University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Colorado School of Mines
|
Table 6.1 Slopes calculated for each test group with respective R2 values
Sample Group Slope: [(mg/m3) / (kW-hr/m3)] / mm R2 values
Concrete 1 4.40 0.952
Concrete 2 2.59 0.996
Concrete 3 12.07 0.914
Limestone 1 0.557 0.796
Limestone 2 6.75 0.946
Limestone 3 2.56 0.885
Sandstone 1 0.265 0.790
Sandstone 2 2.69 0.989
Sandstone 3 3.41 0.375
Average 2.8 0.908
(excluding outliers)
As shown in Figure 5.2 and Table 6.1, there is a positive linear relationship between dust
concentration and specific energy input into the cutting as the pick becomes more worn with a greater tip
(cid:3288)(cid:3282)
(cid:4674) (cid:4675)
(cid:4678) (cid:3288)(cid:3119) (cid:4679)
(cid:3286)(cid:3272)(cid:3127)(cid:3283)(cid:3293) (cid:3417)
(cid:4674) (cid:4675)
radius. With the average slope at 2.8 (cid:3288)(cid:3119) 𝑚𝑚 it can be expected that the concentration of dust per
specific energy unit will increase at a faster rate as the wear of the pick increases concerning the
symmetrical millimeter radius. This value should be considered just a data point, not a concrete number
used to quantify the dust generated in the rock-cutting processes as the picks wear out. This is due to rock
texture and strength changes, pick geometry, cutter head design, and operational settings. Much more data
and perhaps field studies are needed to offer a more suitable model for calculating the dust concentration
generated as a function of SE.
In practice, a system for monitoring power (perhaps an installed power transducer on the input
cable of the machine or at the transformer hook up for the excavation unit) can be used to obtain the
amount of energy used for excavating bulk volume of rock such that cutting specific energy can be
calculated. Then, this value can be compared with the ambient airborne dust across different parts of the
mine, from the heading to major access points into the mine workings. This combination of data could
then unveil the relationship between the dust at the monitoring points with the amount of energy used for
production. With further investigation, this information can subsequently be used to link the power draw
for excavation of the rock of a given volume (which can be calculated from the location of the machine)
to the bit wear conditions and anticipated dust in the mine workings. Ultimately, these preliminary
84
|
Colorado School of Mines
|
CHAPTER 7: CONCLUSIONS AND RECOMMENDATIONS
7.1 Contributions
This study made many contributions, and the findings have addressed the knowledge gaps
discussed in previous sections. The contributions are highlighted in this section and include the study of
various methods along with the design and fabrication of the dust collection system, coordination of the
full-scale cutting experiments, collection of the airborne and fines dust, analysis of the collected samples,
measurement of the quantity of dust, analysis of particle size distributions, analysis of particle shapes,
confirmation of findings in previous studies, examining dust characteristic trends, and comparing the dust
characteristics with the specific energy of rock cutting.
The new and novel methodology for collecting dust in the Linear Cutting Machine can provide rapid
and representative dust samples. This automated dust collection system removed a researcher from
turning knobs and switches while cutting, and therefore, this new system:
Eliminated human error with the exact timing set to control multiple pumps and vacuums;
Allowed for mass collection of dust samples with four ports for four different instruments;
Provided consistent dust collection times for all four instruments with the optimized timed
system;
Decreased the collection time with the consistently optimized timed system;
Ensured a clean environment to collect representative samples before collecting the following
samples with the real-time dust concentration monitor and HEPA filter system;
The automated dust collection system can be utilized at the EMI lab and in similar studies for
future experiments to increase dust collection output, decrease dust collection time, and consistently
provide representative dust samples. With an increase in rapid dust sample collection, this new
methodology can increase the rate of dust studies and understanding of airborne rock dust.
Additionally, with limited previous dust experiments cutting rock in a full-scale environment,
these experiments provided a foundation for studying dust characteristics when cutting samples at the
full-scale level. The results from these experiments provided new dust characteristic data never seen
before because cutting samples at full-scale on concrete, limestone, and sandstone hasn’t been analyzed
before at this level.
These experiments also contributed to the body of knowledge in dust studies by confirming
findings in previous studies. For example, previous studies found that coal dust concentration increases
86
|
Colorado School of Mines
|
with increased symmetrical pick wear for conical picks. The thesis research agrees with this statement,
and the analysis in this study has extended the available information from coal to other rock types,
including concrete, limestone, and sandstone.
The findings in this study offer an incremental improvement in understanding dust characteristics
and their quantitative connection to pick wear. The observations confirmed that dust concentration
increases with the pick tip radius increase. Results also suggest a quantitative relationship and future
testing can expand comparisons and understanding. Additionally, the dust concentration compared to the
cutting forces and specific energy and pick wear reveals future areas for investigation. This quantitative
analysis was one of the novelties of the current studies and can lead to the development of methods for bit
wear management strategies to control respirable airborne dust.
The current study and measurements complimented the limited previous studies and analyses of
the particles generated in the cutting process. These experiments included studying and presenting the
particle size distributions and particle shape analysis as a function of symmetrical pick wear.
The quantitative particle shape analyses performed on the collected dust samples from concrete,
limestone, and sandstone can be used in future studies on the impact of dust on workers’ health in the
construction and mining industries. There is also an understanding that particle shape characteristics are
connected to rock type instead of pick wear.
In the end, the research achieved all the objectives outlined in Section 1.4 Research Objectives
which are as follows:
Verified that the respirable dust concentrations increase as the pick wear increases;
Determined the quantitative measure to track the rise in dust concentration as the pick
wear increases;
Examined the presence of silica within the respirable airborne dust generated from the
rock samples during the cutting process;
Investigated the shifts, or lack thereof, in airborne respirable dust particle size
distributions generated from the three pick wears;
Investigated the particle size distributions of the fines material generated from the three
pick wears;
Investigated the change, or lack thereof, in respirable airborne particle shapes generated
from the three pick wears.
87
|
Colorado School of Mines
|
7.2 Conclusions
For this thesis, full-scale cutting tests generated airborne dust and fines with a new, moderately
worn, and fully worn conical pick. The LCM cut three medium-strength samples, which included a
concrete, limestone, and sandstone sample. Various equipment and instruments collected the dust and
determined the dust concentrations, silica contents, particle size distributions, and particle shapes. A
comparative analysis examined the impacts of pick wear on various airborne and fines dust
characteristics. Following is a summary of the main conclusions of this study concerning the effect of the
symmetrical pick tip wear on characteristics of dust from mechanical excavation systems:
Airborne respirable dust concentrations increased as the pick wear increased for all rock
types.
The concentration of dust samples collected in full-scale testing increased by about 50 mg/m3
for every millimeter increase in pick tip radius in concrete, limestone, and sandstone samples.
The silica content of the dust was a function of the rock type and was not impacted by pick
wear levels. The suspended respirable dust containing quartz or cristobalite was a function of
the original content of these minerals in the rock.
Statistical evidence suggests that the airborne particle size distributions are different from one
another due to pick wear. However, there is no clear trend or connection to pick wear. The
pick wear did not influence significant shifts in airborne particle size distributions within the
respirable size range.
Strong statistical and visual evidence shows that the size distribution of the fines increased as
the pick wear increased. Additionally, the pick wear did not significantly shift the fines
material particle size distributions within the respirable size range.
There is strong statistical evidence that the pick wear did not influence any change in the
particle shapes. Instead, the rock type influenced the particle shapes in the dust samples. All
the picks consistently generated airborne respirable particles in all the rock types with particle
shapes that were slightly oval with mostly smooth edges.
Comparison of the measured dust concentrations with the specific energy of cutting to the
pick wear radii showed a positive linear increase of around 3.0 [(mg/m3) / (kW-hr/m3)] / mm
during symmetrical wear.
88
|
Colorado School of Mines
|
7.3 Recommendations
With this project's scope set to cut and analyze dust generated from three rock samples with three
pick wears, it would be beneficial to continue work in this realm. Following is a list of recommended
follow-up studies:
Cut additional rock types, such as coal and shale, to confirm and extrapolate trends found. This is
because there is a significant focus in research and industry on coal dust. Extending the methods
and analyses in this thesis to coal would be of interest.
Create and use additional pick wears while cutting future samples for dust collection and
analyses. Adding a fourth or fifth pick wear level would increase the resolution and add more
data to verify trends. For example, adding more pick tip radii will improve the dust concentration
analyses by extending the outer bound of wear and increasing the concentration accuracy versus
radius slope rate values.
Cut the samples at different attack angles to the sample surface to see if there is a correlation
between the angle of pick and the dust characteristics.
Perform cutting tests in the field with a water spray system to analyze how the addition of dust
control changes the research results.
Perform cutting tests on highly variable, rough, and angular surfaces to observe any changes in
the mechanisms of dust generation and if there are any changes in the results.
Additional sample collection instruments would increase the number of duplicate samples in the
testing matrix and add informative, real-time results. Adding more cyclones and Tsai Diffusion
Samplers would confirm findings and add duplicates to the studies to improve the accuracy.
Additionally, incorporating a Thermo Scientific Personal Dust Monitor 3700 into the dust
collection setup would provide accurate real-time measurements of the dust concentrations.
Adding in this one instrument can provide a time-dependent study with peak dust concentration
measurements during cutting.
Compare the results of this study with field observations, such as using various dust measurement
and monitoring systems to establish the impact of pick wear on dust generation in operations in
connection to specific energy input to the machines.
Analyze the particle shapes of the fines material. This research only analyzed the airborne particle
shapes, as using the FE-SEM to analyze particles down to 0.25 µm was feasible with the PC
filters. It was more challenging to capture the fines particles on a surface possible for FE-SEM
analysis. With this, the research in this study did not use the Microtrac SYNC laser diffraction
instrument optical analysis feature for samples because the optical range goes as low as 4µm. It
89
|
Colorado School of Mines
|
APPENDIX A
PUBLICATIONS AND PRESENTATIONS
Journal Publications
Slouka, S., J. Brune, and J. Rostami. "Characterization of Respirable Dust Generated from Full-Scale
Laboratory Igneous Rock Cutting Tests with Conical Picks at Two Stages of Wear." Mining, Metallurgy &
Exploration 39.4 (2022): 1801-1809.
Slouka, Syd, et al. "Characterization of Respirable Dust Generated from Full Scale Cutting Tests in
Limestone with Conical Picks at Three Stages of Wear." Minerals 12.8 (2022): 930.
Slouka, Syd, et al. "Preliminary Characterization of Respirable Rock Dust Generated from Cutting Potash
in Laboratory Full-Scale Tests with Radial Picks at Different Stages of Wear." Journal of Environmental
Science and Engineering A11 (2022): 213-219.
Peer-Reviewed Publications
Slouka, S., J. Rostami, and J. Brune. "Characterization of respirable dust samples generated from picks at
differing stages of wear." Mine Ventilation. CRC Press, 2021. 198-207.
Slouka, S., Sidrow, E., C. Tsai, and J. Brune. “Comparing respirable dust characteristics from full scale
cutting tests of three rock samples with conical picks at three stages of wear.” Underground Ventilation.
CRC Press, 2023. 264-273.
Conferences Presentations
“Characterization of respirable dust samples generated from picks at differing stages of wear”. North
American Mine Ventilation Symposium. June 2021.
97
|
Colorado School of Mines
|
ABSTRACT
Fires and explosions in confined spaces are extremely dangerous, destroying homes and
buildings, damaging infrastructure, and posing a fatal risk to civilians and fire responders. In
2016 alone, The National Fire and Protection Association estimated over a million reported fires,
killing 14,650 civilians, 81% of which were home structural fires (Association, 2017). However,
fires and explosions have been a problem across many industries including oil and gas, textiles,
sugar refineries, and retail. A common denominator of the majority of incidents is that the fires
and explosions occurred in confined spaces with complex geometries (i.e. apartment buildings,
homes, industrial facilities, pipelines). This is important because explosions in confined spaces
can quickly accelerate and result in catastrophic events; and obstacles in the path of the flame
could generate a significant amount of turbulence accelerating a high-speed deflagration
resulting ultimately in a detonation.
This is especially important for the coal mining industry where methane gas explosions
are a serious risk in underground mines and can be devastating such as the Upper Big Branch
(UBB) explosion in West Virginia in 2010 which killed 29 miners (Page, et al., 2011), the
Willow Creek explosions in Utah in 2000 which killed 2 miners and injured 8 more (McKinney,
et al., 2001), and the Buchanan Mine in Virginia in 2005 which produced overpressures large
enough to knock down miners (Carico, 2005). Although significant work has been done over the
years to help mitigate these explosions, they still pose a fatal risk to workers. To gain a full
understanding of these gas explosions requires detailed knowledge of mine ventilation schemes,
the movement of methane gas in the mine, and high-speed methane gas deflagrations in the
presence of various obstacles and run-up lengths. Therefore, the main objective of this research
is to build a full-scale, 3D CFD model of a methane explosion in a longwall coal mine and to
help assess risk and potential mitigation methods. The knowledge and experience gained in this
research can easily be applied to other large-scale fires and explosions such as the 2017
Qishayan tunnel explosion in the Guizhou Province of China which killed 12 workers and
injured more (PTI, 2017). The knowledge gained can also help assess the potential risk and
provide guidance for stronger prevention strategies against such disasters; as well as guide future
designs that minimize the potential for such disasters from occurring.
iii
|
Colorado School of Mines
|
Methane deflagration experiments are performed to build a stronger understanding of
how methane flames propagate and interact with obstacles under conditions typically found in a
longwall coal mine. Subsequently the data obtained from these experiments is used to validate
the combustion model across various scales, providing for a robust and accurate model.
Researchers ignited methane-air mixtures in 12cm and 71cm diameter horizontal, cylindrical
flame reactors and a rectangular, experimental box with and without obstacles used to simulate
various gob characteristics. Measurements of methane flame front propagation velocities,
explosion overpressure, and high-speed imaging was performed to develop a comprehensive
understanding of flame behavior and provide multiple points of validation for the continual
improvement of the CFD combustion model. Experimental results show that methane gas
deflagrations in confined spaces are sensitive to the simulated gob characteristics investigated,
which includes but are not limited to ignition location, obstacle location, void spacing, and
obstacle surface topology. Key experiments were reproduced using a two-dimensional (2D) and
three-dimensional (3D) CFD models and results demonstrate the ability of the model to capture
methane flame propagation trends seen in experiments, matching maximum flame front
propagation velocities within 7.5% in some cases. The models predict flame acceleration across
obstacles, capturing the recirculation zone downstream of solid wall-type obstacles and flame
propagation through porous gobs. Additionally, the models capture the effects of surface
topology on local mixing and demonstrates the importance of modeling the gob area discretely
instead of approximating it as Darcy flow porous media when modeling methane flame
propagation in this area.
After validating the different reactor models in 2D and 3D, the CFD combustion model
was combined with a full-scale, 3D ventilation model of an underground longwall coal mine.
Using this combined model, researchers have successfully modeled a methane gas explosion in a
full-scale mine which is the first time this has ever been modeled. The scenario that was modeled
was an ignition at the longwall face, near the headgate drum of the shearer. Results show a large
pressure wave traveling at 350m/s leading an expanding flame front with a velocity of 30-35m/s.
Results show the pressure wave compressing the air ahead of the flame, increasing the
temperature of the unburned gases. This is an important result because increased temperatures
can increase combustion rates and accelerate the flame which could potentially transition the
flame from a deflagration to a detonation. Results from this study also show that the pressure
iv
|
Colorado School of Mines
|
ACKNOWLEDGEMENTS
Firstly, I would like to thank my thesis committee members: Dr. Gregory E. Bogin, Jr.,
Dr. Jürgen F. Brune, Dr. Hugh Miller, Dr. Jason Porter, and Dr. Neal Sullivan for their help and
guidance in bringing this project to fruition. I thank my fellow researchers on this project
including Aditya Juganda for his expertise and helpful discussions.
My sincere gratitude to my parents, Richard and Jean Anne, and my siblings, Martha,
Mary, and Joseph for all their support. Additional thanks to my partner, Patrick Nuessly, and all
my friends for listening to my struggles and making sure I practiced self-love. Thank you to all
the dogs I have sat over the years, for their cuddles, kisses, and warmth.
Finally, this research is made possible with support from the National Institute for
Occupational Safety and Health (NIOSH). Contract #211-2014-600050, and I would like to
thank NIOSH and Dr. Gerrit Goodman for all their support and hard questions. Also, thank you
to HPC at Mines for their use of supercomputing nodes on Mio.
xxx
|
Colorado School of Mines
|
CHAPTER 1
INTRODUCTION
Gas explosions are a hazard in the United States and often result in serious injury, death,
and damage to structures. For example, in 2014 a natural gas explosion in East Harlem, NY
killed 8 people, injured dozens, and leveled two buildings which were over 4 stories tall (Dunlap,
2015; Santora, 2014). In 2010, the San Bruno pipeline exploded in a residential area, resulting in
flames that spread to nearby houses. The Pipeline and Hazardous Materials Safety
Administration (PHMSA), which oversees 2.7 million miles of pipeline, estimates 131 public
fatalities from 2005 to 2018 due to significant incidences which include fatalities or injuries
requiring hospitalization, over $50k in costs, volatile liquid released from 5+ barrels or more,
and liquid released that results in an unintentional fire or explosion (PHMSA, 2018). These
statistics do not include pipeline incidents where a fire/explosion was the source of the incident,
which may increase these numbers. Many of these explosions occur in confined spaces and are
often exacerbated by nearby obstacles. This is especially true for the underground coal mining
industry where methane gas explosions can occur deep within a mine, often in working areas.
It is well known that explosive gas zones (EGZs) of methane and air are present in
underground longwall coal mines and can be a potential hazard to mine equipment, structures,
and workers (Brune, 2014; Karacan, Ruiz, Cote, & Phipps, 2011). In extreme cases these EGZs
can ignite and result in fatal methane gas explosions, as evidenced by the recent mine explosions
detailed subsequently:
• In 2000, the Willow Creek mine in Utah, USA, experienced 4 explosions which reversed
airflow in the mine and resulted in a fire burning behind the shields near the longwall gob
(McKinney, et al., 2001). The mine fire and explosion at Willow Creek was likely due to
a roof fall igniting an EGZ in the longwall gob, coupled with a lack of ventilation air,
resulting in an explosion which killed 2 miners and severely injured 8 others.
• In 2005, the Buchanan mine in Virginia, USA had a mine fire resulting from a roof fall of
thick sandstone releasing methane near the shearer (Carico, 2005). Investigators
hypothesized that the EGZ was ignited from sparks from the shearer or heat from cutting
1
|
Colorado School of Mines
|
bits. Although the Buchanan mine fire resulted in mine damage, no miners were killed in
this accident.
• In 2006, an explosion occurred the Sago mine in West Virginia, USA which killed 12
miners (one from CO poisoning) and injured 1 (Gates, et al., 2006). The investigative
report concluded that an EGZ had formed in an area that had been previously mined and
sealed. The seal was not built properly and investigators believe it may have been able to
only withstand 20psi of pressure forces whereas pressure forces of the explosion were
estimated to be greater than 93psi based on damage in the mine area. The ignition source
was investigated by Sandia National Laboratories who found that likely a lightening
strike caused an arc in a nearby pump cable which ignited the EGZ.
• In 2010, there was an explosion at the Upper Big Branch Mine in West Virginia, USA,
which entrained coal dust leading to a deadly explosion killing 29 miners (Page, et al.,
2011). Reports found that a recent roof fall near the tailgate restricted airflow, allowing
for an EGZ to form near the shearer. The shearer was cutting sandstone and worn shearer
bits left hot smears, which ignited the EGZ resulting in an explosion. The pressure waves
entrained coal dust, due to a lack of rock dust in this area, resulting in a massive coal dust
explosion. Investigators also found that a water barrier did help to stop some of the flame
propagation, but recommended looking into different methods for future sealing.
• In 2014, the Soma Mine explosion in Turkey killed 301 people, trapping hundreds of
miners underground (Tuysuz, Watson, & Smith-Spark, 2014). Also, in 2016 there were
two coal mine explosions in China which killed a total of 65 miners (Luu, 2016; Wang &
Dong, 2016). Unfortunately, there is not enough evidence of these explosions to
determine the EGZ location and cause for ignition.
