinetic
temperature of the non-adiabatic flame as a function of ND2. They calculated the limiting
171
mean kinetic temperature, from the critical particle number density of 4.9 × 10 5 #/cm3 for
alumina (Al2O3) particles of 20 μ m diameter in 10.9% CH4-21.6%O2-67.5%N2 gases, below
which no flame can be self-sustaining. Although their theory is for steady-state, the
parameter as a total heat transfer area between particles and gases seems to play an important
role in the process of ignition.
Yu (1983) also studied the effect of inert copper particles on propane-air mixtures. He
interpreted the parameter, ND2, as the total surface area of particles per volume of particlegas mixtures. He also interpreted the reciprocal 1 /( NπD 2 / 4) as the mean distance over
which a radical of the reacting gases has to diffuse before reaching a solid particle or the
particle-particle distance based on the particle projected area. He concluded that the particlegas mixture cannot be ignited when its characteristic length is shorter than the allowable
minimum length for ignition at a given ignition energy. This concept is similarly interpreted
as a quenching distance associated with particles.
Although the exponential numbers found for this study are not exactly 1 and 2 (but
0.43 and 1.3 shown in Eq. (6.6)), the concept of ND2 will be adopted in this study for the
correlation. Then the attempt is to correlate the additional ignition energy difference Δ Ei
with ND2:
ΔEi = f ( ND 2 )
(6.7)
This is completed by plotting the experimental results of Δ Ei and ND2 using the
semi-log scale, for various aluminum-air ratios. The results are presented in Figure 6.22. A
single best equation can be found by averaging the three values of additional energies shown
in Figures 6.17–6.19 at each of the 6 data values of ND2 and carrying out a regression fit. The
172
regression fit equation (R2=0.99) and 95% confidence interval are also presented in Figure
6.22. When an exponential relationship is assumed, the additional ignition energy can be
written as
ΔEi = 5.20( ND 2 )1.37
(6.8)
This equation can be transformed into following linear equation with a 95%
confidence interval:
LOG (ΔEi ) = 1.37 * LOG ( ND 2 ) + 0.72 ± 0.015 (95% confidence)
(6.9)
The confidence interval calculation detail is presented in appendix. It is evident that
the average additional apparent ignition energy difference also follows an exponential
behavior reasonably well.
2.0
Average Additional Energy
1.37
y=5.20x
Ave(ΔEi ) (J)
1.5
95% confidence interval of fit
1.0
0.5
0.0
0.1
0.15
0.2
2
0.25
ND (#/cm)
0.3
Figure 6.22. Average ( Δ Ei) Vs. ND2 – 1 g.
0.35 0.4 0.45 0.5
173
6.3 Summary
A polynomial best fit equation is given for the aluminum-air quenching distance in
normal gravity using the EPS method. The results are compared with theoretical calculation
based on Ballal’s theory. The experimental and theoretical calculations show agreement.
Quenching distance data for various aluminum particles in microgravity environment are
presented and some typical aluminum ignition video sequences are shown revealing the
ignition and flame development. Based on the maximum suspension results introduced in
Chapter 5, the nominal aluminum concentrations are corrected to the actual suspension
aluminum concentrations. Some interesting new found web/stringer formations of aluminum
particles after ignition were observed during the drop tower experiments. The reason for the
web/stringer formations remains unknown, although one theory is available. The quenching
effect of inert copper particles on aluminum/air mixtures is also investigated using EPS. The
quenching effect of inert copper particle is obvious, since additional ignition energy is
required when inert copper particles are present in a combustible mixture. An attempt to
correlate the additional ignition energy is made with the parameter ND2 adopted from
previous studies. An exponential equation successfully correlates the average additional
ignition energy difference with ND2.
174
CHAPTER 7. CONCLUSIONS AND FUTURE WORK
A new experiment referred to as EPS (electric particulate suspension) designed for
quenching distance measurement of combustible powder suspensions has been taken from
concept to working prototype and validated in both normal gravity and microgravity. A 5-tier
closed EPS system was designed for drop tower microgravity experiments and successfully
tested in drop tower tests.
A moving laser scan method and laser array scan method were used to measure the
particle suspension concentrations in 1 g and 0 g. Suspension stratification was observed for
normal gravity experiments. For the drop tower microgravity experiments, the stratification
effect was eliminated or reduced. It was also confirmed that higher particle concentrations
can be produced compared to normal gravity for the same electric field intensities. For
example, the aluminum particle suspension concentrations are increased by a factor of 4. A
criterion to minimize stratification is developed based on laser profile scans of suspension in
microgravity.
Two newfound phenomena were observed and photographed in microgravity: (1) The
appearance of cellular structures in suspension clouds, due to high electric field intensity and
absence of gravity, indicating the onset of a dynamic cloud instability; and (2) The formation
of “stringer and spider web-like” structures between the electrodes during the post burn
stages of aluminum-air combustion. An explanation for the latter is that chains of charger
and/or polarized particle agglomerated were formed in residual electric fields.
The maximum particle suspension is determined by an external current method. The
relationship of maximum particle suspension concentration and the given electric field
175
intensity follow a linear relationship. A new concept of excess electric field intensity is
introduced to correlate the maximum mass concentrations produced in microgravity,
compared with the corresponding maximum suspension in normal gravity.
It is also confirmed that microgravity can significantly expand the range of testing to
richer powder mixtures compared to normal gravity. Quenching distance versus fuel-air
ratios is determined through trial (yes/no ignition tests) by varying the amount of powder
placed in each of the 5 test cells. Quenching distance curves are obtained for aluminum-air
mixtures. The experimental results are also compared to theoretical prediction and they
match well.
To understand the effect of inert particles on combustible dust mixtures, apparent
ignition energy experiments were carried out in normal gravity. For aluminum-copper-air
mixtures, it is confirmed that more ignition energy is required, compared to aluminum-air
mixtures. A new parameter of ND2 is adopted to correlate the additional ignition energy. The
additional ignition energy and parameter of ND2 follow an exponential relationship
reasonably well.
Various experimental issues are reviewed during the project. The laser array scan
method used in microgravity experiments is limited in resolution, compared to the single
moving laser scan method. The individual photodiode arrays also cannot be “cleaned” for
laser fringes. The humidity is a very important factor affecting suspension quality. During the
drop tower microgravity experiments, high humidity caused suspension problems, especially
for fine powders in the form of buildup and electrical shorting of powders on the Pyrex
retainer walls. A powder coating in turn can interfere with the laser scanner and also
176
premature sparking and ignitions emanating on the walls. Generally, the most reliable time
for testing EPS are the winter months with low humidity (<20%).
Alternative designs using EPS are possible. In the present study, a closed EPS system
utilizing Pyrex container is used. However, an EPS open system can be used to eliminate or
reduce the effect of humidity. A special ring placed on the bottom electrode can be used to
focus and retain particles. This concept has been validated for copper and glass bead particles
reducing the possibility of wall sparking.
In the present study, using inert particles, only one diameter of aluminum is used.
This should be expanded to include other diameters. Other combustible particles such as
titanium and magnesium can also be suspended using EPS expanding correlations.
For the present study only air is employed as the oxidizer. Other possibilities include
carbon dioxide. In this regard, magnesium was tested using carbon dioxide resulting in bright
flames and depositions on retainer walls. This open system can be very useful in this regard
for future tests.
177
BIBLIOGRAPHY
Abbud-Madrid, A., Srround, C., Omaly, P., and Branch, M.C., 1999. Combustion of bulk
magnesium in carbon dioxide and carbon monoxide under normal, partial, and microgravity
conditions. AIAA paper 99-0695.
Abbud-Madrid, A., Modak, A., Branck, A., and Daily, J.W., 2001. Combustion of
magnesium with carbon dioxide and carbon monoxide at low gravity. Journal of Propulsion
and Power, 17: 852-859.
Andrzejak, T.A., Shafirovich, E., and Varma, A., 2007. Ignition mechanism of nickel-coated
aluminum particles. Combustion and Flame, 150: 60-70.
Ballal, D.R., and Lefebvre, A.H., 1975. The influence of spark discharge characteristics on
minimum ignition energy in flowing gases. Combustion and Flame, 24: 99-108.
Ballal, D.R., and Lefebvre, A.H., 1977. Ignition and flame quenching in flowing gaseous
mixtures. Proceedings of Royal Society of London, A357: 163-181.
Ballal, D.R., and Lefebvre, A.H., 1978. Ignition and flame quenching of quiescent fuel mists.
Proceedings of Royal Society of London, A364: 277-294.
178
Ballal, D.R., and Lefebvre, A.H., 1979. Ignition and flame quenching of flowing
heterogeneous fuel-air mixtures. Combustion and Flame, 35: 155-168.
Ballal, D.R., 1980. Ignition and flame quenching of quiescent dust clouds of solid fuels.
Proceedings of Royal Society of London, A369: 479-500.
Ballal, D.R., 1983a. Further studies on the ignition and flame quenching of quiescent dust
clouds. Proceedings of Royal Society of London, A385: 1-19.
Ballal, D.R., 1983b. Flame propagation through dust clouds of carbon, coal, aluminum and
magnesium in an environment of zero gravity. Proceedings of Royal Society of London,
A385: 21-51.