These examples help demonstrate that coal mine explosions are not just a problem in the
United States, but are an international concern. These examples also show the diversity of
explosions and causes, as well as the need for improved understanding of these explosions.
Current methods used to mitigate the likelihood of an explosion include ventilation schemes,
nitrogen inertization, water sprays, and gob ventilation boreholes among some. However, these
methods do not totally prevent methane accumulation in the mine since coal beds contain
methane at high pressures which naturally migrates and outgases during the mining process
(Karacan, Ruiz, Cote, & Phipps, 2011). Thus, one of the goals of this project is to leverage the
2
|
Colorado School of Mines
|
models during the design and developmental stage of the underground longwall mine to try to
reduce the potential for future explosion disasters.
The amount of methane gas emission depends on a variety of factors both geological and
operational including the depth of the coal seam, longwall panel size, amount of coal production,
mining height, and degasification (Karacan, Ruiz, Cote, & Phipps, 2011). These complexities
added to the fact that mining is a transient process makes it difficult to understand exactly where
these EGZs are located and migrate. Detailed investigative accident reports and research has
proven that EGZs exist in the under- and over-lying strata as well as the collapsed strata/gob and
can migrate towards the working longwall face (Brune, 2014).
The active panel of a longwall coal mine consists of an entry, the longwall face, the
headgate and tailgate, the conveyor/belt, and the longwall face as shown in Figure 1.1 and Figure
1.2. The longwall face is typically 300-400m long and 3m high and the active panel can be up to
2000m long. A shearer cuts along the coal face, moving back and forth along the face.
Meanwhile, hydraulic roof supports hold up the strata (roof) and advance forward as the shearer
cuts away at the face. When the roof supports advance forward, the strata that was once
supported collapses creating the gob as shown in Figure 1.3.
Figure 1.1 Schematic of an active panel of a longwall coal mine. Blue arrows represent airflow
pattern. Figure courtesy of CSM research group.
3
|
Colorado School of Mines
|
tandem, Fig, Bogin, Brune, and Grubb (2016) has designed and built the experimental setups for
investigation of high-speed methane deflagrations at both laboratory-scale (5cm diameter, 43cm
length; 9cm diameter, 81cm length; 12cm diameter, 150cm length; 13.6cm diameter, 1.15m
length; 30.5cm diameter 1.2m length) and large-scale (71cm diameter, 6.1m length) which
provides guidance in developing the combustion models. Experimental results show that mine
conditions such as humidity, temperature, and pressure have a significant impact on methane
flame enhancement across all scales (Fig, Bogin, Brune, & Grubb, 2017). Additionally, Fig,
Strebinger, Bogin, and Brune, 2018 have shown methane flame enhancement across a simulated
rock gob.
The research presented in this thesis aims at providing a more detailed description of
which physical properties of the gob and mine environment impact methane flame enhancement;
rock material, void size, void location, obstacle geometry, and porosity. Methane gas explosions
in a longwall coal mine may occur behind the longwall shields such as the Buchanan mine fire in
2005 (Carico, 2005), in an entry/exit, or near the longwall face such as the Upper Big Branch
explosion (Page, et al., 2011) which all have varying degrees of confinement. Thus, this research
also investigates the impact of confinement on methane flame and pressure wave propagation.
Finally, the ultimate goal of this project is to combine the CFD ventilation model and
combustion model and perform large-scale simulations of methane gas explosions in a mine.
6
|
Colorado School of Mines
|
CHAPTER 2
BACKGROUND
Combustion is integral to society and daily life, providing heat for living spaces,
electricity for industrial processes, energy for transportation, and heat for food preparation. There
are four essential pillars of combustion: heat, fuel, oxidizer, and chemical reactions. The
combination of the four produces rapid oxidation of a fuel, producing heat and/or light (Turns,
2012). Since longwall coal mine explosions typically result in a devastating pressure wave or
blast wave and an either subsonic or supersonic chemical reaction zone, this manuscript is
mainly concern with understanding combustion flames and not autoignition events, though they
may be possible.
Flames are typically categorized by the method of fuel and oxidizer mixing. When fuel
and oxidizer are mixed before the addition of heat, the resulting flame is called a premixed
flame, such as spark ignition engines or gas fired furnaces. Non-premixed or diffusion flames
occur when the fuel and oxidizer mix while chemical reactions occur, such as a cigarette lighter
or candle flame. In longwall coal mining, typically the methane gas in the mine mixes with the
fresh ventilation air and as such, any resulting combustion event and flame is either a partially
premixed or premixed flame. The focus of this study is specifically on premixed flames since the
resulting explosion can be more deadly than a non-premixed flame.
There are two types of combustion waves: deflagrations and detonations. Deflagrations
are combustion waves that propagate at subsonic velocities and are categorized as either laminar
or turbulent. Under certain conditions deflagrations can transition to detonations (deflagration to
detonation transition - DDT) which are combustion waves that propagation at supersonic
velocities. The transition of deflagrations to detonations will be discussed in Section 2.3, but will
be introduced here briefly. DDT can occur when a mixture is ignited in a confined space and
sufficient run-up length and/or obstacles are present, resulting in a leading shock wave that is
coupled to the combustion zone traveling at supersonic velocities. The upstream and downstream
properties of deflagrations and detonations are significantly different: the ratio of the
downstream temperature to the upstream temperature for a deflagration is almost 7.5 versus a
detonation is 8-21 and the ratio of the downstream to the upstream density for a deflagration is
7
|
Colorado School of Mines
|
0.13 versus 1.7-2.6 for a detonation (Turns, 2012). Also, the pressure jump across a detonation
wave is over 10 times larger than a deflagration (pressure jump across a deflagration is
approximately 1) which is why detonations can be so devastating (Turns, 2012). For example, in
longwall coal mining the Upper Big Branch explosion in 2010, which entrained coal dust leading
to a more violent explosion, investigators estimated flames traveling upwards of 450m/s and
pressure waves of 170kPa, which are in the DDT range (Page, et al., 2011). Therefore, many
researchers in longwall coal mining are interested in understanding deflagrations and DDTs
(Lee, Knystautas, & Chan, 1985; Oran & Gamezo, 2007; Oran, Gamezo, & Kessler, 2011)
including the CSM research group.
As mentioned, deflagrations are either laminar or turbulent flames which is dependent on
the velocity of the flame, a characteristic length scale dependent on the surrounding
geometry/environment, and fluid properties. Turbulent flows are dominated by fluid movement
which is highly fluctuating and chaotic in both space and time and, thus are often difficult to
define. The turbulent Reynolds number (Re) is used to help quantify a turbulent flow; it is the
ratio of inertial forces to viscous forces and is defined below in Equation (2.1):
(2.1)
where v’ is the root mean square (RMS) fluctuating velocity of the flow (m/s), is the integral
rms
length scale representing the mean size of the large eddies (m), and υ is the kinematic viscosity
(m2/s) (Irbin, Yetter, & Glumac, 2015; Turns, 2012). For turbulent Reynolds numbers equal to or
less than one, the flow is considered laminar, otherwise the flow regime is turbulent.
Turbulent flames have three main reaction regimes: wrinkled laminar flames (flamelets),
flamelets in eddies, and distributed reaction. These regimes are defined by the relation of laminar
flame thickness, δ , (i.e. the thickness of the reaction zone by molecular transport) to turbulent
L
length scales. The largest turbulent length scale is the macroscale, L, which is dependent on the
surrounding geometry (e.g. in a cylindrical tube L would be the tube diameter). As mentioned,
As mentioned, is the integral length scale and is always smaller than L. Next is the Taylor
microscale, , which relates flow characteristics to the mean rate of strain (Irbin, Yetter, &
λ lo
Glumac, 2015; Turns, 2012). Finally, the smallest length scale is the Kolmogorov microscale, ,
κ
8
|
Colorado School of Mines
|
which is the smallest length scale and represents the scale at which turbulent kinetic energy is
transferred to fluid internal energy (Irbin, Yetter, & Glumac, 2015; Turns, 2012). Knowing all
this, the wrinkled laminar turbulent flame regime is when the laminar flame thickness is less than
the smallest turbulent length scale, the Kolmogorov microscale. The flamelets in eddies regime
is intermediate, when the laminar flame thickness is between the integral length scale and the
Kolmogorov microscale (Irbin, Yetter, & Glumac, 2015; Turns, 2012). Then the distributed
reaction regime is when the laminar flame thickness is greater than the integral scale.
Figure 2.1 The expansion of unburned gas through a laminar, planar flame.
Typically in combustion the three turbulent regimes are plotted in relation to the turbulent
Reynolds number, Equation (2.1), and the Damkohler number (Da), Equation (2.2), which is the
ratio of the characteristic flow time to the characteristic chemical time represented by Equation
(2.3).
(2.2)
(2.3)
where S is the laminar flame speed, which is defined as the “speed of an unstretched laminar
L
flame through a quiescent mixture” (Turns, 2012) as shown in Figure 2.1. The flame front moves
perpendicular in the direction of the unburned fuel/oxidizer mixture (reactants) which has a
9
|
Colorado School of Mines
|
(2.7)
Unfortunately, many of the turbulent flame speed models do not fully take into account flame
stretching and stability (Turns, 2012) and so cannot be used for all flows, but are useful for
estimation.
For small Damkohler numbers less than 1 and small turbulent Reynolds numbers, the
turbulent flame regime is a distributed reaction which is a zone between the burned and
unburned gases which has small integral length scales of eddies, but large RMS velocities. At
this point, it is unclear whether or not flames occur in this regime since they would be subject to
large pressure differences and various length scales making the flame front inherently unstable
(Turns, 2012).
The last turbulent flame regime is the flamelets in eddies which is dominated by
moderate Damkohler numbers and high turbulent Reynolds numbers. In this regime, pockets of
burned and almost burned gas pockets are trapped in the flame zone and are transported to the
burned gas by turbulent mixing. The combustion rate in this regime depends on the rate of
unburned gas pockets.
Premixed turbulent flames travel at subsonic velocities, but they can transition to
detonations (DDT) depending on the confinement of the geometry, mixture ratio, and ignition
source (Glassman, Yetter, & Glumac, 2015). For example, in a tube with both ends open or one
end open, if a mixture is ignited from the open end it will always travel at subsonic velocities.
However, if a mixture is ignited from the closed end of a tube that is either open-closed or
closed-closed, then the combustion wave has the potential to reach the speed of sound as long as
the tube is long enough (Glassman, Yetter, & Glumac, 2015). However, it is still unclear whether
or not methane can transition to a detonation in coal mining conditions (Oran, Gamezo, &
Kessler, 2011). Therefore, this research is mainly concerned with methane gas deflagrations and
the potential of methane flames to transition to a detonation. All the flames presented in this
proposal are turbulent premixed methane gas deflagrations in the wrinkled laminar regime or
flamelets in eddies regime.
12
|
Colorado School of Mines
|
2.1 Parameters influencing laminar flames
There are many factors that affect the laminar flame velocity of fuel/oxidizer mixtures
including temperature, pressure, fuel type, and mixture stoichiometry which shall be discussed in
more detail subsequently. The mixture stoichiometry refers to the amount of oxidizer required to
completely burn an amount of fuel. Since methane is the predominant fuel in underground
longwall coal mining, the global equilibrium reaction of methane and air is defined by Equation
(2.8). In the global equilibrium reaction, methane reacts with air to form carbon dioxide, water,
and nitrogen.
(2.8)
From this global reaction the stoichiometric air-fuel ratio is defined as the ratio of the mass of air
to mass of fuel required for a stoichiometric reaction of methane and air, Equation (2.9).
(2.9)
And finally, the equivalence ratio is the ratio of the stoichiometric air-fuel ratio to the actual air-
fuel ratio:
(2.10)
The equivalence ratio is typically used to define the stoichiometry of the mixture. If the
equivalence ratio is equal to unity, then the fuel-air mixture is stoichiometric. If the mixture has
excess air, meaning it is fuel lean, then the equivalence ratio is less than unity. If the mixture has
excess fuel, it is fuel rich and the equivalence ratio is greater than unity.
As previous discussed, the laminar flame speed of a fuel-oxidizer mixture depends on the
density of unburned and burned gases, which is coupled to the stoichiometry of the mixture
through the global reaction. Assuming a single global step chemistry, an estimate of the laminar
13
|
Colorado School of Mines
|
flame speed can be obtained with assumptions; these assumptions are outlined in Turns, 2012,
following Spalding’s theory for laminar flame propagation (Spalding, 1979), and will not be
discussed here (Turns, 2012). After several assumptions, including no pressure change across the
flame, a general 1-D laminar flame theory is derived from conservation laws:
(2.11)
Equation (2.11) shows that the laminar flame speed is a function of thermal diffusivity (α),
viscosity (υ), the average reaction rate, and density (ρ) which means it is also a function of
temperature. As can be seen from this equation, the laminar flame speed of a methane-air
mixture is highly sensitive to the mixture stoichiometry which has been studied experimentally
by Andrews & Bradley (1972). The results of their study are shown in Figure 2.4 and Figure 2.5.
As can be seen from these figures, the maximum flame speed is achieved at slightly fuel rich
conditions, which also corresponds to the highest flame temperatures. Understanding how flame
speed is affected by stoichiometry is extremely important for underground coal explosions since
there is a widely varying distribution of methane in a mine. For example, in the longwall gob
area, methane concentrations can be close to almost 100%, whereas near the edge of the gob, the
methane is diluted from the ventilation air, resulting in more fuel lean mixtures. This research
investigates the effects of mixture stoichiometry on methane flame propagation and pressure.
In addition to mixture stoichiometry, increasing the unburned mixture temperature can
greatly enhance laminar flame speed. Increasing the temperature of the unburned gas mixture
promotes dissociation of minor species, which increases combustion rates and thus, flame speed.
For methane-air mixtures, the laminar flame speed increases parabolically with the temperature
of the unburned gas which was experimentally determined by Andrews & Bradley (1972):
(2.12)
14
|
Colorado School of Mines
|
Although the laminar flame speed of flames has a positive dependence on the unburned
gas temperature, it has a negative dependence on pressure as shown in Equation (2.13), which
was determined by Andrews & Bradley (1972) for pressures above 5atm. This is due to the fact
that an increase in pressure shifts the equilibrium and suppresses dissociation of minor species,
resulting in slower flame speeds. However, this is not the case for turbulent flames where the
increased pressure can help increase local temperatures and flame wrinkling leading to more
violent explosions. This was introduced earlier in this Chapter and will be discussed in later
sections.
(2.13)
Finally, since laminar flame speed is a property of a fuel/oxidizer mixture, each fuel has a
different laminar flame speed. For example, the laminar flame speed of a stoichiometric mixture
of methane and air at room temperature, 300K, and 1atm is approximately 40 cm/s, whereas
hydrogen is significantly faster at 210 cm/s (Turns, 2012). A combination of factors result in
hydrogen’s high flame speeds including a higher thermal diffusivity, higher mass diffusivity, and
fast chemical kinetics.
Thus far the properties of premixed, laminar methane flames have been described in
terms of a propagating flame. However, there are many processes which can hinder the
propagation of a flame, including flame quenching mechanisms, flammability limits, and
minimum ignition energies. The quenching distance of a flame is defined as the critical diameter
of a cylinder where a flame is extinguished due to heat losses to the cool walls of the cylinder
(Irbin, Yetter, & Glumac, 2015; Turns, 2012). The analytical derivation of quenching criteria is a
balance of heat of reaction and heat lost to the cool walls by conduction. However, typically
quenching distances of mixtures are determined experimentally and depend on the flame
thickness which is a function of mixture stoichiometry and initial conditions. For a
stoichiometric mixture of methane and air at 300K and 1atm, the laminar quenching distance is
typically on the order of 2mm (Turns, 2012). Understanding the quenching limits of methane
flames is important for this research because typical coal mine explosions occur near or within
the gob area such as in the Willow Creek explosion in 2000 and the Upper Big Branch explosion
in 2010 (McKinney, et al., 2001; Page, et al., 2011). There has been anecdotal evidence over the
16
|
Colorado School of Mines
|
years from miners who have heard popping or seen flames deep within the gob, which never
propagate to the longwall face (Brune, 2014). Since the gob area consists of different sizes of
rock rubble and varying levels of compaction, perhaps some flames deep within the gob are
quenched by the nearby rock rubble. Thus, it is one of the objectives of this project to model
these types of phenomena and determine whether or not previous anecdotal evidence suggests
methane flames can occur deep within the gob, as well as potentially propagate if originating
near the longwall shields.
Another property of a fuel-oxidizer mixture is the flammability limits of the mixture
which are upper and lower limits in which a flame will propagate given a minimum amount of
energy. The lower limit corresponds to the limit of propagation of a lean mixture and the upper
limit corresponds to flame propagation in the richest mixture. Typically these limits are
determined experimentally and are dependent on the type of experimental apparatus employed.
In general for methane and air mixtures, the flammability limits are between 5-15% for upward
propagation in a vertical cylinder (Coward & Jones, 1952). However, the flammability limits of a
given mixture can vary depending on the apparatus (vertical or horizontal), gravity, and pressure.
For example, Ronney and Wachman (1985) found that the lean flammability limits of methane
and air can change almost 20% traveling upward versus downward.
Taking into account the fuel-oxidizer mixture and environmental conditions, mixtures
require a certain amount of minimum ignition energy for a flame to propagate. The minimum
energy is typically determined by assuming that all heat from an electric spark is transferred to a
critical volume of mixture (Turns, 2012). The critical volume is determined by balancing the
amount of heat released from combustion and heat loss by conduction to the surrounding
environment. Thus, for stoichiometric mixtures of methane and air at 300K, 1atm, the minimum
ignition energy is approximately 0.5mJ which will be compared to experimentally determined
minimum energies in Section 2.7 (Turns, 2012). This is important for longwall coal mining
because there are many sources of ignition in a mine including rock-on-rock friction, machine-
on-rock friction, hot smears, spontaneous combustion, and lightening among some.
2.2 Overview of flame dynamics in smooth cylindrical reactors
Flame propagation has been studied for decades and the apparatus and methods by which
researchers study flames vary depending on the purpose of the research. For example, typically
when measuring the laminar flame speed of a mixture, researchers have used flat flame burners
17
|
Colorado School of Mines
|
or spherical bomb methods (Agnew & Graiff, 1961; Andrews & Bradley, 1972; Rallis &
Garforth, 1980). However, in 1908 Henry Le Chatelier began studying flames in cylindrical
reactors, where the flame traveled along the axis of the reactor (Le Chatelier, 1908). After this,
many researchers began studying flame propagation in cylindrical or rectangular reactors
because it allows researchers to easily vary experimental conditions. Additionally, many
researchers focus their studies on industrial gas pipeline explosions which is also why using
cylindrical reactors has become common in studying fundamental flame propagation. One of the
main advantages of using a cylindrical reactor is that it easily lends to varying end conditions
(i.e. open-open, open-closed, closed-closed) which is important because research has shown end
conditions results in varying flame propagation speeds, overpressure, and flame propagation
phenomena. In a longwall coal mine, methane gas explosions typically occur in rectangular,
horizontal hallways or passageways. Additionally, methane gas explosions can occur in areas
that are either completely confined or partially confined. The main difference with
experimenting in a rectangular chamber is that there will be residual pockets of unburned gas in
the corners of the chamber (Cooper, Fairweather, & Tite, 1986; Solberg, Pappas, & Skramstad,
1981). These pockets can continue to burn after the main flame front has propagated, which can
increase pressure rise in the reactor (Cooper, Fairweather, & Tite, 1986). However, in a longwall
coal mine explosion the main concern is the propagating flame brush and therefore, this research
uses horizontal cylindrical reactors that are open on one end and closed on the other in order to
more closely resemble flames propagating in a mine environment.