Bellenoue, M., Kageyama, T., Labuda, S.A., and Sotton, J., 2003. Direct measurement of
laminar flame quenching distance in a closed vessel. Experimental Thermal and Fluid
Science, 27: 323-331.
Berland, A.L., and Poter, A.E., 1956. Prediction of the quenching effect of various surface
geometries. Sixth Symposium (International) on Combustion, 6: 728-735
Bohren, C.F., and Huffman, D.R., 1983. Absorption and Scattering of Light by Small
Particles. John Wiley & Sons, New York, NY.
179
Boust, B., Sotton, J., Labuda, S.A., and Bellenoue, M., 2007. A thermal formulation for
signle-wall quenching of transient flames. Combustion and Flame, 149: 286-294.
Bregeon, B., Gordon, A.S., and Williams, F. H., 1978. Near limit downward propagation of
hydrogen and methane flames in oxygen-nitrogen mixtures. Combustion and Flame, 33: 3342.
Brooks, P., and Beckstead, M.W., 1995. Dynamics of aluminum combustion. Journal of
Propulsion and Power, 11: 769-780.
Bucher, P., Yetter, R.A., Dryer, F.L., Parr, T.P., and Hanson-Parr, D.M., 1996. Flame
structure measurement of single isolated aluminum particles burning in air. Twenty-Sixth
Symposium (International) on Combustion, 26: 1899-1908.
Bucher, P., 1998. Flame structure measurement and modeling analysis of isolated aluminum
particle combustion. Ph.D. Thesis, Princeton University, Princeton, NJ
Bucher, P., Yetter, R.A., Dryer, F.L., Parr, T.P., and Hanson-Parr, D.M., 1998. PLIF species
and ratiometric temperature measurements of aluminum particle combustion in O2, CO2, and
N2O oxidizers, and comparisons with model calculations. Twenty-Seventh Symposium
(International) on Combustion, 27: 2421-2429.
180
Bucher, P., Yetter, R.A., Dryer, F.L., Vicent, E.P., Parr, T.P., and Hanson-Parr, D.M., 1999.
Condensed-phase species distributions and Al particles reacting in various oxidizers.
Combustion and Flame, 117: 351-361.
Chen, L., and Soo, S.L., 1970. Charging of dust particles by impact. Journal of Applied
Physics, 41: 585-589.
Cho, S.Y., Yetter, R.A., and Dryer, F.L., 1992. A computer model for one-dimensional mass
and energy transport in and around chemically reacting particles, including complex gasphase chemistry, multicomponent molecular diffusion, surface evaporation, and
heterogeneous reaction. Journal of Computational Physics, 102: 160-179.
Chung, S.H., and Law, C.K., 1986. An experimental study of droplet extinction at the
absence of external convections. Combustion and Flame, 64: 237-241.
Colver, G.M., 1976. Dynamics and stationary charging of heavy metallic and dielectric
particles against conducting wall in the presence of a DC applied electric filed. Journal of
Applied Physics, 47:273-289.
Colver, G.M., 1980. Electric suspensions above fixed, fluidized and acoustically excited beds.
Journal of Powders and Bulk Solids Technology, 4: 21-31.
181
Colver, G.M., and Ehlinger, L.J., 1988. Particle speed distribution measurement in an electric
particulate suspension. IEEE Transactions on Industry Applications, 24: 732-739.
Colver, G.M., and Howell, D.L., 1980. Particle diffusion in an electric suspension. IEEE IAS
Annual Meeting, 5: 1056-1062.
Colver, G.M., Kim, S., and Yu, T., 1996. An Electrostatic Method for Testing Spark
Breakdown, Ignition, and Quenching of Powder. Journal of Electrostatics, 37: 151-172.
Colver, G.M., Greene, N., Shoemaker, D., Kim, S., and Yu, T., 2004. Quenching of
combustible dust mixtures using Electric Particulate Suspension (EPS): review of a new
testing method for microgravity. AIAA Journal, 42: 2092-2100.
Colver, G.M., 2007. Quenching of Particle-Gas Combustible Mixtures Using Electric
Particulate Suspension (EPS) and Dispersion Methods. NASA grant NCC3-846 (internal)
Final Report
Colver, G. M., Greene, N., and Xu, H., 2008. Quenching of Fuel-Rich Aluminum PowderAir Mixtures in Microgravity Ignited by High-Energy Sparks Using an Electric Particulate
Suspension (EPS), The Combustion Institution, Central States Section, 2008 Technical
Meeting, (Combustion Fundamentals and Applications), University of Alabama, Tuscaloosa,
Apr. 20-22, paper C2-03 (CD).
182
Cotroneo, J.A., and Colver, G.M., 1978. Electrically augmented pneumatic transport of
copper spheres at low particle and duct Reynolds numbers. Journal of Electrostatics, 5: 205223.
Dewitte, M., Vrebosch, J., and von Tiggelen, A., 1964. Inhibition and extinction of premixed
flame by dust particles. Combustion and Flame, 18: 257-266.
Dreizin, E.L., 1996. Experimental study of stages in aluminum particle combustion in air.
Combustion and Flame, 105: 541-556.
Dreizin, E.L., 1999a. Experimental study of aluminum particle flame evolution in normal and
microgravity. Combustion and Flame, 116: 323-333.
Dreizin, E.L., 1999b. On the mechanism of asymmetric aluminum particle combustion.
Combustion and Flame, 117: 841-850
Dreizin, E.L., 2003. Effect of phase changes on metal-particle combustion processes.
Combustion, Explosion, and Shock Waves, 39: 681-693
Dreizin, E.L., Hoffman, V.K., 1999. Constant pressure combustion of aerosol of coarse
magnesium particles in microgravity. Combustion and Flame, 118: 262-280.
183
Dreizin, E.L., Hoffman, V.K., 2000. Experiments on magnesium aerosol combustion in
microgravity. Combustion and Flame, 122: 20-29
Eapen, B.Z., Hoffman, V.K., Schoenitz, M., and Dreizin, E.L., 2004. Combustion of
aerosolized spherical aluminum powders and flakes in air. Combustion Science and
Technology, 176: 1055-1069.
Eimers, Chad, 2002. Particle Speed Distribution in an Electrostatic Suspension. MS Thesis,
Iowa State University, Ames, IA.
Field, P., 1982. Dust Explosion, Handbook of Powder Technology. Vol. 4. Elsevier,
Amsterdam, Netherland.
Figliola, R.S., and Beasley, D.E., 2006. Theory and Design for Mechanical Measurements,
4th edition. John Wiley & Sons, Hoboken, NJ.
Foelsche, R.O., Burton, R.L., and Krier, H., 1999. Boron particle ignition and combustion at
30-150 ATM. Combustion and Flame, 117: 32-58.
Fukuchi, A., Kawashima, M., and Yuasa, S., 1996. Combustion characteristics of Mg-CO2
counterflow diffusion flames. Twenty-Sixth Symposium (International) on Combustion, 26:
1945-1951.
184
Glassman, I., 1996. Combustion, 3rd edition. Academic Press, Orlando, FL.
Glassman, I., and Papas, P., 1994. Determination of enthalpies of volatilization for metal
oxides/nitrides using chemical equilibrium combustion temperature calculations. Journal of
Materials Synthesis and Processing, 2: 151-159
Goroshin, S., Bidabadi, M., and Lee, J.H.S, 1996. Quenching distance of laminar flame in
aluminum dust clouds. Combustion and Flame, 105: 147-160.
Goroshin, S., Kleine, H., and Lee, J.H.S, 1995. Microgravity combustion of dust clouds:
quenching distance measurements. 3rd International Microgravity Combustion Workshop,
August 1995, NASA CP-10174, pp. 141-146.
Goroshin, S., and Lee, J.H.S., 1999. Laminar dust flames: a program of microgravity and
ground based studies at McGill. 5th International Microgravity Combustion Workshop, May
1999, NASA CP-208917, pp. 123-126.
Greene, Nathanael, 2004. Cloud Stratification Measurement of an Electric Particulate
Suspension for Studies in Gravity and Microgravity, MS Thesis, Iowa State Univ., Ames, IA.
Hertgerg, M., 1980. The Theory of Flammability Limits Conductive and Convective Wall
Losses and Thermal Quenching. Report of Investigations No. 8469, USDI-Bureau of Mines.
185
Holland, L., 1966. The Properties of Glass Surfaces. Chapman and Hall, London, UK
Holm, J.M., 1932. On the initiation of gaseous explosions by small flames. Philosophical
Magazine, 14: 18-56.
Jorosinski, J., 1984. The thickness of laminar flames. Combustion and Flame, 56: 337-342.
Jarosinski, J., Lee, J.H., Knystautas, R., and Crowley, J.D., 1986. Quenching of dust-air
flames. Twenty-first Symposium (International) on Combustion, 21:1917-1924.
Kim, K.T., Lee, D.H., and Kwon S., 2006. Effects of thermal and chemical surface-flame
interaction on flame quenching. Combustion and Flame, 146: 19-28.
Kim, Se-Won, 1986. Spark Ignition of Aluminum Powder. MS Thesis, Iowa State University,
Ames, IA.
Kim, Se-Won, 1989. Theoretical and Experimental Studies on Flame Propagation and
Quenching of Powdered Fuels. Ph.D. Thesis, Iowa State University, Ames, IA.
King, M.K., 1978. Modeling of single particle aluminum combustion in CO2-N2 atmosphere.