Depending on the end conditions of the experimental setup, the flame propagation
dynamics can be extremely different. For a mixture ignited at the open end of a horizontal
cylinder, the flame first expands spherically as can be seen in Figure 2.6. As the flame comes
closer to the edges of the reactor, the reaction front near the cool walls begins to lose heat due to
the temperature gradient. At this point the flame front is slightly retarded and some hot exhaust
gases flow out of the open end. During this time, some of the hot product gases also rise to the
top of the reactor due to buoyancy. When this happens, the top of flame front is slightly tipped
over at an angle as it continues propagating towards the closed end of the reactor. As the flame
travels towards the closed end of the reactor it combusts all of the mixture. Near the closed end
of the reactor the flame becomes unstable as it compresses and burns the residual mixture. This
phenomena has been observed by many researchers (Ellis & Wheeler, 1928; Gerstein, Levine, &
18
|
Colorado School of Mines
|
Wong, 1951; Guenoche & Jouy, 1953). Additionally, Guenoche and Jouy (1953) found that if
there is a small orifice on the closed end of the reactor it helps stabilize the flame. Further
discussion on flame instabilities will be provided in Section 2.4.
Figure 2.6 Images of a stoichiometric methane-air flame traveling from the open end of a 12cm
diameter quartz reactor towards the closed end. Flame travels from left to right. CH =
4
9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole at the
ign
closed end.
If a mixture is ignited from the closed end of the reactor, it will also initially expand
spherically as shown in Figure 2.7. During flame expansion, pressure waves emanate from the
flame front causing turbulence to develop upstream of the flame front increasing the transport of
unburned gases to the reaction zone, accelerating the flame. Then, due to the small volume and
hot product gases expanding, the pressure behind the flame increases and further accelerates the
flame front. Since there is no obstruction ahead of the flame, the flame is able to freely propagate
towards the open end as shown in Figure 2.7. Due to the confinement and fluid movement ahead
of the flame from the pressure generated in this type of explosion the flame accelerates much
faster than a flame propagating from the open end of the reactor. This type of flame propagation
has been observed and studied by many other researchers using reactors with open-closed end
conditions and closed-closed end conditions (Clanet & Searby, 1996; Ellis & Wheeler, 1928;
Xiaoping, Minggao, Wentao, Meng, & Juniie, 2015).
19
|
Colorado School of Mines
|
Figure 2.7 Images of a stoichiometric methane flame traveling from the closed end of a 12cm
diameter quartz reactor towards the open end. Flame travels from right to left. CH = 9.5±0.3%.
4
Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole at the closed end.
ign
The cases described thus far have been for flames propagating in smooth cylindrical
reactors. However, there are many processes which can affect the propagation of a flame, some
of which includes flame instabilities (both intrinsic and external), the degree of confinement,
environmental conditions, obstacles, and ignition energies and sources. Understanding these
effects are extremely important for developing a comprehensive combustion model of a longwall
coal mine explosion. For example, in a real longwall coal mine, the passages and corridors are
made out of irregular shaped rock, meaning the boundary conditions are not smooth. Also, there
are many obstacles in a mine, including mine equipment, mine workers, and pillars which can
either enhance or retard the flame. For brevity, the following sections will attempt to show some
of the major effects of these different phenomena on flame propagation.
2.3 Transition of a deflagration to detonation
As previously discussed, deflagrations, both laminar and turbulent flames, travel at
subsonic velocities, but have the potential to transition to a detonation which travels at
supersonic velocities. The shock wave, or pressure wave, ahead of a deflagration is supported by
the constant expansion of combustion products as was discussed and shown in Figure 2.7. The
combustion process of a deflagration is different than a detonation, where the shock wave
produced by a detonation increases the temperature and pressures of the nearby gases such that
chemical reactions occur in the form of a flame front. In a detonation, the chemical reaction front
follows the shock wave and travels with the shock wave at an almost constant speed. The
minimum speed that a detonation can travel at is called the Chapman-Jouguet (CJ) detonation
20
|
Colorado School of Mines
|
state (Ciccarelli & Dorofeev, 2008). The CJ velocity can be thought of as the sonic velocity,
usually above 1000m/s, and depends on the fuel and mixture composition because it is a choked
state that supports the minimum entropy increase across the detonation wave (Ciccarelli &
Dorofeev, 2008).
Detonations also have a complex structure and can be visualized many ways including
soot tracking (Ciccarelli & Dorofeev, 2008; Kuznetsov, et al., 2002). Figure 2.8 shows soot
tracking images of methane-air detonations with vary stoichiometries taken by Kuznetsov, et al.
2002. As demonstrated in these images, the detonation wave has a structure to it which is
cellular; typically the size of the cell is determined by the distances between points where the
pressure waves interact and is detonated by λ (Turns, 2012). The cell size changes depending on
fuel, mixture compositon, and initial conditions, which can be seen in these images; the cell size
changes as a function of mixture stoichiometry. The cell sizes are important because a detonation
inside a cylindrical reactor will not be sustained if the cell size, λ, is larger than the diameter of
reactor (Ciccarelli & Dorofeev, 2008).
Figure 2.8 Soot tracking visualization of methane-air detonations with varying mixture
stoichiometries. From the left: CH 8.5% by volume, 11%, and 12%. Image credit: (Kuznetsov,
4
et al., 2002).
Detonations can also occur if the ignition energy source is large enough or a deflagration
transitions to a detonation (Ciccarelli & Dorofeev, 2008; Kuznetsov, et al., 2002; Zipf, et al.,
2013). However, in an underground longwall coal mine, there are not many ignition energy
sources strong enough to immediately produce a detonation, which is why this research is mainly
concerned with DDT.
Ciccarelli and Dorofeev (2008) wrote an overview of flame acceleration and transition to
detonation, noting that the first step to DDT is flame acceleration. To illustrate this consider the
21
|
Colorado School of Mines
|
discussion regarding a confined, closed-end ignition in Figure 2.7. In this scenario, the
confinement forces the combustion products to expand in all directions and the pressure waves
emmanating from the flame front creates movement in the upstream, unburned gases. Together
these higher pressures and temperatures from the confinement accelerates the flame front down
the reactor. The flame acceleration is highly dependent on the reactor size, the mixture
composition, reactor wall/tube roughness, initial conditions, and presence of obstacles (Ciccarelli
& Dorofeev, 2008; Kuznetsov, et al., 2002; Silvestrini, Genova, Parisi, & Trujillo, 2008). As
discussed, the detonation cell size has to be smaller or on the order of the reactor diameter in
order to sustain a detonation. The flame acceleraton process has a critical distance, the run-up
length, at which the flame inside the reactor will transition to a detonation. Previous researchers
have compared the length to diameter (L/D) ratios necessary for flames to transition in smooth
reactors and for methane-air mixtures this is typically above an L/D ratio of 50 (Ciccarelli &
Dorofeev, 2008; Lee J. , 1984). However, as the diameter of the reactor increases, the run-up
distance decreases as shown in Figure 2.9.
Figure 2.9 Run-up distance versus reactor diameter in a smooth reactor for different fuels. Figure
credit: (Ciccarelli & Dorofeev, 2008).
After achieving flame acceleration, the transiton to a detonation can happen in two main
ways: 1) from shock reflection or focusing or 2) from instabilities and mixing (Ciccarelli &
Dorofeev, 2008; Glassman, Yetter, & Glumac, 2015). In shock reflection or focusing, the
pressure waves continually emmanating from the flame front eventually coalesce creating a
22
|
Colorado School of Mines
|
shock wave. As discussed in Section 2.1, the pressures and temperatures across a detonation
wave are hugely different and can autoignite unburned pockets of fuel-air mixture creating a
detonation. Shock focusing is when the shock waves are focused due to a concave wall and
again, autoignites the unburned gas.
The second method of transitioning an accelerating flame to a detonation is by
instabilities and mixing processes. In general, one of the main ways the run-up distance can be
decreased is by roughened walls or the presence of obstacles, which will change depending on
the obstacle configuration (Ciccarelli & Dorofeev, 2008; Chapman & Wheeler, 1926; Glassman,
Yetter, & Glumac, 2015; Kuznetsov, et al., 2002; Oran, Gamezo, & Kessler, 2011; Silvestrini,
Genova, Parisi, & Trujillo, 2008; Zipf, et al., 2013). Significant research has gone into flame
acceleration by roughened tubes and obstacles and this will be discussed in more depth in
Section 2.6. In general, obstacles tend to increase mixing and fluid movement which increases
unburned gases to the flame front in addition to stretching the flame front. Stretching of the
flame front increases combustion rates leading to flame acceleration. This flame acceleration is
important because in a longwall coal mine, the longwall face can be 300m long and the
entryways and corridors in the mine can be kilometers long. Added to the fact that the mine walls
are roughened rock, the run-up distance of flame acceleration in an underground coal mine will
be significantly shorter than experiments in smooth reactors.
There have been many researchers investigating methane flame acceleration to DDT for
application to mine explosions including Zipf, et al. (2013) and Oran, Gamezo and Kessler
(2011). Zipf, et al. (2013) performed experiments in a reactor 1.03m in diameter and 73m long
with obstacles and was able to achieve detonation speeds near the CJ velocity. However, to get
an immediate detonation, a pocket of methane and oxygen mixture was located near the spark
and so it is difficult to determine whether or not this scenario is perfectly indicative of a coal
mine explosion. Oran, Gamezo and Kessler (2011) developed a complex CFD model of methane
flame acceleration to DDT and found transition to detonation by flame interaction with obstacles
and boundary layers which helped to create hot spots for autoignition.
The research presented in this manuscript is mainly concerned with high-speed
deflagrations and determining which aspects of a longwall coal mine environment,
environmental, physical, etc., impact methane flame propagation and the potential to transition to
a detonation. Previous researchers on this project, Fig (2019), investigated the impact of
23
|
Colorado School of Mines
|
environmental conditions and flame propagation as a function of scale by performing
experiments in cylindrical reactors at a variety of scales, 5cm diameter to 71cm diameter, with a
fixed L/D = 8.5. Research showed that methane flame propagation velocities increases non-
linearly as a function of scale (Fig, 2019) which agrees with DDT theory and other reseachers
(Ciccarelli & Dorofeev, 2008; Chapman & Wheeler, 1926; Kuznetsov, et al., 2002; Silvestrini,
Genova, Parisi, & Trujillo, 2008). The research in this manuscript is focused on how different
obstacles, shapes, sizes, and configurations, can impact methane flame propagation as well as
ignition location, energy, and confinement.
2.4 Impact of flame instabilities
Although there may be instances in which premixed flames are planar as shown in Figure
2.1 (page 9), not all flames are planar. Premixed flames are subject to many different types of
flame instabilities including instabilities which are inherent to the combustion process.
Hydrodynamic instabilities are natural for premixed flames because the unburned gas expands
across the flame front, which means the burned gas velocity is higher and density is lower. For
turbulent premixed flames, there is also a larger pressure difference across the flame and it is this
pressure difference eventually disturbs the gas flow such that the flame becomes wrinkled as
shown in exaggeration in Figure 2.3 (page 11). For a wrinkled flame, the gas flow moves
perpendicular to the flame front, which means that the flow converges and diverges so that the
burned gas flow has areas where the flow is accelerated (Jarosinski & Veyssiere, 2009). This
type of instability is called the Darrieus-Landau (D-L) instability and will most likely grow with
time.
Simultaneously, there are thermodynamic-diffusive effects which affect the flame in what
is termed the diffusive zone (Jarosinski & Veyssiere, 2009). In this diffusive zone, which is the
size of the flame thickness, heat is diffused away from the flame front towards the unburned gas.
At the same time, mass is diffused and convected towards the flame front and always opposite of
the heat flux. In areas of the wrinkled flame, where the flame front is concave towards the
unburned gas, the heat flux is locally convergent, helping to stabilize the flame, and the mass or
species flux is locally divergent (Jarosinski & Veyssiere, 2009). The balancing of hydrodynamic
(D-L) instabilities and thermo-diffusive effects is typically evaluated in terms of the Lewis
number (Le), which is a non-dimensional parameter and is the ratio of the thermal diffusivity and
species diffusivity. For example, for a Le < 1 the species diffusion is greater than the stabilizing
24
|
Colorado School of Mines
|
thermal diffusion and a wrinkled flame front will continue to grow and become unstable,
increasing flame speeds. However for Le > 1, the flame may become more stable, depending on
other environmental and mixture conditions, decreasing flame wrinkling and flame propagation
velocities.
The D-L instability phenomena can cause other instabilities such as the Rayleigh-Taylor
(R-T) instability. The R-T instability occurs when there are large density differences along an
interface between two fluids pushing against each other. For a flame, this means that the
wrinkling becomes exaggerated, which can also lead to Kelvin-Helmholtz (K-H) instabilities. As
the flame becomes more wrinkled, the rises and troughs become larger which means that the
velocity differences between the burned and unburned gases are greater. This larger difference in
velocity gives way to the K-H instability which typically results in the shedding of vortices along
the velocity interface.
One example of the impact of the D-L and R-T instabilities on flame propagation is the
tulip flame phenomena as shown in Figure 2.10. This phenomena has been realized for many
years in different experimental setups and different fuel-oxidizer mixtures (Ellis & Wheeler,
1928; Guenoche & Jouy, 1953; Starke & Roth, 1989) and has been a subject of CFD modeling
(Bychkov, Akkerman, Fru, Petchenko, & Eriksson, 2007; Gonzalez, 1996; Gutkowski, 2013).
Figure 2.10 Images of the development of a tulip flame in a stoichiometric mixture of methane
and air. Refer to Section 4.3 for more details. Flame moves from left to right. CH = 9.5±0.3%.
4
Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole at the closed end.
ign
One of the most recent, comprehensive experimental studies of the development of the
tulip flame was performed by Clanet and Searby (1996). They used vertical cylindrical Pyrex
tubes with open-closed end conditions and propane-air mixtures. The reactors had varying
25
|
Colorado School of Mines
|
diameters and lengths: 2.5-5cm diameter and 0.6-6m in length. The mixtures were ignited from
the closed end of the reactor such that the flame propagated from the bottom to the top of the
vertical tube. They discussed four main stages to the tulip development, which have also been
observed in certain experiments performed by the CSM research group as shown in Figure 2.10.
The first stage is a hemispherical expansion of the flame which can also be seen at t=10ms in
Figure 2.10. At this point in time the flame is unaffected by the walls of the reactor. As the flame
develops, it starts to approach the walls of the vessel and creates a finger shape seen at t=30ms.
When the flame interacts with the cool wall, it quenches and rapidly loses heat to the walls as
seen at t=40ms. When the flame loses heat to the walls, the surface area and velocity of the
leading flame front decreases, leading to large density gradients. The large differences in density
and velocity lead to the R-T instability and the flame front becomes inverted, i.e. the tulip flame
seen at t=50ms (Clanet & Searby, 1996). Finally, in the fourth stage of the process, acoustic
effects dominate, K-H instabilities, and the tulip flame is further distorted, sometimes producing
multiple inversions along the flame front as shown at t=60ms.
It is important to note that acoustics play a large role in the stability of a flame. Other
researchers studying detonations have found the development of tulip flames due to pressure
waves or shock wave interacting with the flame front, inverting it and increasing flame
acceleration (Markstein, 1957; Salamandra, Bazhenova, & Naboko, 1959). Understanding the
effects of interactions between flames and acoustics is extremely important, especially for the
transition of a flame from deflagration to detonation. As shown in Figure 2.10, acoustics can
grossly disturb the flame front and the fluid velocity in the unburned and burned mixtures. This
is important because stretching of the flame front leads to greater differences in local fluid
velocities along the flame front and increases combustion rates, leading to faster flames.
Understanding flame instabilities and the interaction of acoustic waves and flames is
extremely important for developing a combustion model of methane gas explosions in longwall
coal mines. In a coal mine, there are many walls for pressure waves to interact with and
reverberate off, which means that in the event of an explosion the pressure waves can compress
to form a shock wave and potentially a detonation. Additionally, many reports of mine
explosions have shown that ventilation controls were destroyed (McKinney, et al., 2001; Page, et
al., 2011), likely from the leading pressure wave. There has also been anecdotal evidence from
mine workers who have seen flames moving back and forth within the gob area (McKinney, et
26
|
Colorado School of Mines
|
al., 2001). This could also be due to the acoustic oscillation of the flame as it continues to burn
methane emanating from the gob area. In general, however, one of the main purposes of this
research is to help explain under what conditions flame instabilities and acoustic instabilities
may exacerbate the propagation of a methane flame in a mine.
2.5 Impact of confinement on flame propagation
The environment of an underground coal mine is very confined, which is why mine fires
and explosions can be so devastating. As previously discussed, ignitions in confined spaces can
result in higher temperatures, pressures, and fluid motion, thereby increasing combustion and the
speed of the propagating flame and can lead to large overpressures and/or blast waves. The effect
of degree of confinement, or venting, on flame propagation has been the topic of research in a
variety of industrial processes (e.g. oil refineries, chemical plants).
In the 1970s, Bradley & Mitcheson developed a simplified theory of venting for spherical
reactors and compared the theory to a significant amount of experimental data from other
researchers (Bradley & Mitcheson, 1978). The theory assumes isentropic compression and
expansion, the flame is symmetric and expands spherically, and there are no interactions between
the flame and pressure waves. Despite the fact that their model is simplified, many of their
findings agree with other researchers. For example, they find that central ignition produces the
largest overpressure, consistent with Fairweather, Hargrave, Ibrahim and Walker (1999) as well
as Kindracki, Kobiera, Rarata, and Wolanski (2007). This is due to the fact that the hot exhaust
gases must travel significantly further to vent from the reactor, which increases local
temperatures and pressures, accelerating the flame away from the vent. Additionally, they find
that as the area of the vent on one end of the vessel increases, the max overpressure decreases
similar to Cooper, Fairweather, and Tite (1986) and Zhang and Ma (2015). Finally, they also
investigated how the pressure of the vent itself (or weight of the vent) affects the overpressure.
They found that as the vent pressure increases, the overpressure of the explosion increases in
addition to the velocity of fluid ahead of the flame (Bradley & Mitcheson, 1978). This has been
studied by many other researchers who further quantified the effect of vent relief pressure or
weight of the vent (Bao, et al., 2016; Cooper, Fairweather, & Tite, 1986). However, Bradley &
Mitcheson (1978) also discuss how their assumptions are simplified; noting that pressure-flame
interactions must be incorporated into future venting theory.
27
|
Colorado School of Mines
|
Soon after the studies by Bradley and Mitcheson (1978), McCann, Thomas, and Edwards
(1985) performed a study in two, small-scale cubical vessels investigating the effects of Taylor
instabilities and other acoustic instabilities (Helmholtz) and how they affect propagating flames
and explosion overpressure. They found that as the vent relief pressure increased, the speed of
sound also increased as well as the frequency of the oscillations (McCann, Thomas, & Edwards,
1985). This finding is important because as the frequency of pressure oscillations increase, they
may disturb the flame front even more, creating cellular structures. Solberg, Pappas, and
Skramstad (1981) performed studies investigating the effects of instabilities for large scale
explosions and they found that high frequency pressure oscillations are much more important in
large scale explosions and the onset of cellularity occurs much sooner, leading to faster
combustion and flame speeds which has been observed by other researchers (Bauwens, Chaffee,
& Corofeev, 2008). Because of this, there has been much research on the elimnation of these
high frequency oscillations and vanWingerden and Zeeuwen (1983) found that lining a vessel
with glass wool helped “damp the acoustic wave and prevent the coupling between the acoustic
wave and combustion to occur”. Though this may be a useful technique for some, whether or not
a similar technique can be employed in a mine environment is unlikely.
Figure 2.11 Example of a typical pressure-time profile of an explosion in a near cubic vessel
with a single vent. Figure modified from: (Cooper, Fairweather, & Tite, 1986).