Seventeenth Symposium (International) on Combustion, 17: 1317-1328.
186
King, M.K., 1993. A review of studies of boron ignition and combustion phenomena at
Alantic Research Corporation over the past decades. In Boron-Based Solid Propellants and
Solid Fuels (Kuo, K.K., ed.), CRC Press, Boca Raton, FL.
Kosnake, K.L., Kosnake, B.J., and Dujay, R.C., 2000. Pyrotechnic particle morphologiesmetal fuels. Journal of Pyrotechnics, 11: 46-52.
Kroll, R.M., 2006. The Measurement of Burning Velocity in an Electrostatic Particulate
Suspension. MS Thesis, Iowa State University, Ames, IA.
Kydd, P.H., and Foss, W.I., 1964. A comparison of the influence of heat losses and threedimensional effects on flammability limits. Combustion and Flame, 8: 267-273.
Law, C.K., 1973. A simplified model for the vapor-phase combustion of metal particles.
Combustion Science and Technology, 7: 197-211.
Lawton, J. and Weinberg, F.J., 1969. Electrical Aspects of Combustion, Clarendon Press,
Oxford, UK
Law, C.K., and Williams, F.A., 1974. On a class of models for droplet combustion. AIAA
Paper 74-147.
187
Legrand, B., Marion, M., Chauveau, C., Gokalp, I., and Shafirovich, E., 2001. Ignition and
combustion of levitated magnesium and aluminum particles in carbon dioxide. Combustion
Science and Technology, 165: 151-174.
Legrand, B., Shafirovich, E., Marion, M., Chauveau, C., and Gokalp, I., 1998. Ignition and
combustion of levitated magnesium particles in carbon dioxide. Twenty-Seventh Symposium
(International) on Combustion, 27: 2413-2419.
Lewis, B., and von Elbe, G., 1961. Combustion, Flames and Explosions of Gases. Academic
Press, NY.
Li, S.C., and Williams, F.A., 1991. Ignition and combustion of boron in wet and dry
atmospheres. Twenty-Third Symposium (International) on Combustion, 23:1147-1154.
Liang, Y., and Beckstead, M,W., 1998. Numerical simulation of quasi-steady, single
aluminum particle combustion in air. Thirty-Sixth Aerospace Sciences Meeting and
Exhibition, Reno, NV, AIAA 98—0254.
Liu, X.K., and Colver, G.M., 1991. Capture of fine particles on charged moving spheres: a
new electrostatic precipitator. IEEE Transaction on Industry Applications, 27: 807-815.
Macek, A., 1966. Fundamentals of combustion of single aluminum and beryllium particles.
Eleventh Symposium (International) on Combustion, 11: 203-214
188
Marion, M., Chauveau, C., and Gokalp, I., 1996. Studies on the burning of levitated
aluminum particles. Combustion Science and Technology, 115: 369-390.
McBride, B.J., and Gordon, S., 1996. Computer program for calculation of complex
chemical equilibrium and applications. NASA Reference Publication No. 1311, Lewis
Research Center, Cleveland, OH.
Mulholland, G.W., Yang, J.C., and Scott, J.H., 2001. Kinetics And Structure Of
Superagglomerates Produced By Silane And Acetylene, Sixth International Microgravity
Combustion Workshop, May 2001, NASA/CP—2001-210826, pp. 289-292
Nagy, J., and Verakis, J.M., 1983. Development and Control of Dust Explosion. Marcel
Dekker Inc, New York.
Palmer, K.N., 1973. Dust Explosions and Fires. Chapman and Hall, London, England.
Pyaman, W., and Wheeler, R.V., 1923. The combustion of complex gaseous mixtures:
mixtures of carbon monoxide and hydrogen with air. Journal of Chemical Society of London,
23: 1251-1259
Potter, A.E., and Berland, A.L., 1956. The effect of fuel type and pressure on flame
quenching. Sixth Symposium (International) on Combustion, 6: 27-36.
189
Price, E.W., 1984. Combustion of metalized propellants. Progress in Astronautics and
Aeronautics, 6: 479-495.
Pu, Y., Podfilipski, J., and Jarosinski, J., 1998. Constant volume combustion of aluminum
and cornstarch dust in microgravity. Combustion Science and Technology, 135: 255-267.
Rossi, S., Dreizin, E.L., and Law, C.K., 2001. Combustion of aluminum particles in carbon
dioxide. Combustion Science and Technology, 164: 209-237.
Sarhan, Ahmed, 1989. Effect of Electrically Driven Particles in Air Flow in a Rectangular
Duct. Ph.D. Thesis, Iowa State University, Ames, IA.
Schoenitz, M., Dreizin, E.L., and Shtessel, E., 2003. Constant volume explosions of aerosols
of metallic mechanical alloys and powder blends. Journal of Propulsion and Power, 19: 405412
Shafirovich, E.Y., and Goldshleger, U.I., 1992. Combustion of magnesium particles in
CO2/CO mixtures. Combustion Science and Technology, 84: 33-43.
Shafirovich, E.Y., Shiryaev, A.A., and Goldshleger, U.I. Magnesium and carbon dioxide: a
rocket propellant for Mars missions. Journal of Propulsion and Power, 9: 197-203.
190
Shafirovich, E.Y., Mukasyan, A., Thiers L., Varma, A., Legrand B., Chauveau, C., and
Gokalp I., 2002. Ignition and combustion of Al particles clad by Ni. Combustion Science and
Technolopy, 174: 125-140.
Shafirovich, E.Y., Bocanegra, P.E., Chauveau, C., Gokalp I., Goldshleger, U., Rosenband V.,
and Gany A., 2005. Ignition of single nickel-coated aluminum particles. Proceedings of
Combustion Institute, 30: 2055-2062.
Shoshin, Y., and Dreizin, E., 2002. Production of well-controlled laminar aerosol jets and
their application for studying aerosol combustion processes. Aerosol Science and Technology,
36: 953-962.
Sloane,T.M., and Schoene, A.Y., 1983. Combustion studies of end-wall flame quenching at
low pressure: the effects of heterogeneous radical recombination and crevices. Combustion
and Flame, 49: 109-122.
Sotton, J., Boust, B., Labuda, S.A., and Bellenoue, M., 2005. Head-on quenching of transient
laminar flame: heat flux and quenching distance measurement. Combustion Science and
Technology, 177: 1305-1322.
Spalding, D.B., 1955. Some Fundamentals of Combustion. Butterworths, London, UK.
191
Spalding, D.B., 1957. A theory of inflammability limits and flame quenching. Proceedings of
Royal Society of London, A240: 83-100
Steinberg, T.A., Wilson, D.B., and Benz, F., 1992. The burning of metals and alloys in
microgravity. Combustion and Flame, 88: 309-320.
Tadahiro, A., Atsumi, M., Terushige, O., Takehiro, M., Yoshio, N., and Mitsuaki, I., 1997.
Evaluation of the explosion strength of Al/KClO3 firework compositions. Proceedings of the
23rd International Pyrotechnics Seminar, 23: 26-36.
White, H.J., 1963. Industrial Electrostatic Precipitation. Addison-Wesley Publishing,
Reading, MA.
Williams, F. A., 1985. Combustion and Theory, 2nd Edition. Benjamin Cummings, Menlo
Park, CA.
Williams, F.A., 1997. Some aspects of metal particle combustion. In Physical and Chemical
Aspects of Combustion: A Tribute to Irv Glassman (Dryer, F.L. and Sawyer, R.F., eds).
Gordon and Breach, Netherland.
Wang, J.S., and Colver, G.M., 2003. Elutriation control and charge measurement of fines in a
gas fluidized bed with ac and dc electric fields. Powder Technology, 135: 169-180.
192
Xu, H., Greene, N., and Colver, G. M., 2008. The Generation of High Concentration and
Uniform Powder Suspension in Microgravity Using an Electric Particulate Suspension (EPS)
for Study of Fuel-Rich Combustion, The Combustion Institution, Central States Section,
2008 Technical Meeting, (Combustion Fundamentals and Applications), University of
Alabama, Tuscaloosa, Apr. 20-22, paper C2-02 (CD).
Xu, Renliang, 2000. Particles Characterization: Light Scattering Methods. Kluwer
Academic Publishers, Norwell, MA.
Yeh, C.L. and Kuo, K.K., 1996. Ignition and combustion of boron particles. Progress in
Energy and Combustion Science, 22: 511-541.
Yu, Tae-U, 1983. Electrical Breakdown and Ignition of an Electrostatic Particulate
Suspension. Ph.D. Thesis, Iowa State University, Ames, IA
Yuasa, S., and Isoda, H., 1988. Ignition and combustion of metals in a carbon dioxide stream.
Twenty-Second Symposium (International) on Combustion, 22: 1635-1641.
Yuasa, S., and Fukuchi, A., 1994. Ignition and combustion of magnesium in carbon dioxide
streams. Twenty-Fifth Symposium (International) on Combustion, 25: 1587-1594.
Yuasa, S., Sogo, S., and Isoda, H., 1992. Ignition and combustion of aluminum in carbon
dioxide streams. Twenty-Fourth Symposium (International) on Combustion, 24: 1817-1825.
193
Zeldovich, Y.B, 1944. Theory of Combustion and Detonation. USSR: Publication of
Academy of Sciences.