Finally, one of the most fundamental studies on pressure generated by vented explosions
was by Cooper, Fairweather, and Tite (1986) who performed experiments in a near cubic vessel,
the results of which are reproduced in Figure 2.11. The first peak pressure rise, P , is the result of
1
28
|
Colorado School of Mines
|
the initial kernel expansion when the production of combustion gases exceeds the volume
removed by venting . The second major pressure peak, P , occurs after initial venting when the
2
flame front ignites unburned gases increasing the rate of combustion and pressure rise in the
vessel. After the second peak, Helmholtz modes are excited and the flame front undergoes bulk
motion (Cooper, Fairweather, & Tite, 1986; McCann, Thomas, & Edwards, 1985). Figure 2.11
also shows two additional pressure peaks, P and P . After the Helmholtz oscillations, the flame
3 4
expands, turbulence and combustion increases, and large density gradients are created in the
vessel which increases the pressure, P . Finally, as the production of burned gases decreases, the
3
pressure in the vessel decreases and couples “with the acoustic modes of the vessel” sustaining
pressure oscillations creating a high frequency fourth peak, P (Cooper, Fairweather, & Tite,
4
1986; vanWingerden & Zeeuwen, 1983). In addition to evaluating the physical mechanisms
behind each peak, Cooper, Fairweather, and Tite (1986) also discusses the practical implications
of each peak, noting that the presence of obstacles can produce pressure peaks similar to P . This
3
is important, especially for explosions in a mine environment because pressure waves can
destroy mine structures and equipment and potentially harm workers.
Figure 2.12 Experimental setup used by Guo, Wang, Liu, and Chen (2017) exploring the impact
of multiple vents on explosion overpressure. PT refers to the location of the pressure transducer.
Image credit: (Guo, Wang, Liu, & Chen, 2017).
Though much of the research discussed has only been in regards to a single vent and
investigating the factors leading to more deadly explosions, there has been some research into
the effects of multiple vents as shown in Figure 2.12. Results from Guo, Wang, Liu, and Chen
(2017) show that multiple vents can slightly help reduce explosion overpressure and they can
largely reduce the flame length of the explosion extending out of the vent.
29
|
Colorado School of Mines
|
In general, previous research has shown that compartment geometry, ignition location,
vent size, venting pressure, and reactor scale can greatly affect the max overpressure, flame
propagation velocity, and acoustic interaction with the flame. These are all important
considerations for explosions in underground coal mines. Depending on the ignition location,
whether it is in a passageway, in the gob, or behind the shields, the resulting explosion can be
significantly different. To this end, it is the goal of this research to determine which aspects of a
longwall mine environment has the greatest affect pressure generation and flame propagation.
Thus, experiments are performed with and without obstacles, changing the ignition location and
degree of confinement. Additionally, pressure traces can be correlated to flame propagation
velocities which allows for a thorough understanding of possible flame dynamics in a mine
environment.
2.6 Impact of obstacles on flame propagation
The environment of a longwall coal mine has a variety of obstacles (i.e. mine equipment,
workers, rock piles) as shown in Figure 1.2 and Figure 1.3. The complex geometries make it
difficult to determine how a flame interacts with the obstacle since different mechanisms can be
at play simultaneously. Since many explosions occur in or around the gob area, the CFD model
developed in this work must be able to model methane gas explosions in confined spaces, under
different environmental conditions, and with rock rubble and mine equipment/structures. There
has been a significant number of studies on flame interaction with obstacles and porous media,
varying different parameters such as number of obstacles, obstacle location, and blockage ratio
(BR = Area unobstructed/Total cross sectional area) among some (Chapman & Wheeler, 1926;
Dong, Bi, & Zhou, 2012; Dunn-Rankin & McCann, 2000; Evans, Schoen, & Miller, 1948;
Ibrahim & Masri, 2001; Kindracki, Kobiera, Rarata, & Wolanski, 2007; Masri, Ibrahim, Nehzat,
& Green, 2000; Moen, Lee, Hjertager, Fuhre, & Eckhoff, 1982; Xiaoping, Minggao, Wentao,
Meng, & Juniie, 2015; Yibin, Fuquan, Xiaoyan, Xin, & Hongbin, 2011). This discussion shall
detail some of the more notable studies on flame interaction with obstacles and in following, will
discuss how this manuscript investigates the impact of obstacle parameters on methane flame
propagation and explosion overpressure and how it differs from previous researchers.
One of the first, most fundamental studies on flame propagation across an obstacle was
performed by Chapman and Wheeler (1926). They performed methane-gas experiments in a
horizontal, 5cm diameter brass cylinder, 240cm long open at both ends. The obstacles were brass
30
|
Colorado School of Mines
|
orifice plates (or annuli) that were 1mm thin. In the first part of their study they placed the orifice
plates 40cm from ignition and varied the orifice diameter (3.6, 2.5, 1.5cm). When they compared
flame speed measurements, they found that the flame was slightly retarded upstream of the
obstacle, and then accelerated across the obstacle and downstream. Different orifice diameters
also produced different accelerations; the greatest acceleration was found to be when the orifice
diameter was half the tube diameter, meaning a 50% BR (Chapman & Wheeler, 1926). They also
varied the location of the orifice plate and found as they moved it further from ignition the
velocity recorded at the end of the reactor increased, however this was for a closed-closed reactor
end condition. Varying the thickness of the brass orifice plate decreased the downstream
velocity, though this could be due to the fact that brass has a high thermal conductivity and thus
absorbed heat from the flame, slowing it down. Additionally, they added two or more restrictions
in the tube and found even greater flame speeds than a single restriction. Decreasing the distance
between the restrictions from 70 to 30cm accelerated the flame to even greater speeds. They even
attempted to reach detonation velocities by adding 12 restrictions, but shattered the glass viewing
windows, recording pressures upwards of 3.9atm (Chapman & Wheeler, 1926). Later, Robinson
and Wheeler (1933) performed similar experiments in a slightly larger, steel tube, 30.5cm in
diameter and 32.3m long with 11 steel orifice plates. They were unable to reach detonation
velocities, but they did note a central flame core traveling through the restrictions and residual
burning towards the wall between the obstacles (Robinson & Wheeler, 1933). This is important
and there has been a significant area of research investigating how unburned gas pockets trapped
between obstacles impacts flame acceleration (Ciccarelli and Dorofeev, 2008 (Moen, Lee,
Hjertager, Fuhre, & Eckhoff, 1982)).
In 1982, Moen, Lee, Hjertager, Fuhre, and Eckhoff performed a large-scale experimental
study on the effect of obstacle BR on flame speed and pressure generation, also looking at the
mechanisms of burning pockets of mixture between obstacles and downstream of obstacles.
They ignited stoichiometric methane-air mixtures in a cylindrical steel tube, 2.5m in diameter,
10m long with one end open and one end attached to an ignition tube and the mixtures were
ignited using planar ignition. Some of their results have been reproduced here shown in Figure
2.13 and Figure 2.14. In summary, they found that even a single plate with a low BR can
enhance methane overpressure, the value of which highly depends on the location of the obstacle
relative to ignition. In most circumstances, increasing the BR increased overpressure and flame
31
|
Colorado School of Mines
|
In their theoretical description of the processes leading to enhanced overpressure and
flame speeds due to obstacles, they noted that as the flame travels across the obstacle it results in
a “nonuniform flow field” which “increased burning rate” and overpressure (Moen, Lee,
Hjertager, Fuhre, & Eckhoff, 1982). In their discussion of this finding, they noted, similar to
Chapman and Wheeler (1926) that “the rate of burning increases due to the larger flame surface
area, the flame induced flow velocity also increases, creating stronger flow field gradients”
generating turbulence and a feedback between the unburned pockets downstream of the obstacles
and the propagating flame brush (Moen, Lee, Hjertager, Fuhre, & Eckhoff, 1982). This feedback
loop between the unburned recirculating gases trapped between obstacles and the main flame
brush is one of the main mechanisms of flame acceleration and has been observed and studied by
other researchers as well (Ciccarelli & Dorofeev, 2008).
Figure 2.14 Flame time of arrive versus distance from ignition for various blockage ratios (BR)
and varying number of orifice plates. Figure credit: (Moen, Lee, Hjertager, Fuhre, & Eckhoff,
1982).
33
|
Colorado School of Mines
|
Figure 2.15 Schematic of turbulent flame propagation in an obstacle filled tube. Figure credit:
(Moen, Lee, Hjertager, Fuhre, & Eckhoff, 1982).
In general, the experimental results from Moen, Lee,Hjertager, Fuhre, & Eckhoff, 1982
showed that BR, the number of obstacles, obstacle spacing, and obstacle location can have a
huge effect on methane flame propagation and overpressure. Their results laid a strong
foundation for other researchers including Lee, Knystautas, and Chan (1985), Phylaktou and
Andrews (1991), and Fairweather, Hargrave, Ibrahim, and Walker (1999). Lee, Knystautas, and
Chan (1985) performed experiments in different steel cylinders ranging 5-30cm in diameter and
11-17m in length using a variety of fuels at different stoichiometries (including methane and air).
They determined there were four major propagation regimes. The first regime was the quenching
regime, in which the obstacle BR were large enough to initially accelerate the flame, but
eventually the flame was quenched due to heat loss to the cool unburned mixture entrained in the
turbulent flame brush (Lee, Knystautas, & Chan, 1985). The second regime was the choking
regime and the flame accelerates the entire length of the reactor, noting that eventually the flame
may reach the local speed of sound (Lee, Knystautas, & Chan, 1985). The third regime is the
quasi-detonation regime, where flame speeds are so fast and the orifice diameter is large enough
to transition to detonation which is dependent on the detonation cell size and reactor/obstacle
geometry (Lee, Knystautas, & Chan, 1985). Finally, the fourth regime is the C-J detonation
regime, and again the size of the orifice to the detonation cell size must be greater than several
times the cell size, which is dependent on the fuel mixture (Lee, Knystautas, & Chan, 1985).
These findings and discussion are extremely important for understanding the potential for
deflagration transition to detonation.
Other researchers have also experimented with obstacles that are staggered, in the center
of flow, and only on one side of the flow (Xiaoping, Minggao, Wentao, Meng, & Juniie, 2015).
34
|
Colorado School of Mines
|
They found that obstacles that are staggered or central to flow produce significantly higher
overpressures and the flame propagation itself is more tortuous.
Understanding the potential for a methane mixture to transition to detonation via obstacle
induced turbulence is the subject of much research in the coal mining industry (Oran, Gamezo
and Kessler, 2011 (Zipf, et al., 2013)). Although certain experimental setups and models have
shown that methane gas mixtures can reach detonation velocities, it is still unclear from the
studies discussed above, whether or not this is possible during a real mine explosion. For
example, in the UBB mine explosion in 2010, the pressure waves entrained coal dust on the floor
of the mine, which transitioned the methane explosion to a methane-coal dust explosion resulting
in 29 deaths and significant damage to miles of mine structures and equipment (Page, et al.,
2011). Because of this, there has also been some research on understanding how coal dust can
exacerbate a methane gas explosion and to determine better techniques for depositing inert rock
dust over coal dust (Bai, Gong, Liu, Chen, & Niu, 2011; Dong, Bi, & Zhou, 2012; Sapko, Weiss,
Cashdollar, & Zlochower, 2000).
In addition to using solid orifice plate type obstacles, researchers have also investigated
the effects of a grid, different shapes (cylinders, rectangles, etc), and porous media. Evans,
Schoen, and Miller (1948) looked at the effect of copper grids on propane-air flame phenomena
in pyrex tubes, varying in cross sectional shape and lengths. They found that grids accelerated
the flame and a “grid flame” consisted of “large cells or globules at the front, and fine-grained
eddies behind the front” as compared to a flame propagating in an unobstructed tube (Evans,
Schoen, & Miller, 1948). Stark and Roth (1989) investigated the development of a tulip flame in
a tube obstructed with a grid. They found that as the grid was moved further from ignition, the
tulip inversion continued to be affected by the obstacle until a certain distance and the inversion
was suppressed (Starke & Roth, 1989).
There have also been several studies investigating the effect of different solid shaped
obstacles on flame propagation (Ibrahim & Masri, 2001; Masri, Ibrahim, Nehzat, & Green, 2000;
Yibin, Fuquan, Xiaoyan, Xin, & Hongbin, 2011). Researchers looked at the effects of cylinders,
squares, diamonds, triangles, and wall/plates on flame and pressure enhancement and found that
changing the blockage ratio of the rectangular or plate type obstacles had the largest effect on
flame propagation velocity and overpressure (Ibrahim & Masri, 2001; Masri, Ibrahim, Nehzat, &
Green, 2000). Interestingly Masri, Ibrahim, Nehzat, and Green (2000) found significant
35
|
Colorado School of Mines
|
differences in the size and length of vortices produced downstream of the different obstacles as
shown in Figure 2.16, though they were unable to draw any significant conclusions about the
downstream vortices effects. In a second paper, however, results of the overpressure versus BR
in Figure 2.17 seem to suggest that off-spherical objects (wall/plate) increases the overpressure
of an explosion (Ibrahim & Masri, 2001). This is of significant interest for this research because
some coal mine explosions have originated near or within the gob, e.g. the Willow Creek
explosion in 2000, and so it will be important to understand how rock, which has a varying
degrees of sphericity, impacts flame propagation. Up to this point, the CSM group has been
unable to find other researchers investigating flame propagation across a rock pile. Due to the
lack of experiments, the main focus of this research is to perform experiments with idealized
rock and actual rock typically found in a mine to determine whether these shapes change flame
acceleration mechanisms.
Figure 2.16 Vortex pair behind a triangular obstacle. Image credit: (Masri, Ibrahim, Nehzat, &
Green, 2000).
Figure 2.17 Overpressure versus blockage ratio for different obstacle geometries. Figure credit:
(Ibrahim & Masri, 2001).
36
|
Colorado School of Mines
|
Finally, it has been shown that obstacles, confinement, and ignition location can
influence explosion overpressure and flame propagation. In a real longwall coal mine, the
explosion can occur from the gob area (confined) and propagate towards the face or entries.
Therefore, one of the main areas of concern of this research is flame propagation between
obstacles and this particular research has only been able to find one paper were the ignition was
between obstacles by vanWingerden and Zeeuwen, (1983). In their study, they had a square,
wooden plate with cylindrical sticks mounted on it. The obstacle was located in the middle of a
larger vessel such that the flame could propagate in all directions (similar to a spherical bomb).
Experiments were performed with ignition located 1) on the top of a single plate, 2) centered
between two plates, 3) on a single plate with sticks, and 4) between two plates with sticks
supported between them. For all fuels tested, including methane, the fastest flame velocity was in
configuration 4) and produced an overpressure twice that found for a stagnant mixture (van
Wingerden & Zeeuwen, 1983). This is of no surprise since the flame was not only confined, but
accelerated due to the obstacles. However, it must be noted that the flame propagated in all
directions, there was no principle direction like would be the case for a mine explosion where the
flame might tend to travel in the open passageways. Therefore, one of the objectives of this
research is to further understand the differences with flame propagation between obstacles in
addition to across obstacles.
Although most of the obstacles typically found in a mine are solid obstacles, the gob
area, walls, and rock on the belt has different levels of compaction and it is unclear whether they
can be treated as a porous media. Thus, one of the major focuses of this research is whether or
not the gob and other such obstacles act similar to a porous media. Because of this, it is
important to understand some of the differences in flame propagation mechanisms between
completely solid obstacles and porous media. Some of the most fundamental studies on porous
media were performed by Babkin, Korzhavin, and Bunev (1991) and Howell, Hall, and Ellzey
(1996). Babkin, Korzhavin, and Bunev (1991) performed experiments in square reactors filled
with different porous media including steel polished balls, polyurethane foam, porous material
made of foam, and aluminum glued combs. They show that methane flames are much faster at
higher initial pressures and pore cavity size. However, it is unclear whether the thermal
properties of the material had an effect. Additionally, they show that porous media can accelerate
flames “as effectively as in rough tubes, or in tubes with periodical obstacles such as spirals”
37
|
Colorado School of Mines
|
(Babkin, Korzhavin, & Bunev, 1991). This is extremely important because the walls of a mine
consist of rock rubble and as such will have a significant effect on flame propagation. Another
notable finding of Babkin, Korzhavin, and Bunev (1991) and more fully explained by Howell,
Hall, and Ellzey (1996) is the feedback mechanism between the flame and porous media. As the
flame propagates through the porous media, pockets of unburned mixture are trapped and then
heated by radiation and conduction. These pockets are entrained by the passing flame, thus
accelerating the flame in a feed-back loop (Howell, Hall, & Ellzey, 1996). Babkin, Korzhavin,
and Bunev (1991) also note that “the most probable stabilizing factor” is quenching of the flame
(Babkin, Korzhavin, & Bunev, 1991).
Further studies by Ciccarelli, Hlouschko, Johansen, Karnesky, and Shepherd (2009)
experimented with a layer of 12.7mm diameter ceramic-oxide beads along the entire bottom of a
horizontal, rectangular reactor investigating how bead layer affects the transition to DDT. They
helped confirm that the interaction of the flame with the bead layer “drives the flame
acceleration in the gap until DDT” (Ciccarelli, Hlouschko, Johansen, Karnesky, & Shepherd,
2009). Also they found reducing the height of the gap above the bead layer resulted in a faster
acceleration to DDT because the trapped gases in the bead layer have a stronger coupling to the
main flow and may act similar to a piston (Ciccarelli, Hlouschko, Johansen, Karnesky, &
Shepherd, 2009).
Fig (2019) performed similar experiments to Cicarelli, et al. (2009) in different diameter
reactors using layers of rock. In these experiments, rock rubble was arranged in a non-reacting
metal cage and inserted into a reactor at different locations. The length and height of the rock
pile were explored and it was found that lining the entire length of the reactor resulted in the
greatest flame acceleration for large vessels. This research aims at taking a step back and using
spheres as an idealized rock rubble. The spheres take away any different in flame propagation
due to surface topology so that this research can determine which property of the rock pile has
the largest impact on flame acceleration.
2.7 Impact of spark energy on methane flame propagation
There are many sources of ignition in a longwall coal mine such as rock-on-rock friction,
metal-on-rock friction (from a carbide cutting tips on the longwall shearer), spontaneous
combustion, welding components, or any electrical communication system/remote. Frictional
ignition from rocks and metal-on-rock friction are two of the most common methods of ignition
38
|
Colorado School of Mines
|
(Page, et al., 2011) and are difficult to quantify since the energy is highly dependent on the type
of rock, the speed of the rock or machine, and contact point among some. This research aims to
understand how these different types of energy and energy durations may affect methane gas
deflagrations using the experimentally validated CFD model.
In 1979, Maly and Vogel wrote a paper describing the three discharge modes of an
electric spark, detailing the properties of each mode and where the majority of losses come from.
The three discharge modes are the breakdown phase, arc phase, and glow discharge phase as
shown in Figure 2.18. The energy supplied by the spark system during the breakdown phase
ionizes the gas between the electrodes, creating a plasma. Approximately 94% of the energy
during the breakdown phase is transferred to the plasma which helps provide the “conductive
path between the electrodes necessary” for the arc and glow discharge phases (Maly & Vogel,
1979). During the arc and glow discharge phases, a significant amount of energy is lost through
conduction to the electrodes and there is less dissociation of species (Maly & Vogel, 1979).
Understanding these processes is important because it has helped researchers design more
efficient methods of spark ignition, such as the plasma spark plug which is designed to transfer
the majority of the energy during the breakdown phase.
Figure 2.18 Schematic of voltage and current versus time of a typical spark ignition system. U
refers to the energy in unit volts (V). Figure credit: (Maly & Vogel, 1979).
39
|
Colorado School of Mines
|
Most studies of premixed combustion employ spark ignition, especially for determining
the minimum ignition energy required to ignite mixtures (Blanc, Guest, von Elbe, & Lewis,
1947; Shmelev, 2009). However, there are several parameters of the design of the spark ignition
system which can affect measurements of minimum ignition energy, including the electrode
material, distance between the electrodes, and the shape of the electrodes, whether they are
plates, hemispherical, or to a point. In general, most researchers agree that for a stoichiometric
mixture of methane and air at 1atm and 300K, the minimum ignition energy is approximately
0.3-0.5mJ (Shmelev, 2009; Turns, 2012). As discussed earlier, the minimum ignition energy is
highly dependent on pressure, temperature, and mixture stoichiometry.