Zeldovich, Y.B., and Barenblatt, G.I., 1959. Theory of flame propagation. Combustion and
Flame, 3: 61-73
Zenin, A., Kusnezov, G., and Kolesnikov, V., 1999. Physics of aluminum particle
combustion at zero-gravity. In 37th AIAA Aerospace Sciences Meeting and Exhibition,
January 11-14, Reno, NV, AIAA paper 99-0696.
Zhou, W., Yetter, R.A., Dryer, F.L., Rabitz, H., Brown, R.C., and Kolb, C.E., 1998. Effect of
fluorine on the combustion of “clean” surface boron particles. Combustion and Flame, 112:
507-521.
Zhou, W., Yetter, R.A., Dryer, F.L., Rabitz, H., Brown, R.C., and Kolb, C.E., 1999. Boron
particle combustion model. Combustion and Flame, 117: 227-243.
Zhu, C., and Soo, S.L., 1992. A modified theory for electrostatic probe measurements of
particle mass flows in dense gas-solid suspensions. Journal of Applied Physics, 72: 20602062.
194
Zvuloni, R., Gomez, A., and Rosner, D.E., 1991. High temperature kinetics of solid boron
gasification by B2O3(g): chemical propulsion implications. Journal of Propulsion and Power,
7: 9-13.
195
APPENDIX: UNCERTAINTY ANALYSIS
Whenever measurements are made, uncertainties in the raw data and curve fit results
are inevitable. To estimate the accuracy of the experimental data, it is necessary to determine
the total uncertainty through the use of statistics. It is also important to limit the significant
digits in curve fit formulas and to evaluate how well a curve fit equation fits the data set.
A.1 Significant Digits
The significant digits in curve fit formulas are usually limited by the accuracy of
instrument displays or original data. If a quantity (dependent variable) is calculated from
other quantities (independent variables), the significant digits of the dependent variable are
limited by the independent variable with least significant digits.
A.1.1 Electric field intensity
In this study, the electric field intensity E was not measured directly, but calculated
from the applied suspension voltage V and ring height H. The suspension voltage was
measured by a voltmeter, which has a resolution of 0.01 kV. The height was measure by a
caliper with a resolution of 0.01 mm. Consequently, the calculated electric field intensity has
a resolution of 0.01 kV/m, i.e., two significant digits.
A.1.2 Particle number density and mass concentration
Particle number density can be calculated by particle loading mass and the volume of
combustion chamber, also through a curve fit related to electric field intensity (or excess
electric field intensity). In the curve fit equation Eq. (5.8), the only dependent variable is
196
excess electric field intensity, which has two significant digits giving all the coefficients in
Eq. (5.8) two significant digits.
Particle mass concentration also can be calculated by particle loading mass and
volume of combustion chamber. The mass scale has a resolution of 0.0001 gm, while the
calculated volume (limited by length resolution) has a resolution of 0.01 mm3. Limited by the
independent variable with lest significant digits, the particle mass concentration has only two
significant digits. As a result, all coefficients in related curve equations (Table 5.2 and Eq.
(6.2)) have two significant digits.
A.1.3 Quenching distance
The quenching distance is a function of particle mass concentration in Eq. (6.1). Since
the quenching distance is actually the minimum plate separation allowing flame passage, it
should have the same significant digits as the separation distance of the chamber, i.e., two
significant digits. All coefficients in Eq. (6.1) have two significant digits.
A.1.4 Ignition energy
In this study, the ignition energy was calculated from the capacitance and ignition
voltage. The capacitance has a resolution of 0.1 nF, while the ignition voltage has resolution
of 0.01 kV, giving the ignition energy only one significant digit.
In the correlation equation Eq. (6.8), since both the particle number density and
particle diameter have two significant digits, the coefficients retain two significant digits.
197
A.2 Regression Fit Confidence Interval
A measured variable is often a function of one or more independent variables that are
controlled during the measurement. A regression analysis can be used to establish a function
relationship between the dependent variable and the independent variable, which will hold on
average. The following discussion presents the concepts of regression analysis, its
interpretation, and its limitations (Figliola and Beasley, 2006).
The regression analysis for a single variable of the form y=f(x) provides an mthorder polynomial fit of the data in the form
y c = a 0 + a1 x + a 2 x 2 + ⋅ ⋅ ⋅ + a m x m
(A.1)
where y c refers to the value of y predicted by the polynomial equation for a given value of x.
The values of a0 , a1 ,…, a m are determined by the analysis.
In general, the polynomial found using a regression analysis will not pass through
every data point exactly, so there will be some deviation between each data point and the
polynomial. A standard error of the fit is used to evaluate how closely a polynomial fits the
data set
N
S yx =
∑(y
i =1
i
− y ci ) 2
ν
(A.2)
where N is the total data points, and ν is the degrees of freedom of the fit and ν =N-(m+1).