In majority of the studies on the effects of spark ignition on flame propagation,
researchers note that the ignition energy only affects the initial flame kernel development until a
critical radius (Lintin & Wooding, 1958). To note, many of these researchers used small test
chambers or spherical bombs, making it difficult to determine if ignition energy does not affect
flame propagation in other apparatus (Blanc, Guest, von Elbe, & Lewis, 1947; Lintin &
Wooding, 1958). However, some researchers have found that for lean limits and high spark
energies, ignition may affect the subsequent flame propagation, suggesting that researchers
“perform experiments with several ignition energies to determine conditions at which the flame
is not affected” (Lawes, Sharpe, Tripathi, & Cracknell, 2016). Additionally, other researchers
have found that a planar ignition of match heads produces overpressures and flame speeds faster
than a single point ignition (Hjertager, Fuhre, & Bjorkhaug, 1988). Though the ignition source is
a match head, it still lends to the point that the distribution of energy and quantity of energy
could affect flame propagation. Therefore, it is the goal of this research to determine whether
varying spark ignition energy and duration can affect methane gas deflagrations, especially for
confined spaces.
2.8 Modeling methane flames
There are three major ways to model complex fluid flow problems: direct numerical
simulation (DNS), large eddy simulation (LES), and moment models. As the name implies, DNS
is the most accurate method of simulating fluid flow because it directly solves the Naiver-Stokes
equations. Unfortunately, the major drawback to DNS is that it requires a significant amount of
computational time and energy due to the wide range of temporal and spatial scales of most
flows, which makes it less desirable as compared to LES and moment models. LES models
40
|
Colorado School of Mines
|
resolve the large-scale turbulent structures directly by solving the filtered Navier-Stokes
equations and models the smaller scales (sub-grid scales – SGS). In general, LES models are
good for turbulent flows that are characterized by large-scale eddies, allowing for a much coarser
grid. Finally, the third type of models, which are the most widely used, are moment models that
solve the fluid flow equations using statistical analysis. The most common method of averaging
the Naiver-Stokes equations is by using ensemble averages which solve for mean velocities
producing a set of equations called Reynolds-Averaged Navier-Stokes equations (RANS) and
uses 1st order closure models to obtain a complete set of equations.
Many researchers over the decades have developed various models of premixed methane-
air (or other mixtures) deflagrations and detonations using DNS (Gonzalez, 1996; Hawkes &
Chen, 2004), LES (Bi, Dong, & Zhou, 2012; Chen, Luo, Sun, & Lv, 2017; Di Sarli, Di
Benedetto, & Russo, 2012; Xiao, Makarov, Sun, & Mokov, 2012), RANS (Fairweather, Ibrahim,
Jaggers, & Walker, 1996; Fairweather, Hargrave, Ibrahim, & Walker, 1999; Fig, Bogin, Brune,
& Grubb, 2016; Jerome, Christophe, & Guillaume, 2017; Kozubkova, Krutil, & Nevrly, 2014;
Wang, Ma, Shen, & Guo, 2013), and other reduced order models (Sezer, Kronz, Akkerman, &
Rangwala, 2017). Some researchers have written in-house codes (Catlin, Fairweather, &
Ibrahim, 1995; Fairweather, Ibrahim, Jaggers, & Walker, 1996) while most use commercial
softwares such as Fluent (Gutkowski, 2013; Kozubkova, Krutil, & Nevrly, 2014) or AutoReaGas
(Hong, Lin, & Zhu, 2016; Zhang & Ma, 2015). In general, most researchers are concerned with
accurately solving the flame time of arrival and the overpressure of the explosions for gas
explosions with and without obstacles. To accurately model these phenomena, the majority of
researchers performed their own experiments to help validate their model (Dunn-Rankin &
McCann, 2000; Fairweather, Hargrave, Ibrahim, & Walker, 1999; Jerome, Christophe, &
Guillaume, 2017; Kozubkova, Krutil, & Nevrly, 2014; Yu, Sun, & Wu, 2002), but there are
many researchers who rely on experiments from other researchers as points of model validation
(Catlin, Fairweather, & Ibrahim, 1995; Di Sarli, Di Benedetto, & Russo, 2012; Hong, Lin, &
Zhu, 2016; Li & Hao, 2017; Oran, Gamezo, & Kessler, 2011; Salvado, Tavares, Teixeira-Dias,
& Cardoso, 2017; Sezer, Kronz, Akkerman, & Rangwala, 2017; Valiev, Bychkov, Akkerman,
Law, & Eriksson, 2010).
To build a comprehensive 2D & 3D combustion model of a longwall coal mine methane
gas explosion, it is important to be able to capture the effects of obstacles on flame propagation
41
|
Colorado School of Mines
|
and pressure enhancement. Figure 2.19 is a summary of the numerical simulations of premixed
combustion which model obstacles versus those which do not (Note this is not meant to be an
entire review of all flame experiments ever performed, but a summary of those discussed in this
manuscript). As can be seen, there are less researchers modeling obstacles, though it is becoming
much more common in recent years, especially with the improvements made to commercial
codes, allowing for much easier meshing around complex objects and coupling of more complex
chemical reactions. However, most research groups modeling obstacles consider only a couple of
obstacles of simple geometry as shown in Table 2.1. This research not only models a larger
variety of obstacles, but validates the model across a much wider range of reactor volumes.
Fairweather, et al. 1996 & 1999 have modeled premixed combustion and interaction with
obstacles and validated the model across different reactor sizes (Fairweather, Ibrahim, Jaggers, &
Walker, 1996; Fairweather, Hargrave, Ibrahim, & Walker, 1999). However, their code was
developed in-house which makes it more difficult to replicate their settings. Additionally,
although they have validated their model with a handful of reactors, the range is still limited
when it is compared to all other experimental combustion researchers, Figure 2.20 and Figure
2.21.. This is extremely important because flame dynamics, especially across obstacles, do not
scale linearly with reactor size (Fig, Bogin, Brune, & Grubb, 2016). This also helps demonstrate
how novel and important this research is to the combustion community; modeling the effects of
obstacles across a wide range of scales and validating the model with experiments is a small
percentage of the total number of experimental and numerical combustion research. It is also
important to note that this research was the first to couple a combustion model with a ventilation
model of a longwall coal mine and model a methane gas explosion (Bogin, 2015). Although
there are still improvements to be made, the coupled model developed in this research is not only
applicable to the mining industry, but to other small- and large-scale industrial explosions or
even to flame propagation through a building.
42
|
Colorado School of Mines
|
Health Administration (MSHA). Although the number of incidents and fatalities from coal mine
disasters have declined over the years due to better safe practices and equipment enhancements,
underground coal mine methane gas explosions still pose a risk to miners and equipment.
Figure 2.22 Number of coal mine disasters and fatalities from 1900-2016. Figure credit: (Center
for Disease Control and Prevention, 2017).
Methane gas explosions have several mechanisms of blast injury which affect the human
body including primary, secondary, tertiary, and quaternary effects. Primary injuries are injuries
from the initial blast shock wave interacting with the human body. The most common primary
injuries from coal mine explosions are blast lung, eardrum/middle ear rupture and concussion,
but other injuries may include eye rupture or abdominal hemorrhage (Centers for Disease
Control and Prevention, 2017; Institute of Medicine, 2014). As the blast wave travels through the
body and blood, it can cause air to expand in the blood, creating air embolus which in many
cases, end up in the lungs (blast lung) (Institute of Medicine, 2014). If the victims do not seek
medical attention right away including a chest x-ray, they may not know they have blast lung
until it is too late. Secondary injuries may result from airborne debris that is thrown from the
explosion, which may result in blunt force trauma or perforation of the body (Centers for Disease
45
|
Colorado School of Mines
|
Control and Prevention, 2017; Institute of Medicine, 2014). Tertiary injuries are a result of
individuals being through from the blast and include bone fractures and head injuries (Centers
for Disease Control and Prevention, 2017; Institute of Medicine, 2014). For example, in the
Willow Creek mine explosion of 2010, several miners close to the second explosion were thrown
and one miner lost their cap lamp (McKinney, et al., 2001). Finally, quaternary injuries are those
which are not a direct result of the blast, but are indirect including burns or inhalation of toxic
fumes such as smoke, dust, or carbon monoxide (Centers for Disease Control and Prevention,
2017). Again, in the Willow Creek mine explosion one miner “was asphyxiated as a consequence
of carbon monoxide poisoning” and another miner was “seriously burned and received a massive
head injury” (McKinney, et al., 2001).
However, in addition to injuries to miners, blast effects of methane gas explosions can
damage/destroy mine structures, damage mine equipment, and reverse airflow in the mine
(Brune, 2014; McKinney, et al., 2001; Page, et al., 2011). For example, pressures between 0.1-
5psi can shatter single-strength glass and pressures between 1-2psi can crack plaster walls or
buckle sheet steel (Owen-Smith, 1981). Also, pressures between 2-3psi can crack cinder-block or
concrete block walls and upwards of 8psi they can crack brick wall (Owen-Smith, 1981). These
pressure examples are important because the demonstrate the overpressure effects on common
mine materials which the pressures from the UBB explosion in 2010 easily exceed (Page, et al.,
2011). Also important is the pressure wave fluctuations which may reverse airflow in the mine or
possibly temporarily stop ventilation fans which can be dangerous as CO levels increase from the
explosion.
In summary, there are many different types of hazards from a methane gas explosion in
an underground longwall coal mine. Although advancements throughout the years have helped
reduce the number of fatalities from these explosions, they still pose a risk in active longwall
coal mines. It is the goal of this research to better predict these hazards and understand the
potential for human loss and mine damages in order to build stronger mitigation strategies for
improved worker safety.
46
|
Colorado School of Mines
|
CHAPTER 3
DESIGN OF EXPERIMENTAL COMBUSTION REACTORS
Due to safety hazards of performing methane-air explosions in a mine, this research
performs experiments in both large- and small-scale reactors of varying diameter and length. The
large-scale reactor is 71cm in diameter, 6.1m in length (L/D=8.5, Volume = 2.4m3) and is
located at Edgar Experimental Mine in Idaho Springs, CO. The benefit of performing
experiments in the large-scale reactor is that the resulting methane flame speeds and pressure rise
are more indicative of those found in a longwall coal mine explosion. However, this reactor
requires large amounts of gases which increases cost and requires significant amounts of time to
set up and perform experiments. Therefore, small-scale experiments are also performed to
provide additional insights across a wide range of conditions. There are several small-scale
reactors including 1) 5cm diameter, 43cm long (L/D=8.5), 2) 5cm diameter, 1.5m long (L/D=30)
steel reactor 3) 9.5cm diameter, 81cm long (L/D=8.5), steel reactor, a 4) 12cm diameter, 1.5m
long (L/D=12) quartz reactor, 5) 13.6cm diameter, 1.5m long (L/D=11) quartz reactor, and a 6)
30.5cm diameter, 1.15m long (L/D=8.5) steel reactor. Additionally, a small experimental box,
51x34x15cm (LxWxH), was setup to explore the impact of reactor shape on methane flame
propagation and interaction with a simulated gob. The main advantage of performing
experiments at different scales is it allows researchers to understand the scalability of methane
flame properties which is important for validating the combustion model before incorporation
into the mine-scale ventilation model. The scalability of methane flame behavior is one of the
major objectives of M. Fig’s research (Fig, 2019) and will be presented in part of this
dissertation.
This document will present the work performed in the 12cm diameter quartz flow reactor,
71cm diameter large-scale reactor, and laboratory-scale experimental box. The main advantage
of the quartz reactor and experimental box is that they allow full visualization of the flame as it
interacts with rock rubble and other obstacles. This is important because it provides further
insight into methane flame dynamics and allows researchers to use imaging to validate the CFD
models.
47
|
Colorado School of Mines
|
3.1 Laboratory-Scale Experimental Setup and Procedure
The laboratory-scale experimental system consists of industrial-grade compressed
methane and zero-grade air, mass flow controllers, a mixing tank, flame arrestors, a reactor, ion
sensors, a pressure transducer, an ignition system, and data acquisition system as shown in
Figure 3.1.
Figure 3.1 Flow diagram of laboratory experimental setup. NI USB refers to a National
Instruments USB used for recording data.
Before any experiments are performed, compressed building air flows through all lines to
check for any possible leaks. After all safety checks have been performed, the experimental
process begins by flowing methane and air at specified rates for a desired stoichiometry. The
flow is controlled by mass flow controllers (MFC); a Bronkhorst EL-FLOW® Select (0-50
SLM) controls the air and an Alicat Scientific MC Series (0-5 SLPM) controls the methane. The
methane and air mix enter a mixing chamber, which is filled with turbulence inducing media,
until homogeneity is reached. Before the mixture flows into the quartz reactor, an aluminum foil
cover is placed over the open end of the reactor. A small perforation is made in the foil so that
48
|
Colorado School of Mines
|
while filling the reactor is purged of the ambient air by the premixed methane-air and provides
an undiluted mixture without increasing the overall pressure in the reactor. Finally, the premix
flows into the quartz reactors at 30psi for 5 minutes which is equivalent to approximately 2-3
volume fills depending on the reactor size. Homogeneity of the mixture inside the reactor is
verified using an Infrared Industries IR-6000 gas analyzer and mixture stoichiometry is
confirmed using a gas chromatograph with thermal conductivity detector (GC-TCD). At the
closed end of the reactor, where the premix inlet is located, there is no stratification in the radial
direction. Along the top of the reactor, the methane-air mixture did not vary in concentration;
near the open end of the reactor the mixture varies in the radial direction by less than 0.2%
(within error of the MFCs). After the reactor is completely filled with the methane-air mixture,
the mixture is allowed to settle for 40 seconds to insure stagnant conditions. The mixture is then
ignited using a capacitive spark ignition system as described in Section 3.3, providing
approximately 60±5mJ of energy to the system. After each experiment, compressed building air
flows through the reactor to 1) cool down the reactor to ambient conditions and 2) to flush out
any remaining combustion products. During the shutdown procedure all lines are purged to the
laboratory exhaust system and compressed building air flows through the lines to ensure no
residual methane or premix is left in the system. Note all experiments are performed at 294±1K
and 83±1kPa, which are ambient atmospheric conditions for Golden, CO – elevation
approximately 1,730m above sea level.
The DAQ used in this system consists of a National Instruments (NI) USB-6008 and
USB-6009 DAQ board, both capable of sampling at 48,000 samples per second. The ion sensors
are each wired to a single input on the USB-6008 board and a Kistler © 4260 piezoresistive
pressure transducer is wired separately to the USB-6009 board. Additionally, high-speed
imaging is taken using a GoPro Hero4 capable of sampling at 240 frames per second at 720 pixel
resolution. In order to ensure repeatability, each experimental set consists of 4-5 experimental
runs. The flame front propagation velocities reported are the average of all experimental runs and
the error bars represent the standard deviation of the mean of the set. The overpressure traces
shown are a single run from an experimental set and are meant to show the general trend of the
explosion overpressure. The maximum overpressure reported is always the average of all the
experimental runs and the error bars represent the standard deviation of the mean of the set.
49
|
Colorado School of Mines
|
Finally, there are several safety design features in the laboratory-scale experimental setup. The
compressed methane and zero-grade air are housed in an exhaust cabinet. High-pressure tubing is
routed from the regulators to the electronic solenoid valves as shown in Figure 3.1. The solenoid
valves are normally closed so that in the event of an emergency, all flows are stopped.
Additionally, these valves are wired to an emergency stop button so that the entire system can be
shut down. Downstream of the solenoid valves, stainless steel tubing is used and can withstand
pressures up to 35MPa which is significantly greater than pressures used in this system.
Upstream of the mixing chamber is a pressure relief valve that is set to relieve at pressures
greater than 35kPa and evacuate the gases to the laboratory exhaust system in case of an
accidental overpressure. Downstream of the mixing tank is a 3-way electronic solenoid valve,
which is wired to the operator’s table and either fills the reactor or purges the premix into a high-
pressure line vented in the exhaust cabinet. Two flame arrestors are employed; one flame arrestor
is located in the reactor fill line, upstream of the premix inlet and a second flame arrestor is
located in the purge line. Additionally, exhaust hoods connected to a point exhaust system (150
cfm) are located on either end of the quartz reactor and a methane sensor is located near the open
end and can measure concentrations less than 25% of the lower explosive limit (LEL) of
methane. Another safety design feature is the Plexiglas enclosure around the quartz reactor
which 1) protects users in the case of cracking or breaking of the quartz and 2) keeps any leaking
premix inside the system so that it can be exhausted from either the open or closed end of the
system. Lastly, there are three fire extinguishers placed throughout the laboratory; one is near the
operator’s table, a second near the exhaust cabinet housing the compressed methane and zero-
grade air, and a third near the laboratory exit. Safety is the number one concern of this lab and
the measures described have been evaluated by the Environmental Health and Safety office on
campus and exceed requirements. Additionally, the Standard Operating Procedures (SOPs) for
operating the gas flame tubes and spark ignitions systems have been created, signed, included in
a binder in the lab, and taped to the operator’s table and gas cabinet.
3.1.1 Design of Quartz Flow Reactor
The horizontal, cylindrical quartz flame rector used in this study has an inner diameter of
12cm, 0.25cm wall thickness, and length of 1.5m as shown in Figure 3.2 and Figure 3.4. One end
of the quartz reactor is open to atmosphere and the other end is closed. The premix fuel-air inlet
and pressure transducer are mounted on the closed end of the reactor as shown in Figure 3.3,
50
|
Colorado School of Mines
|
Figure 3.4 Image of the 12cm diameter quartz flow reactor laboratory setup.
There are several techniques to measure the propagation velocity including IR/UV
sensors, pressure sensors, ion sensors, and high-speed imaging and many researchers use a
combination of sensors in order to check the reproducibility of an experiment. For example,
Fairweather, et al. (1999) and Robinson and Wheeler (1933) used imaging to determine flame
time of arrival. Many other researchers, including Hjertager, Fuhre, and Bjorkhaug (1988),
Ciccarelli et al. (2009), and Moen, et al. (1982) used ion sensors to measure the flame speed.
This research uses high-speed imaging and ion sensors made from copper wire as shown in
Figure 3.5. The electrodes are housed in a ceramic tube and placed into the sensor ports such that
the electrodes are flush with the top of the reactor. The sensor ports are 0.8cm in diameter and
the ceramic tubes are 0.5cm in diameter. Therefore, silicone was wrapped around the ceramic
tubes to 1) hold the sensors in place and 2) not allow any premix or combustion gases to escape
out of the top of the tube. The IR gas analyzer was used to verify no gases escape from the tube
via the sensor ports.
The ion sensors were wired to a voltage source and resistor network as shown in Figure
3.6. As the flame passes across the electrodes, the ions trigger a voltage drop across the resistor
which is recorded by the DAQ system. There are three main components to the design of the ion
sensors: the wire, the resistor, the voltage source. A study was done using 18 gage copper wire of
varying lengths and compared the results to using coaxial cable which has significantly more
insulation. The additional insulation on the coaxial cable reduced the standard deviation of the
recordings, but did not impact the results.
52
|
Colorado School of Mines
|
Figure 3.5 Image of custom-made copper ion sensors (left) and example of raw output from ion
sensors (right).
Figure 3.6 Ion sensor circuit.
The vertical location of the ion sensor is extremely important and by performing
experiments in the 12cm diameter quartz reactor researchers took high-speed imaging to
determine the actual location of the flame front so that reported flame front propagation
velocities are accurate. As can be seen in Figure 3.7, when ignition is 11cm from the open end
the flame travels towards the closed end and the flame front is at the top of the reactor.
Therefore, for all experiments when ignition is from the open end the ion sensors are placed at
the top of the reactor. However, when ignition is 11cm from the closed end, the flame shape is
much different as shown in Figure 3.8. Results of testing the location of the ion sensor in the
vertical direction for ignition from the closed end are shown in Figure 3.9. As can be seen in this
figure, when the sensors are placed at the top of the reactor, the flame front propagation
53
|
Colorado School of Mines
|
velocities recorded seem step-wise. This is because the flame expands more rapidly in the axial
direction than in the radial direction. Thus, by placing the sensors further down into the reactor,
at 1cm as shown in Figure 3.8, the recorded flame front propagation velocities are more accurate
and the standard deviation of the mean is reduced significantly. In conclusion, for all closed-end
ignition experiments the ion sensors are placed 1cm down from the top of the reactor. This study
was extremely important because researchers are unable to visualize the flame shape in all the
other explosion reactors, including the 71cm diameter large-scale reactor at Edgar Experimental
Mine.
Figure 3.7 Flame front shape when ignition is at the open end of the 12cm diameter quartz
reactor. Flame propagates from the open end to the closed end (left to right). CH = 9.5±0.3%.
4
Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
ign
Figure 3.8 Flame front shape when ignition is at the closed end of the 12cm diameter quartz
reactor. Flame propagates from the closed end to the open end (right to left). Yellow line
represents the top of the quartz reactor. CH = 9.5±0.3%. Operating conditions 294±1K,
4
83±1kPa. E =60±5mJ. One (1) relief hole.
ign
54
|
Colorado School of Mines
|
Figure 3.9 Impact of vertical location of ion sensor on methane flame front propagation velocity
for a closed-end ignition. Ignition location: 1.39cm from the open end. Flame travels towards
open end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief
4 ign
hole.
3.1.2 Design of Experimental Box
Most of the experimental setups conducted for this research have been horizontal
cylindrical reactors to understand the impact of mine conditions on the bulk flow of a high-speed
methane gas explosion. However, in a real mine environment the entryways, crosscuts, longwall
face, etc. are rectangular in shape. To begin to understand how the reactor shape and multiple
pathways affect methane gas flame propagation velocities an experimental box was setup as
shown in Figure 3.10. The experimental box is made of steel and has a Plexiglas covering on top
in order to take high-speed imaging of the flame during experiments. The box is 51x34x15cm
(LxWxH) and in the center of the box is a simulated porous gob (porous medium) consisting of 2
layers of lava rock, average size 6.1±0.2cm x 4.4±0.2cm x 3.2±0.2cm, and 2 upper layers of river
rock, average size 6.1±0.3cm x 4.2±0.2cm x 2.7±0.1cm, arranged in a non-reacting metal cage
that is 28x22x15cm (LxWxH). Although lava and river rock are not commonly found in a mine
environment, they were chosen due to ease of availability and due to the fact they have similar
thermal properties of rock in underground mines (thermal conductivity is between 1.5-5 W/m-
K), along with irregular shapes meant to represent the complex pathways between the rocks in a
gob.
55
|
Colorado School of Mines
|
Figure 3.10 Image of experimental box setup with a simulated gob (porous medium) consisting
of lava rock and river rock.
As can be seen in Figure 3.10, a small opening 7x12cm (HxW) was made in the bottom-
left corner of the box which allowed for explosion relief. The experimental procedure is the same
as for the quartz reactor described earlier in this section. In summary, the box is filled with
premixed methane-air mixture and allowed to settle before ignition using a capacitive discharge
electric spark providing 60±5mJ of energy. Note that the spark electrodes were located at a
height of H=7.5cm from the bottom of the box (total height is H=15cm). In this experimental
box setup, ignition was initiated near the opening (open-end ignition at x=0.06m, y=-0.14m) in
the bottom-left corner of the box and conversely, opposite the opening (closed-end ignition at
x=0.45m, y=0.14m) in the top-left corner of the box. High-speed imaging was taken using a
GoPro Hero4 capable of sampling at 240 frames per second and 720 pixel resolution and average
flame speeds and standard error of the mean were estimated using video recordings. All
experiments were performed using a methane-air mixture stoichiometry of 9.5±0.3% by volume.
Experiments are performed at 293±1K and 83±1kPa, which are ambient atmospheric conditions
for Golden, CO.
3.2 Design of Large-Scale Experimental Reactor
The large-scale experimental system consists of industrial-grade compressed methane,
zero-grade air, industrial grade compressed nitrogen, mass flow controllers, NEMA 7 explosion
proof solenoid valves, a mixing tank, flame arrestors, a reactor, ion sensors, pressure transducers,
an ignition system, and data acquisition system as shown in Figure 3.11, Figure 3.12, and Figure
56
|
Colorado School of Mines
|
Figure 3.13: Image of the closed end of the 71cm diameter, 6.1m long steel reactor showing the
mixing tank, pressure relief valve, premix access ports, flame arrestors, and sensor ports and
wiring.
Before beginning the experimental process, an aluminum foil cover is placed over the
open end of the reactor as shown in Figure 3.13 and a steel grid is placed over the end to
suppress the flame exiting the reactor. Additionally, the reactor is angled towards a rock burm
for safety concerns and to help redirect noise away from the city of Idaho Springs. A small
perforation is made in the foil so that while filling the premix is enclosed in the reactor without
increasing the overall pressure in the reactor. Similar to the laboratory-scale reactor system, the
experimental process begins by flowing methane and air at specified rates for a desired
stoichiometry. The flow is controlled by mass flow controllers (MFC); a Bronkhorst EL-
FLOW® Select (0-50 SLM) controls the methane and a MKS 1559A controls the air and
nitrogen. The methane and air are delivered at 70psi and enter a mixing tank, which includes a
pressure relief valve and a NEMA 7 explosion proof solenoid valve connected to a purge line for
safety. The mixture then flows into two lines on which are flame arrestors are attached in case of
flashback. Finally, the premix then flows into the reactors for at least 45 minutes which is
equivalent to approximately 1.5-2 volume fills. Homogeneity of the mixture inside the reactor is
58
|
Colorado School of Mines
|
verified using an Infrared Industries IR-6000 gas analyzer on site. After the reactor is completely
filled with the methane-air mixture, the mixture is allowed to settle for 30 seconds to insure
stagnant conditions. During this settle time, air flows through all the lines and through the gas
mixture tank, purging to the outside to ensure no premix is inside the system. The mixture inside
the reactor is then ignited using a capacitive spark ignition system as described in Section 3.3,
providing approximately 60±5mJ of energy to the system. Note that the spark electrodes are
located in the first sensor port, which is 28.5cm from the closed end. During the shutdown
procedure all lines are vented to the outside and compressed air flows through the lines to ensure
no residual methane or premix is left in the system. Note all experiments are performed at
295±1K and 79±1kPa, which are ambient atmospheric conditions for Idaho Springs, CO.
The DAQ used in this system consists of a NI CDAQ and NI USB-6009 DAQ board. The
ion sensors are wired in parallel to the NI CDAQ and two Kistler © 4260 piezoresistive pressure
transducers are wired separately to the USB-6009 board. The ion sensors are located in the center
of the reactor and reported flame front propagation velocities are the average of all experimental
runs and the error bars represent the standard error of the mean of the set. The overpressure
traces shown are a single run from an experimental set and are meant to show the general trend
of the explosion overpressure. The maximum overpressure reported is always the average of all
the experimental runs and the error bars represent the standard deviation of the mean of the set.
Finally, there are several safety design features in the large-scale experimental setup. The
compressed gases and flow controllers are housed away from the reactor and are wired
underground in a PVC pipe back to the control center. This allows the operator to run the entire
system from the control center, in which a fire extinguisher is located. High-pressure tubing is
routed from the regulators to the NEMA 7 rated electronic solenoid valves and the solenoid
valves are normally closed so that in the event of an emergency, all flows are stopped.
Downstream of the solenoid valves, stainless steel tubing is used and can withstand pressures up
to 35MPa which is significantly greater than pressures used in this system. Upstream of the
mixing tank is a pressure relief valve which vents to the surrounding atmosphere. Downstream of
the mixing tank is a NEMA 7 rated electronic solenoid valve, which is wired to the operator’s
table and either fills the reactor or purges the premix into a high-pressure line vented to the
outside. Two flame arrestors are employed downstream of the mixing tank before the premix
enters the reactor. Safety is the number one concern of this research and the measures described
59
|
Colorado School of Mines
|
have been evaluated by the Environmental Health and Safety office on campus and exceed
requirements.
3.3 Design of Spark Ignition System
For this manuscript, a spark ignition was be employed since it is one of the most common
methods of igniting a combustible mixture due to its reliability, ease of setup, and ease of
implementation (Turns, 2012). The current spark ignition system consists of a voltage source, a
1Ohm resistor, a 2μF capacitor, a manual switch, and single ignition coil as shown in Figure
3.14. To produce a spark, the manual switch is activated so that the current builds up in the
capacitor, storing the spark energy. After the switch is deactivated, the energy in the capacitor is
released and a magnetic field breakdown occurs between the capacitor and ignition coil. The
ignition coil acts similar to a transformer and steps up the voltage across the spark electrodes.
After reaching a sufficient voltage breakdown across the electrodes, which is dependent on the
mixture composition and stoichiometry, a spark is produced and ignition is initiated.
Figure 3.14 Schematic of spark ignition system. The primary side includes the 12V battery,
1Ohm resistor, fuse, capacitor, and manual switch. The secondary side includes the output from
the ignition coil and spark electrodes.
It is important to note that typically the battery, resistor, switch, and capacitor are called
the primary side. This primary side is low voltage (12V), but has a high current (5A). The
ignition coil has primary and secondary windings with a typical turns ratio of almost 100:1. The
60
|
Colorado School of Mines
|
low voltage and high current from the primary side flow through the primary windings and a
magnetic field builds up between the primary and secondary windings, such that during the spark
discharge process, the low voltage is stepped up to 10,000V or more and the current is reduced to
milliamps.
To estimate the amount of spark energy, it is assumed that the amount of energy stored in the
capacitor is completely transferred to the spark. In reality, some of the stored energy is dissipated
during the dielectric breakdown across the electrodes and other processes such as heat
dissipation to the electrodes. The amount of energy stored in the capacitor can be estimated by
the product of the primary current (i ) and voltage source (V), Equation (3.1):
p
(3.1)
The measured primary spark current is 5.2 ± 0.2A and measured battery voltage is approximately
12 ± 0.5V which corresponds to spark energy of 62.4 ± 1.7mJ; however, over time this may
fluctuate slightly and researchers have recorded energies 60 ± 5mJ. The spark current is
measured using an oscilloscope and the voltage is measured using a voltmeter.
In summary, a spark ignition system capable of handling currents upwards of 30A,
secondary voltages in excess of 20,000V, and spark durations between 0.5-6.5ms has been
designed. The configuration of the spark ignition system is shown in Figure 3.14 includes a 12V
battery, 1Ohm resistor, 2μF capacitor, a 30A fuse upstream of the ignition coil, and an ignition
coil, supplying approximately 60 ± 5mJ of energy through the spark. Safety features of the
system include no. 10AWG multicore copper wiring, a 30A fuse, a separate plastic box for the
ignition coil in case of overheating, a handlebar switch on the positive junction to the battery, a
secondary switch to activate the system, a thermometer in case components overheat, and a large
plastic container with a Plexiglas lid to allow visualization of components. After each
experimental session, the circuit is disconnected from the battery and ignition coil and all
components are cooled down to ambient.
61
|
Colorado School of Mines
|
CHAPTER 4
EXPERIMENTAL INVESTIGATION OF METHANE FLAME PROPAGATION AND
PRESSURE GENERATION
Previous related experimental research was performed with piles of rock rubble to help
simulate a real longwall coal gob as shown in Figure 4.1 (Fig, Strebinger, Bogin, & Brune, 2018;
Strebinger, et al., 2017). Results from these experiments have shown that for all reactors and
across all scales the rock pile, meant to represent fallen rock rubble in the gob, accelerated the
methane flame for all stoichiometries investigated as shown in Figure 4.2. Additionally,
increasing the length of the rock pile further accelerated the flame and had the most significant
increase for the largest reactor. However, as can be seen from Figure 4.1, the rock piles have
varying rock material, rock orientation, rock size, rock pile porosity, void spacing between rocks,
and void location. These variabilities make it difficult to determine what gob parameters had the
greatest impact on increasing the flame front propagation velocities. Therefore, one of the main
objectives of this research is to gain a strong, fundamental understanding of the impact of
different gob parameters on methane flame propagation.
Figure 4.1 Image of methane flame passing across rocks in the 13.6cm diameter quartz reactor
(top) and 71cm diameter steel reactor at Edgar Mine (bottom). Image credit: (Strebinger, et al.,
2017).
62
|
Colorado School of Mines
|
Figure 4.2 Impact of rock pile length in the 13.6, 30.5 and 71cm diameter reactors. The barrier
lengths were varied from 3 to 7% of the overall reactor length. The impact of the length of the
barrier is larger for larger reactors than for smaller reactors. Ignition from the open end of the
reactor. Operating conditions 294K, 83kPa. Figure credit: (Fig, 2019).
4.1 Empty 12cm Diameter Reactor: Impact of mixture stoichiometry on open- and closed-
end ignition
Methane can emanate from various sources in the longwall coal mine, including the gob
area, face, walls, etc. among some (Karacan, Ruiz, Cote, & Phipps, 2011). Previous research has
shown that the ventilation air can leak into the gob area (Krog, Schatzel, & Dougherty, 2014),
and can mix with methane, creating EGZs with different stoichiometries (Gilmore, et al., 2016;
Juganda, Brune, Bogin, Grubb, & Lolon, 2017). Therefore, to understand the impact of
stoichiometry on methane flame propagation for model validation, researchers investigated lean,
stoichiometric, and rich mixtures of methane and air.
To obtain a base case for the 12cm diameter quartz reactor, experiments were performed
igniting different methane-air mixture volume fractions from either the open end of the reactor
(open-end ignition) or the closed end of the reactor (closed-end ignition). In the case of open-end
ignition, the mixture is ignited 11cm from the open end of the reactor, centered in the radial
direction, and the flame travels from the open end to the closed end. For closed-end ignition, the
mixture is ignited 1.39m from the open end of the reactor (11cm from the closed end of the
reactor) and the flame travels from the closed end to the open end. The stoichiometry, or volume
63
|
Colorado School of Mines
|
fraction, of the methane-air mixture was confirmed using a GC-TCD and the volume fraction
reported is within ±0.3%, which is limited by the accuracy of the mass flow controllers. Note
that one of the relief holes (D,=1±0.2cm,A=1.13cm2) is open on the closed end of the reactor,
which helps in flame stability.
Figure 4.3 Impact of mixture stoichiometry on methane flame front propagation velocity versus
distance for open-end ignition. Ignition location is 11cm from the open end. Reported mixture
volume fractions are within ±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One
ign
(1) relief hole. Each data point is the average of 5 data points. Standard deviation range is
between 1-15% of the mean.
Figure 4.4 Impact of mixture stoichiometry on methane flame front propagation velocity versus
time for open-end ignition. Ignition location is 11cm from the open end. Reported mixture
volume fractions are within ±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One
ign
(1) relief hole. Each data point is the average of 5 data points. Standard deviation range is
between 1-15% of the mean.
64
|
Colorado School of Mines
|
Results show for open-end ignition the stoichiometric flame, 9.5% methane by volume,
produced the fastest flame front propagation velocities, followed by the lean flame, 7.5%, and
the rich flame, 11.5%. According to theory, the stoichiometric flame should produce the fastest
burning velocities, followed by the rich and then lean flame (Turns, 2012). However, in this
experimental setup, the rich flame produces a diffusion flame which burns on the open end of the
reactor. This is important because the burning of the diffusion flame acts as a counterbalance to
the main flame front traveling towards the closed end, thus retarding the rich flame. The lean
flame does not produce a diffusion flame and is able to propagate freely down the reactor. Also
to note, for all cases the flame front propagation velocity reaches a peak just over half way down
the reactor. At this point in time, the pressure resistance due to compressed unburned gases on
the closed end of the reactor begin to heat up and increase in pressure, pushing back against the
propagating flame.
Figure 4.5 Images of a stoichiometric methane flame traveling from the open end of the 12cm
diameter quartz reactor towards the closed end. Flame travels from left to right. Ignition location
is 11cm from the open end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa.
4
E =60±5mJ. One (1) relief hole.
ign
Interestingly, the resulting methane flame from an open-end ignition has a unique shape
as shown in Figure 4.5. Initially the flame expands spherically and after the onset of venting of
exhaust gases out the open end of the reactor, the hot, buoyant exhaust gases near the reaction
front rise to the top of the reactor. As the hot product gases rise, they push over the top of the
flame and the resulting traveling flame has a unique angled shape. This has also been observed
by other researchers (Ellis & Wheeler, 1928; Guenoche & Jouy, 1953).
65
|
Colorado School of Mines
|
Figure 4.6 Impact of mixture stoichiometry on methane flame front propagation velocity versus
time for closed-end ignition. Ignition location is 1.39m from the open end. Reported mixture
volume fractions are within ±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One
ign
(1) relief hole. Each data point is the average of 5 data points. Standard deviation range is
between 0-11% of the mean.
Figure 4.7 Impact of mixture stoichiometry on methane flame front propagation velocity versus
time for closed-end ignition. Ignition location is 1.39m from the open end. Reported mixture
volume fractions are within ±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One
ign
(1) relief hole. Each data point is the average of 5 data points. Standard deviation range is
between 0-11% of the mean.
When the ignition location is moved from the open end of the reactor to the closed end of
the reactor, the stoichiometric mixture produced the fastest flame propagation velocities,
followed by the rich and lean flame as shown in Figure 4.6. However, the magnitude of the
velocities produced are significantly different with movement of the spark location. A closed-end
66
|
Colorado School of Mines
|
ignition results in flame propagation velocity magnitudes almost 50 times greater than an open-
end ignition. This is because a closed-end ignition is an ignition from a confined space which
means the hot exhaust gases are not able to flow out of the reactor as easily. The increase
confinement increases the temperature and overpressure of the explosion in addition to
increasing fluid motion ahead of the flame with little to no pressure resistance leading to faster
flame propagation velocities. As can be seen in Figure 4.6, the flame propagation velocities
accelerate fairly linearly, but near the open end of the reactor they begin to fall off. This roll-over
effect was investigated with the 5cm diameter reactor and have observed that as the length of the
reactor gets longer, the roll-over point moves further as well which helps show that the roll-over
is affected by the open end condition (Fig, 2019).
Figure 4.8: Images of a stoichiometric methane flame traveling from the closed end of the 12cm
diameter quartz reactor towards the open end. Flame travels from right to left. Ignition location is
1.39m from the open end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa.
4
E =60±5mJ. One (1) relief hole.
ign
In addition to being quantitatively different then an open-end ignition flame, the closed-
end ignition flame shape is also qualitatively different as shown in Figure 4.8. Because of the fast
expansion of the flame, the buoyant exhaust gases do not rise to the top of the reactor as quickly
as the flame expands axially. The shape of the flame is often referred to as “finger shape” and
has been observed by many other researchers (Clanet & Searby, 1996; Ellis & Wheeler, 1928).
67
|
Colorado School of Mines
|
Figure 4.9 Pressure-time history of a closed-end ignition (CEI) versus open-end ignition (OEI)
with no obstacle. P (OEI, 1 experiment) = 0.26kPa. P (CEI, 5 experiments) = 3.24±0.15kPa.
max max
CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
4 ign
Figure 4.10 Example of a typical pressure-time profile of an explosion in a near cubic vessel
with a single vent. Modified from: (Cooper, Fairweather, & Tite, 1986).
As has been discussed in Chapter 2, Section 2.5, the degree of confinement plays a major
role in the propagation of methane flames, specifically in the amount of pressure rise of the
explosion. Figure 4.9 helps to demonstrate the large overpressure produced by an explosion from
the closed-end of the reactor. As can be seen by both pressure traces, there are two major
pressure peaks followed by decaying oscillations. These researchers have not observed the third
pressure peak described by Cooper, Fairweather, and Tite (1986) and have only observed the
fourth pressure peak in some of the experiments presented in this manuscript.
68
|
Colorado School of Mines
|
Figure 4.11 Pressure-time history of closed-end ignition with varying mixture stoichiometry.
P (CH = 9.5±0.3%) = 3.24±0.15kPa. P (CH = 7.5±0.3%) = 1.56±0.06kPa. P (CH =
max 4 max 4 max 4
11.5±0.3%) = 1.39±0.38kPa. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1)
ign
relief hole.