If both the independent and dependent variables are considered, then the confidence
interval of curve fit, due to random data scatter about the curve fit, is estimated by
± tν , P
S yx
N
(P%)
(A.3)
198
For the particle number density correlation introduced in Eq. (5.8), the 95%
confidence interval is also included. The number of measurements, N, is 6, so the degree of
freedom of the fit ν is 4. The 95% confidence interval is indicated in Eq. (5.8) and Figure
5.11.
For the maximum particle suspension mass concentrations at given electric field
intensities (presented in Figures 5.12-5.20, Table 5.2), the number of measurements and the
degree of freedom of fit are 22 and 20, respectively. The 95% confidence intervals are also
presented in Figures 5.12-5.20 and Table 5.2.
For the quenching distance correlation introduce in Eq. (6.1), the number of
measurements and the degree of freedom of fit are 12 and 9, respectively. The 95%
confidence interval is also presented in Eq. (6.1) and Figure 6.1.
For the additional ignition energy correlation introduced in Eq. (6.8), the number of
measurements and the degree of freedom of fit are 6 and 4, respectively. The 95% confidence
interval is also presented in Eq. (6.9) and Figure 6.22.
A.3 Uncertainty Propagation
Each individual measurement error will combine in some manner with other errors to
increase the uncertainty of the measurement. For a measurement of x which is subject to
some k elements of error, ek , a realistic estimate of the uncertainty in the measurement due
to these elemental errors can be computed using the root-sum-squares method:
u x = ± e12 + e22 + ⋅ ⋅ ⋅ + ek2 (P%)
(A.4)
In this study, the major error comes from the resolution error with a numerical value
of half the instrument resolution. The precision error, which relates to the student t-test and
199
standard deviation of means, is normally important for design stage uncertainty. However, it
is not considered in this uncertainty propagation study due to the limited number of
measurements executed.
For the calculated variable, the total uncertainty is determined from:
⎡⎛ ∂R
uR = ± ∑ ⎢⎜
⎜
i =1 ⎢⎝ ∂xi
⎣
L
⎞ ⎤
⎟u x ⎥
⎟ i⎥
x=x ⎠
⎦
2
(A.5)
where ∂R/∂xi is the partial derivative of the number density with respect to each variable, l is
the total number of variables and u xi is the total estimated error for the respective variable.
In this appendix, only the particle number density measurement uncertainty
propagation is introduced. The total uncertainty of Glass Bead particles using moving laser
scan and laser array scan methods will be detailed next.
The particle number density is calculated by
N =−
ln(
It
Ii
)
AC lγ
(A.6)
The total uncertainty of the number density can be determined by the independent
variables I i , I s , Ac and l. For the parameters l, the uncertainty comes from the caliper
resolution error, i.e., ± 0.005mm.
For the projected cross area Ac is calculated using Eq. (A.5), since it is calculated
from particle diameter
Ac =
π
4
d p2
(A.7)
200
The particle diameter was determined by a particle size analyzer, which has
resolution of 0.1 μ m. So the diameter resolution error is ± 0.05 μ m. Consequently, for the
glass bead particles with diameter of 31.4 μ m, the total uncertainty is ± 2.28 μ m.
For the initial laser intensity and suspension laser intensity I i and I s ,the uncertainty
is different in different scan methods used. For the moving laser scan, the laser power-meter
has a resolution of 0.005 mW (mV), giving resolution error of ± 0.0025 mW (mV). On the
other hand, the laser photodiode array, used in laser array scan method, only has a resolution
of 0.01 mW (mV), giving resolution error of ± 0.005 mW (mV). Consequently, the total
uncertainty would be different even for the same particle and identical testing conditions.
The total uncertainty for glass beads at 500 kV/m electric field intensity using the
moving laser scan method, shown as error bars, is presented in Figure A.1. The total
uncertainty for the same testing conditions using the laser array scan method is presented in
Figure A.2. It is evident that the laser array scan method has a larger error compared to the
moving laser scan method due to the larger instrument error from the laser photodiode array.
201
4.0
3.5
Height (cm)
3.0
Glass Bead
E=500 kV/m
2.5
2.0
Moving laser scan
1.5
1.0
0.5
0.0
0
5000
10000
15000
20000
3
Particle Number Density (#/cm )
Figure A.1. Total uncertainty for Glass Bead using moving laser scan – 1 g.
2.0
Height (cm)
1.5
Glass Bead
E=500 kV/m
1.0
Laser array scan
0.5
0.0
2000
4000
6000
3
8000
Particle Number Density (#/cm )
Figure A.2. Total uncertainty of Glass Bead using laser array scan – 1 g.