Figure 4.11 shows the pressure-time history of a closed-end ignition for various methane-
air mixture concentrations. As expected the 9.5% (black) stoichiometric mixture produced the
largest pressure rise, P , which agrees with other researchers (Bao, et al., 2016; Kindracki,
2
Kobiera, Rarata, & Wolanski, 2007). In this case, peak P is approximately twice that of the lean
2
(red) or rich (blue) cases, respectively, which helps to show how dangerous explosions can be if
the flammable mixture is stoichiometric.
4.2 Impact of gob factors on methane flame propagation in the 12cm diameter quartz
reactor
Methane gas explosions in longwall coal mines may occur in different areas of the mine
and can interact with a variety of obstacles including rock rubble, mine structures, and mine
workers. It is well known that many of these explosions occur in confined spaces and can
potentially occur in the gob area directly behind the shields and propagate towards the working
face. Therefore, it is extremely important to understand how various gob parameters such as rock
material, void spacing and location, gob porosity, and rock orientation can affect methane flame
propagation. To study these effects, several simulated gob inserts of varying geometry and
materials have been made to explore the impact of these parameters as shown in Figure 4.12 and
Figure 4.13.
69
|
Colorado School of Mines
|
Figure 4.12 Images of simulated gob materials.
In order to study the effect of gob parameters on methane flame acceleration, this
research uses solid, smooth spheres of varying material and size and compares the results to
granite rock of similar size. Smooth spheres were used for several reasons: 1) spheres take away
any differences in results due to surface topology 2) spheres are easy to model for combustion
model validation and 3) spheres can be easily arranged in a variety of geometries. Glass was
used since it has a low thermal conductivity of 1W/m-K similar to common rock types found in
mines: sandstone (1.7W/m-K) and limestone (1.3W/m-K). The glass spheres will be compared to
granite rock (2.85W/m-K) of averaged similar size. The effects of simulated gob material on
methane flame propagation will be discussed in Section 4.3.2.
Figure 4.13 Schematic of simulated gob-wall geometries.
A non-reacting metal cage 12cm in diameter and 2.54cm in length was used to control the
orientation of the spheres so that researchers may investigate different simulated gob-wall
geometries shown in Figure 4.13. The cage geometry serves as a base case to understand the
effect of the wall and checkerboard geometries. Spheres were arranged in a checkerboard
geometry in order to study the effects of porosity on methane flame propagation. Adding or
removing spheres from the checkerboard geometry is equivalent to changing the porosity of the
70
|
Colorado School of Mines
|
simulated gob geometry without changing the effective void spacing. The wall geometry was
used to study the effects of void spacing on methane flame propagation. By changing the height
of the wall, the effective void spacing changes. The wall geometry can also be oriented in
different ways to study the effect of void spacing at the top of the reactor versus the bottom of
the reactor. Although a void at the bottom of the reactor may not be perfectly analogous to a void
in a longwall coal mine gob, it is still important to investigate for potential future modeling.
Finally, these gob inserts were located at different distances from the open end of the reactor to
test how the relative location of the obstacle to ignition affects methane flame front propagation
velocity.
All experimental results in subsequent sections are performed at 9.5±0.3%% methane by
volume since Section 4.1 showed a stoichiometric mixture produces the most explosive flame of
the stoichiometries investigated.
4.2.1 Impact of void spacing and void location
As shown in Figure 1.3 on page 4, the rock rubble in the gob can be varying sizes,
shapes, and have different void spacings. Therefore, the main goal of these experiments was to
understand the impact of void spacing, or BR, on methane flame front propagation.
To isolate the impact of the simulated gob geometries on methane flame propagation, the
cage was inserted into the reactor 37cm from the open end, between the first two ion sensors. In
Figure 4.14, the mixture was ignited from the open end of the reactor and results show flame
acceleration across the cage. This acceleration is due to the fact that the cage induces some fluid
movement in the nearby unburned gases which accelerates combustion as the flame passes
across the cage. Downstream of the cage, the effect of induced fluid motion is lessened.
In the next set of experiments, researchers investigated how void spacing of a single
simulated gob wall impacts methane flame propagation for both an open-end ignition and closed-
end ignition. Figure 4.15 shows the results of an open-end ignition when the simulated gob wall
is located at 37cm from the open end and the void space decreases from 73% (H=3.8cm) to 13%
(H=9.8cm). Results show that the simulated gob wall with 73% void space (H=3.8cm) had no
significant effect on methane flame front propagation velocity across the obstacle. However,
when the void spacing was decreased to 13% (H=9.8cm) there is an enhancement of methane
flame front propagation velocity across the obstacle. This enhancement is due to flame stretching
around the obstacle, increasing the surface area of the flame, thereby increasing combustion
71
|
Colorado School of Mines
|
rates. For the taller wall, H=9.8cm, the flame surface area increases significantly more to move
around the wall, versus the smaller wall, H=3.8cm. The taller wall, H=9.8cm, also had more
consistent flame propagation run-to-run compared to the wall with H=3.8cm which shows larger
deviations from the mean and a slight slow down downstream of the wall.
Figure 4.14 Impact of cage on methane flame front propagation velocity. Obstacle = Cage.
Obstacle location=37cm. Ignition location is 11cm from the open end. CH = 9.5±0.3%.
4
Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole. Each data point is the
ign
average of 5 data points. Standard deviation range is between 2-8% of the mean.
Figure 4.15 Impact of obstacle height (or amount of void space) on methane flame front
propagation velocity for an open-end ignition. Obstacles = 6.35mm diameter glass spheres in a
wall geometry, L=6.35mm, H = 3.8cm, 9.8cm. Obstacle location=37cm. Ignition location is
11cm from the open end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa.
4
E =60±5mJ. One (1) relief hole. Each data point is the average of 5 data points. Standard
ign
deviation range is between 2-14% of the mean.
72
|
Colorado School of Mines
|
Although the 3.8cm tall simulated gob wall (73% void space) did not have as a
significant impact on an open-end ignition as the 9.8cm tall simulated gob wall, it greatly
accelerated the flame for a closed-end ignition. This is because a closed-end ignition results in a
turbulent flame that when passing over an obstacle wall results in flow separation as shown in
Figure 4.17. This flow separation is important because it forms eddies on the downstream side of
the wall, which increases temperatures and fluid motion promoting flame acceleration, agreeing
with observations made by other researchers (Chapman & Wheeler, 1926) (Moen, Lee,
Hjertager, Fuhre, & Eckhoff, 1982). For an open-end ignition the maximum overpressure rise is
0.26kPa as shown in Figure 4.9 and does not significantly affect the unburned mixture upstream
of the flame. Thus, when the flame passes over the obstacle wall, the flame continues to burn
radially and tangentially as shown by the arrows in Figure 4.17.
Figure 4.16 Impact of amount of void space on methane flame front propagation velocity for a
closed-end ignition. Obstacles = Cage and 6.35mm diameter glass spheres in a wall geometry,
L=6.35mm, H = 3.8cm. Obstacle location=37cm. Ignition location is 1.39cm from the open end.
CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
4 ign
Each data point is the average of 5 data points. Standard deviation range is between 1-4% of the
mean.
In addition to accelerating the methane flame across the simulated gob wall, the 3.8cm
high gob wall (73% void space) also resulted in a, likely, reflected pressure wave whose peak is
greater than the empty reactor and cage obstacle as shown in Figure 4.18. Based on the speeds of
the flame, it is most probable this reflected pressure wave occurred after the flame exited the
reactor. What is also important is that after the reflected pressure wave from the obstacle wall,
greater pressure oscillations were sustained while residual methane-air mixture is burned even
73
|
Colorado School of Mines
|
though the main flame front has exited the reactor. In summary, these results help demonstrate
that obstacles can increase the maximum overpressure, sustain pressure oscillations, increases
secondary burn duration and percentage of burned methane-air mixture, therefore increasing the
total heat release during an explosion. This is important because large overpressures can knock
out ventilations controls, damage mine equipment, and cause harm to workers (Brune, 2014;
McKinney, et al., 2001; Page, et al., 2011). Additionally, the large positive and negative pressure
oscillations induced by the obstacle wall may entrain air, which in a real mine situation, could
sustain a methane fire or reverse ventilation airflow.
Figure 4.17 Image of methane flame propagating across a simulated gob wall for an open-end
ignition (left) and a closed-end ignition (right). Obstacle: 6.35mm diameter glass spheres in a
wall geometry, L=6.35mm, H = 3.8cm. Obstacle location=37cm. CH = 9.5±0.3%. Operating
4
conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
ign
Figure 4.18 Pressure-time history of closed-end ignition with and without an obstacle. Obstacle:
Cage and 6.35mm diameter glass spheres in a wall geometry, L=6.35mm, H=3.8cm. Obstacle
location = 37cm. P (Empty) = 3.24±0.15kPa. P (Cage) = 2.98±0.50kPa. P (Wall,
max max max
H=3.8cm) = 5.24±0.45kPa. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief
ign
hole.
74
|
Colorado School of Mines
|
Finally, after investigating the impact of void spacing on OEI an CEI, researchers were
interested in examining the impact of a longer void space, meaning increasing the obstacle
length. Figure 4.19 shows the results of increasing the simulated gob wall length from 6.35mm to
12.7mm. Results show that by increasing the length of the reduced void spacing further
accelerates the flame. This is because the flame must travel through a longer void, which means
the flame front is being stretched even more, increasing combustion rates and unburned gases to
the flame front. Also to note, the standard deviation range for these experiments was a bit larger
than previous experiments because of older spark ignition circuitry. Since then, improvements
have been made to the circuitry.
Figure 4.19 Impact of obstacle length on methane flame front propagation velocity for an open-
end ignition. Obstacles = 6.35mm diameter glass spheres in a wall geometry, H = 7.62cm,
L=6.35mm or 12.7mm. Obstacle location=37cm. Ignition location is 11cm from the open end.
CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
4 ign
Each data point is the average of 4-8 data points. Standard deviation range is between 2-13% of
the mean.
Thus far all experiments considering the simulated gob wall have been performed with
the void at the top of the reactor. However, in a reality, the void spacing can be distributed in
different locations and although a bottom void is unlikely, a fundamental understanding is still
important. Therefore in this next experiment, the simulated gob wall void was tested at both the
top and bottom of the reactor. Results from Figure 4.20 show that the location of the void
spacing can significantly affect methane flame propagation. In this experimental setup, the flame
front is at the top of the reactor, so when the void location was at the bottom of the reactor, the
75
|
Colorado School of Mines
|
flame had to travel around the obstacle as shown in Figure 4.21. After the flame has propagated
through the bottom void, the flame front propagation velocity is much slower than the top void.
This is because, as can be seen in c) of Figure 4.21, the flame is more concave and has not fully
developed into the elongated angled flame typically observed with a top void. Since the flame
that has passed through the bottom void has less flame area, the downstream propagation
velocities are less than the flame which traveled through the top void. As the bottom void flame
continues to develop, it accelerates down the length of the reactor until reaching the same
velocity as the flame which passed through the top void.
Figure 4.20 Impact of void location on methane flame front propagation velocity for an open-end
ignition. Obstacle: 6.35mm diameter glass spheres in a wall geometry L=12.7mm, H=9.8cm.
Obstacle location=37cm. Ignition location is 11cm from the open end. CH = 9.5±0.3%.
4
Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole. Each data point is the
ign
average of 3-4 data points. Standard deviation range is between 1-18% of the mean.
Figure 4.21 Images of methane flame propagation around an obstacle wall with a bottom void.
Obstacle: 6.35mm diameter glass spheres in a wall geometry L=12.7mm, H=9.8cm. Obstacle
location=37cm. Ignition location is 11cm from the open end. CH = 9.5±0.3%. Operating
4
conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
ign
76
|
Colorado School of Mines
|
4.2.2 Impact of porosity
A longwall coal mine gob consists of varying types of rock rubble which depends on the
geological location and during the mining process, leftover rock rubble collapses and is
compacted by the roof above. The gob itself is a confined space and there is no access to the
majority of the gob and limited visual access to rock rubble near the gob fringes. Due to the
limited access, it is difficult to exactly determine the size, shape, and porosity of the gob. Despite
this challenge, there have been some researchers who have used tracer gas studies to determine
airflow leakage into the gob (Krog, Schatzel, & Dougherty, 2014) and others have used photo-
analysis to get a rock size distribution (Pappas & Mark, 1993a). Because of this challenge and
lack of experimental evidence, researchers have turned to CFD modeling to determine porosities
and permeabilities of the gob (Marts, et al., 2014; Ren & Edwards, 2000; Tanguturi, Balusu, &
Bongani, 2017; Yuan, Smith, & Brune, 2000). In general, it has been found that near the edge of
the gob, especially behind the longwall shields, the rock rubble is less compacted with porosities
near 40%, versus the center of the gob which is much more compacted with porosities near 15%
(Marts, et al., 2014). Since gob porosities near areas of potential explosion risks, i.e. fringe zones
and behind the longwall shields, can be upwards of 40%, this research is interested in the impact
of small changes in porosity on methane flame propagation.
To explore this experimentally, simulated gob-wall checkerboard geometries were made
using 6.35mm diameter glass spheres with 77% porosity and 67% porosity as shown in Figure
4.22. The simulated gobs were located 37cm from the open end of the reactor and results show
that even a small decrease in porosity can significantly retard the methane flame across the gob
(Figure 4.23). These results are important because in a real longwall coal mine there are different
levels of compaction in different areas of the gob where explosions may occur. Understanding
the effects of porosity of methane flames is key to developing a physically accurate combustion
model. Although only a single obstacle was used in these experiments, after model validation,
simulations can be run to investigate the impact of multiple checkerboard obstacles with less
porosity, which would be more representative of the gob.
77
|
Colorado School of Mines
|
Figure 4.22 Schematic of checkerboard geometry with 77% porosity (left) and 67% porosity
(right).
Figure 4.23 Impact of obstacle porosity on methane flame front propagation velocity for an
open-end ignition. Obstacle: 6.35mm diameter glass spheres in a checkerboard geometry
L=6.35mm, Porosity=67% or 77%. Obstacle location=37cm. Ignition location is 11cm from the
open end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief
4 ign
hole. Each data point is the average of 4 data points. Standard deviation range is between 1-7%
of the mean.
4.2.3 Impact of simulated gob location
Experiments presented thus far have only shown results when the simulated gob is
located 37cm from the open end. However, in a real methane gas explosion, the location of
obstacles (and ignition) is unknown. Thus, to build a comprehensive model capable of predicting
these explosions requires knowledge of the impact of obstacle location relative to ignition.
To do this experimentally, the cage and wall obstacles were tested at 37cm, 62cm, and
87cm from the open end of the reactor. The results in Figure 4.24 and Figure 4.25 show that the
location of the simulated gob-wall (or obstacle) relative to ignition location can significantly
impact results. For all locations tested, the cage and wall resulted in acceleration across the
obstacle. However, as the cage and wall were moved further from the ignition location the effect
of acceleration across the obstacle was less pronounced. This is due to the fact that at
78
|
Colorado School of Mines
|
approximately 50cm from the open end the methane flame is almost fully developed.
Additionally, the pressure resistance built up on the obstacle increases and suppresses the impact
of local fluid motion enhanced by the obstacle. Because of these effects, the impact of
acceleration across the obstacle decreases further from ignition.
Figure 4.24: Impact of obstacle location on methane flame front propagation velocity for an
open-end ignition. Obstacle: Cage. Obstacle location=37cm, 62cm, 87cm. Ignition location is
11cm from the open end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa.
4
E =60±5mJ. One (1) relief hole. Each data point is the average of 4 data points. Standard
ign
deviation range is between 1-12% of the mean.
Figure 4.25: Impact of obstacle location on methane flame front propagation velocity for an
open-end ignition. Obstacle: 6.35mm diameter spheres in a wall geometry, H=9.8cm,
L=6.35mm. Obstacle location=37cm, 62cm, 87cm. Ignition location is 11cm from the open end.
CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
4 ign
Each data point is the average of 4-5 data points. Standard deviation range is between 0-8% of
the mean.
79
|
Colorado School of Mines
|
Finally, comparing the results from Figure 4.24 with the empty cage versus the obstacle
wall in Figure 4.25, the average velocities for a given experiment are almost the same. However,
as has been shown, the cage has much more open void space than the wall, which begs the
question as to why the obstacle wall does not produce faster average velocities than the cage.
This difference is due to the fact that these obstacles are very thin, 6.35mm; because the main
flame acceleration mechanism in these cases is flame stretching, the amount of distortion of the
flame traveling across the thin cage is similar to traveling through the void space above the thin
wall.
4.2.4 Impact of simulated gob bed
There are many different types of obstacles in a longwall coal mine and thus far a single
obstacle has been studied. This section aims to gain a better understanding of how a rock pile or
porous pile can affect methane flame propagation and explosion overpressure. To this end, the
obstacle used in these studies is a simulated gob bed consisting of 1cm diameter glass spheres as
shown in Figure 4.26. Similar to the simulated gob wall, researchers varied the height and length
of the simulated gob bed to simulate varying rock piles in a longwall coal mine.
Figure 4.26 Images of glass spheres and an example of simulated gob bed. Both 15cm and 30cm
lengths are tested with 1cm and 2cm heights.
The experiments in Figure 4.27 use a simulated gob bed that is 30cm in length and 1cm
in height (1 layer of 1cm diameter glass spheres) and has a void spacing of 96%. The simulated
gob bed was located 11cm from the open end directly under the ignition location and then it was
moved further down the reactor to investigate how location of the rock pile affects open-end
ignition. Results show that when the simulated gob bed is closer to the point of ignition it
accelerates the flame across the pile more than when it is located further from ignition. This is
because during the initial kernel expansion of the flame there is a small pressure rise, as shown in
80
|
Colorado School of Mines
|
Figure 4.9 on page 68, which creates some movement in the nearby gases, promoting unburned
gases to the flame front. The gob bed acts similarly to a porous media and the small movement in
the gases is accentuated by the spheres which continues to enhance methane flame propagation
as described by other researchers studying porous media (Babkin, Korzhavin, & Bunev, 1991).
When the simulated gob bed is located further from ignition, at 44cm, there is little-to-no flame
enhancement because the pressure resistance of the gob counteracts the fluid motion induced by
the obstacle.
Figure 4.27 Impact of simulated gob bed location on methane flame front propagation velocity
for an open-end ignition. Obstacle: 1cm diameter glass sphere bed, L=30cm, H=1cm. Obstacle
location=11cm, 44cm. Ignition location is 11cm from the open end. CH = 9.5±0.3%. Operating
4
conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole. Each data point is the average of
ign
5 data points. Standard deviation range is between 1-6% of the mean.
In contrast, when the ignition location is moved to the closed end of the reactor, the
simulated gob bed located at 44cm enhances the flame propagation whereas the gob bed at 11cm
had little-to-no impact (Figure 4.28). This acceleration is due to the fact that when the methane
flame interacts with the simulated gob bed it creates a turbulent boundary layer as shown in
Figure 4.29; this turbulent boundary layer was observed for all cases tested, but this figure is
only showing images from an experiment with a L=15cm, H=2cm gob for visualization
purposes. Fluid motion and combustion rates are increased outside of this boundary layer, which
continue to enhance methane flame propagation such that the flame front across the simulated
gob bed moves faster. For the simulated gob bed at 11cm, no flame enhancement was recorded,
however this may mainly be due to the fact that there was only one ion sensor above the bed.
81
|
Colorado School of Mines
|
Figure 4.28 Impact of simulated gob bed location on methane flame front propagation velocity
for a closed-end ignition. Obstacle: 1cm diameter glass sphere bed, L=30cm, H=1cm. Obstacle
location=11cm and 44cm from open end. Ignition location is 1.39cm from the open end. CH =
4
9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole. Each data
ign
point is the average of 5 data points. Standard deviation range is between 1-10% of the mean.
Figure 4.29: Image of the methane flame before (t=42ms) it encounters a simulated gob bed
15cm in length and 2cm in height (Void space = 89%) and image of methane flame as it interacts
(t=0.46ms) with the simulated gob bed during a closed-end ignition. Flame moves from right to
left. Ignition location is 1.39cm from the open end. CH = 9.5±0.3%. Operating conditions
4
294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
ign
Additional experiments investigating how simulated gob bed length affect methane flame
propagation have been performed and similar to the single obstacle, increasing the gob bed
length greatly increases methane flame front propagation velocity and pressure rise as shown in
Figure 4.30 and Table 4.1. Increasing the height of the simulated gob bed decreases the amount
of void spacing from 96% to 89% and even this small change increases the maximum flame front
propagation velocity from 79m/s to 82m/s (Figure 4.31,Table 4.1). Also, increasing the height of
82
|
Colorado School of Mines
|
the simulated gob bed increased the reflected pressure wave and sustained larger oscillations in
the overpressure of the explosion as shown in Figure 4.32. This finding is important because
larger pressure oscillations could possibly reverse airflow in a mine causing continual burning of
an explosive gas mixture and/or damage ventilation controls.
Figure 4.30 Impact of simulated gob bed length on methane flame front propagation velocity for
a closed-end ignition. Obstacle: 1cm diameter glass sphere bed, L=15cm and 30cm, H=1cm.
Obstacle location is 44cm from open end. Ignition location is 1.39cm from the open end. CH =
4
9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole. Each data
ign
point is the average of 5 data points. Standard deviation range is between 1-12% of the mean.
Figure 4.31 Impact of simulated gob bed height on methane flame front propagation velocity for
a closed-end ignition. Obstacle: 1cm diameter glass sphere bed, L=15cm, H=1cm and 2cm.
Obstacle location is44cm from open end. Ignition location is 1.39cm from the open end. CH =
4
9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole. Each data
ign
point is the average of 5 data points. Standard deviation range is between 1-9% of the mean.
83
|
Colorado School of Mines
|
Figure 4.32: Impact of simulated gob bed height on the pressure-time history of a methane-gas
closed-end ignition. Obstacle: 1cm diameter glass sphere bed, L=15cm, H=1cm and 2cm.
Obstacle location is44cm from open end. Ignition location is 1.39cm from the open end.
Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
ign
Table 4.1 Table summarizing the maximum flame front propagation velocity, maximum
overpressure, and minimum overpressure recorded for the closed-end ignition experiments with
and without different obstacles. Averages of 5 experimental runs. CH = 9.5±0.3%. Operating
4
conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole:
ign
Maximum Flame
Maximum Minimum
Obstacle Front
Overpressure Overpressure
Conditions Propagation
(kPa) (kPa)
Velocity (cm/s)
Empty 6500 ± 270 3.24 ± 0.15 -2.45 ± 0.60
Wall L=6.35mm,H=3.8cm 8100 ± 360 5.24 ± 0.45 -2.56 ± 0.46
Bed L=15cm,H=1cm 7900 ± 340 4.56 ± 0.71 -3.13 ± 0.88
Bed L=30cm,H=1cm 8400 ± 720 4.79 ± 0.55 -3.61 ± 1.10
Bed L=15cm,H=2cm 8200 ± 650 6.23 ± 0.87 -5.74 ± 0.45
Bed L=30cm,H=2cm 8600 ± 630 5.52 ± 0.96 -5.50 ± 1.20
Finally, Table 4.1 summarizes the impact of a single simulated gob wall and simulated
gob bed on methane flame propagation. In summary all simulated gob conditions increased
methane flame propagation velocity and increased the overpressure of the explosion. Interesting
to note is that a simulated gob bed with height 2cm, void spacing 89%, produced similar or
greater flame propagation velocities and overpressures than a simulated gob wall of almost
double the height. This is important because it shows that not only do mine structures and other
84
|
Colorado School of Mines
|
solid obstacles enhance mine explosions, but the overall mine environment (i.e. rock rubble in
the gob, rock rubble on the belt, corridors made on rock) can generate sufficient turbulence with
low blockage ratio and have a major impact on the explosion.
4.3 Ignition between simulated gobs in the 12 cm diameter quartz reactor
All experiments performed have been considering methane flame propagation dynamics
with a single obstacle. However, in a real longwall coal mine the EGZs typically are near the gob
area which is composed of varying types of rock rubble. Explosions that originate nearby or from
within the gob can be caused by static discharge of falling rock rubble, hot smears left by metal-
on-rock friction, or a flame created by spon-com (Page, et al., 2011). The resulting flame then
travels from the gob towards the working face. In order to understand how a flame may
propagate in these areas, experiments were performed where the ignition electrodes were placed
between different simulated gob geometries as shown in Figure 4.33. In these experiments,
referred to as in-gob ignition experiments in this manuscript, the ignition location was in Port 1,
25cm from the open end of the reactor. Simulated gob cages, checkerboard geometries, or wall
geometries (no spacing between the spheres) were centered on either side of the ignition
electrodes either D=15cm or D=30cm apart.
Figure 4.33 Image of in-gob ignition in Port 1 with obstacles on either side, a distance D apart.
4.3.1 Impact of simulated gob location
The first set of experiments shown in Figure 4.34 compares the results of igniting a
stoichiometric mixture of methane and air 25cm from the open end to results of in-gob ignition
between empty cages located a distance D=15cm or D=30cm away from each other. Results
show that igniting between empty cages enhances methane flame front propagation velocity
across the obstacles due to enhanced fluid motion induced by the cage. Downstream of the
85
|
Colorado School of Mines
|
simulated gob, the flame front propagation velocities are unaffected since the empty cages have
shown to induce movement only in nearby gases. When the cages are moved further from the
ignition point they have less of an effect on the initial kernel development, inducing less motion
in the overall gases resulting in slightly slower propagation velocities, though still faster than the
open tube without a simulated gob.
Figure 4.34 Impact of simulated gob location on methane flame front propagation velocity for an
in-gob ignition. Obstacle: Cage. Obstacle location represented by bars, D=15cm, 30cm. Ignition
location is 25cm from the open end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa.
4
E =60±5mJ. One (1) relief hole. Each data point is the average of 4-5 data points. Standard
ign
deviation range is between 1-12% of the mean.
Replacing the in-gob simulated gob geometries with 6.35mm diameter glass spheres in a
checkerboard geometry (77% porosity) results in slower flame propagation velocities than the
empty cage cases (Figure 4.35). This is because the glass checkerboard geometry has a larger
pressure resistance which allows less unburned gases to reach the flame front, resulting in slower
downstream propagation velocities. Downstream of the obstacles, the compression of burned
gases builds up, similar to a closed-end ignition, accelerating the flame towards the end of the
reactor. Additionally, the increased pressure resistance of the glass checkerboard geometry also
changes the flame speed trends; as the glass checkerboard geometries are moved further from
ignition they do not inhibit the initial kernel expansion, but continues to induce turbulence in
nearby gases thereby accelerating the flame. However, due to the increased pressure resistance to
the flame, the in-gob experiments using the glass checkerboard geometries results in a higher
peak overpressure than the open tube or empty cages (Figure 4.36). Also to note in this figure,
86
|
Colorado School of Mines
|
these overpressure rises are while the flame is expanding and escaping out of the in-gob
arrangement and the flame continues to travel towards the closed end after 0.15s.
Figure 4.35 Impact of simulated gob location on methane flame front propagation velocity for an
in-gob ignition. Obstacle: 6.35mm diameter glass spheres in a checkerboard geometry (77%
porosity). Obstacle location represented by bars, D=15cm, 30cm between obstacles. Ignition
location is 25cm from the open end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa.
4
E =60±5mJ. One (1) relief hole. Each data point is the average of 5 data points. Standard
ign
deviation range is between 4-20% of the mean.
Figure 4.36 Pressure-time history of in-gob ignition with and without obstacles. Obstacle:
Cages, 6.35mm diameter glass spheres in a checkerboard geometry with 77% porosity. Obstacle
location, D=15cm. Ignition location is 25cm from the open end. P (Empty) = 1.23±0.1kPa.
max
P (Cages) = 1.48±0.1kPa. P (Glass Checkerboard) = 1.97±0.1kPa. CH = 9.5±0.3%.
max max 4
Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
ign
87
|
Colorado School of Mines
|
Figure 4.37 compares the overpressure trace of the checkerboard obstacles spaced 15cm
apart and 30cm apart. Although the peak pressures of these experiments are within the standard
deviation, the initial pressure rise due to the kernel expansion occurs for a longer duration when
the obstacles are spaced 30cm apart, which makes sense since there is more physical room for
the flame to expand. Also, the peak pressure when the obstacles are spaced 15cm apart occurs
earlier than when the obstacles are twice the distance away from ignition.
Figure 4.37 Pressure-time history of in-gob ignition with glass checkerboard obstacles (77%
porosity) spaced 15cm and 30cm apart, centered on Port 1. Obstacle: 6.35mm diameter glass
spheres in a checkerboard geometry with 77% porosity. Obstacle location, D=15cm, 30cm.
Ignition location is 25cm from the open end. P (15cm) = 1.97±0.1kPa. P (30cm) =
max max
2.08±0.1kPa. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1)
4 ign
relief hole.
Finally, the most interesting aspect of all the in-gob ignition experiments is that they
produce a tulip flame as shown in Figure 4.38. This is important because a tulip flame is an
inversion of the flame front due to hydrodynamic instabilities and acoustic interactions with the
flame front (Clanet & Searby, 1996; Ellis & Wheeler, 1928). The tulip inversion stretches the
flame front, which increases combustion rates and pressure resulting in typically faster flame
speeds. Tulip flames have been studied over the past decades and the development of the tulip
inversion has been identified by four major steps (Clanet & Searby, 1996). First, there is the
initial kernel expansion of the flame as shown at t=0.01s in Figure 4.38. Next the flame travels
towards the closed end of the reactor in a finger shape shown at t=0.03s. Third, the edges of the
finger-shape flame reach the cool walls of the reactor and the flame front is flattened as shown at
88
|
Colorado School of Mines
|
t=0.04s. Because of this cooling, the flame front starts to become hydrodynamically unstable and
the tulip inversion is initiated as shown at t=0.05s. Finally, after the onset of the tulip flame,
acoustic-flame interactions dominate which in some cases produce multiple inversions as shown
at t=0.06s. Understanding this phenomena is extremely important because stretching and
inversion of the flame front can result in faster combustion rates and larger peak overpressures
that could lead to more violent explosions. Additionally, matching up the location and shape of
the flame with the overpressure traces is extremely useful for understanding this phenomena and
is one of the major advantages of performing these experiments in the quartz flow reactor.
Figure 4.38 Images of tulip flame resulting from in-gob ignition between two glass checkerboard
geometries (77% porosity). Obstacle location, D=30cm. CH = 9.5±0.3%. Operating conditions
4
294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
ign
4.3.2 Impact of simulated gob material
A simplified experiment of an in-gob ignition was carried out with the ignition electrodes
placed between two “rock walls” with a porosity of 77%, as shown in Figure 4.33. The main goal
of these experiments is to better understand the major differences between solid, smooth spheres
and rough, irregular rock of similar size and thermal conductivity. Initial results have shown that
there is a competing effect of induced turbulence by an obstacle and the pressure restriction from
the obstacle. Results in Figure 4.39 show that ignition between two empty cages enhances
turbulence in nearby unburned gases which helps accelerate the flame within 25cm upstream of
the obstacles. The glass spheres induce movement in nearby gases, but the pressure restriction
from the obstacle slows the flame down as compared to the cage. Furthermore, granite pebbles,
due to their surface roughness, induce more fluid movement in the nearby gases resulting in
higher flame front propagation velocities in the first 25cm as compared to the glass spheres.
89
|
Colorado School of Mines
|
Further downstream of the obstacles, the induced turbulence diminishes and the flame front
propagation velocities approach a similar value due to the pressure resistance experienced by the
flame from the closed end of the reactor.
Simulated gobs have a large impact on the pressure-time history inside the quartz reactor
as shown in Figure 4.40 and Table 4.2. The granite pebbles produced the highest peak pressures
for the longest duration, followed by the glass spheres, and the empty wire mesh. Though the
experimental set-ups vary, these results agree with previous researchers who found obstacles can
increase overpressure (Kindracki, Kobiera, Rarata, & Wolanski, 2007; Moen, Lee, Hjertager,
Fuhre, & Eckhoff, 1982), enhance turbulence (Fairweather, Hargrave, Ibrahim, & Walker, 1999;
Moen, Lee, Hjertager, Fuhre, & Eckhoff, 1982), and promote combustion and burning velocity
(Kindracki, Kobiera, Rarata, & Wolanski, 2007; Fairweather, Hargrave, Ibrahim, & Walker,
1999; Moen, Lee, Hjertager, Fuhre, & Eckhoff, 1982). The rate of decay of the pressure waves
after the second major peak is greatest for the glass spheres, followed by the cage and the granite
rock. These results help to confirm that the granite rock increases mixing for a longer period of
time in the unburned gases allowing for more complete combustion and faster flame front
propagation.
Figure 4.39 Impact of simulated gob material on methane flame front propagation velocity for an
in-gob ignition. Obstacle: Cages, granite rock in a checkerboard geometry (77% porosity), and
6.35mm diameter glass spheres in a checkerboard geometry (77% porosity). Obstacle location
represented by grey bars, D=15cm between obstacles. Ignition location is 25cm from the open
end. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief
4 ign
hole. Each data point is the average of 5 data points. Standard deviation range is between 1-20%
of the mean.
90
|
Colorado School of Mines
|
Also, the recorded flame front propagation velocities of the glass checkerboard obstacles
resulted in much larger standard deviations than the other materials used, a maximum of 20%.
However, when exploring the overpressure traces, shown in Figure 4.41, the maximum
overpressure and resulting pressure oscillations did not significantly change, within 10%. Recall
that in these experiments there is always a balance between induced turbulence by the obstacle
and pressure resistance from the obstacle. For the cages, there is little to no pressure resistance,
so the induced fluid motion is more dominant, resulting in fast flame speeds and less error. For
the granite checkerboard obstacles, there is pressure resistance, but the irregularity of the
obstacle surface induces significant nearby fluid motion, thus dominating the flame acceleration
process. The smooth glass spheres arranged in a checkerboard pattern had a competition between
fluid motion and pressure resistance, resulting in larger standard deviations over 5 experiments.
Figure 4.42 Images of entrained unburned gases and autoignition event resulting from in-gob
ignition between two granite checkerboard geometries (77% porosity). Obstacle location,
D=15cm. CH = 9.5±0.3%. Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief
4 ign
hole.
Another interesting discovery during these experiments was an autoignition event during
an in-gob ignition between the granite checkerboards (Figure 4.42). As shown in these images,
the flame begins propagating into a finger shape (t=25ms), after which the pressure oscillations
entrain unburned gases (t=33.3ms). After t=33.3s, pressure oscillations continue to entrain
unburned gases resulting in an autoignition of the gases between the granite checkerboard
obstacles. Finally, after t=58.3ms, the flame continues to burn down the length of the reactor. To
note, the resulting flame speeds from this experiment did not lie outside the standard deviation of
the mean and thus, were included in the data presented here. Also, the likely autoignition did not
present itself as a sudden spike on the overpressure traces, so unfortunately, there is only
92
|
Colorado School of Mines
|
photographic evidence of this event. However, what is most important about this experiment is
that it shows how the pressure fluctuations from a gas explosion can continue to re-entrain
unburned gases, sustaining combustion and burning leading to an autoignition. This is important
for longwall coal mining because an ignition from within the gob may continue to be sustained if
there is nearby, fresh air being entrained by an initial explosion.
4.4 Impact of ignition location in the 12cm diameter quartz reactor
Since methane gas explosions can occur in a variety of locations with varying degrees of
confinement, it is important to understand how methane flame propagation velocities and
overpressures change depending on ignition location. Previous researchers have ignited mixtures
at the open and closed ends of their combustion chambers (Solberg, Pappas, & Skramstad, 1981)
(Kindracki, Kobiera, Rarata, & Wolanski, 2007), and some in the middle, but none have
thoroughly investigated the impact of varying the ignition along the length of the reactor.
Therefore, to gain a better understanding of the impact of ignition location, in the next set of
experiments the ignition location was varied along the length of the reactor in Ports 1, 2, and 3
(ignition 25, 50, and 75cm from the open end respectively). Figure 4.43 demonstrates the
influence of ignition location. As ignition is moved further from the open end, the maximum
flame front propagation velocity towards both the open and closed end increases. Figure 4.44
shows that the peak overpressure in the explosion vessel also increases as the ignition point
moves away from the open end towards the center of the vessel which agrees with previous
researchers (Bradley & Mitcheson, 1978; Cooper, Fairweather, & Tite, 1986). Ignition within
Port 3 was centered in the quartz reactor and produced the largest pressure rise and sustained
high frequency pressure oscillations which were also observed in high-speed imaging shown in
Figure 4.45.
Researchers also found that ignition in Port 2, 50cm from the open end of the reactor, can
cause a higher pressure rise in the reactor than a closed-end ignition, but has a slower flame front
propagation velocity. The duration of the pressure-time history of ignition in Port 2 is longer than
that for a closed-end ignition. These differences are mainly due to the large acoustical waves
produced by explosion, which in the case of ignition in Port 2, continue to interact with the walls
of the vessel and flame front thereby increasing the overall pressure rise. This result is important
because in a real longwall coal mine, pressure waves may travel throughout the mine, reverberate
off walls and interact with other mine structures and thus increase the overpressure.
93
|
Colorado School of Mines
|
Figure 4.43 Impact of ignition location on methane flame front propagation. Ignition location is
11cm from the open end, 25cm from the open end in Port 1, 50cm from the open end in Port 2,
75cm from the open end in Port 3, and 1.39m from the open end. Dotted lines indicate ignition
location and arrows indicate propagation direction of recorded flame fronts. CH = 9.5±0.3%.
4
Operating conditions 294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole. Each data point is the
ign
average of 5 data points. Standard deviation range is between 1-22% of the mean.
Figure 4.44 Pressure-time history of ignition within various ports. Ignition location is 25cm
from the open end in Port 1, 50cm from the open end in Port 2, 75cm from the open end in Port
3, and 1.39cm from the open end (CEI). P (Port 1)=1.23±0.1kPa, P (Port 2)=5.69kPa,
max max
P (Port 3)=11.99±2kPa. P (CEI) = 3.24±0.15kPa CH = 9.5±0.3%. Operating conditions
max max 4
294±1K, 83±1kPa. E =60±5mJ. One (1) relief hole.
ign
94
|
Colorado School of Mines
|
Figure 4.45 Images of methane flame from ignition in Port 3. Ignition location is 75cm from the
open end. P (Port 3)=11.99±2kPa, P (Port 3)=-12.1±2.6kPa. CH = 9.5±0.3%. Operating
max min 4
conditions 294K, 83kPa. E =60mJ. One (1) relief hole.
ign
4.5 Impact of relief holes in the 12cm diameter quartz reactor
There have been many researchers who have investigated the impact of venting/relief on
flame propagation and overpressure, finding, in general, that increased relief/venting decreases
the overall pressure rise of the explosion (Bao, et al., 2016; Guo, Wang, Liu, & Chen, 2017;
McCann, Thomas, & Edwards, 1985; van Wingerden & Zeeuwen, 1983). As discussed in
Chapter 2 Section 2.5, much of the research into venting is for designing safe pipelines in oil/gas
transport, but this research can also be applied to large-scale industrial explosions including
methane gas explosions in longwall coal mines.
In order to develop a strong understanding on the impact of venting/relief on the 12cm
diameter quartz reactor, experiments were performed changing both ignition location and the
number of relief holes (D=1.0±0.2cm) on the closed end of the reactor. As shown in Figure 4.46
and Figure 4.47, there was no measurable change in flame front propagation velocity or peak
pressure during a closed-end ignition, CH = 9.5%, when the number of relief holes was changed
4
from 0-2 because the expansion of the reaction gases is greater than the volume of unburned
gases vented from the closed end.
95
|
Colorado School of Mines
|
Figure 4.46 Impact of number of relief holes on methane flame front propagation for a closed-
end ignition. Ignition location is 1.39cm from the open end. CH = 9.5±0.3%. Operating
4
conditions 294±1K, 83±1kPa. E =60±5mJ. Each data point is the average of 5 data points.
ign
Standard deviation range is between 0-8% of the mean.
Figure 4.47 Impact of number relief holes on pressure-time history for a closed-end ignition.
Ignition location is 1.39cm from the open end. P (0 holes)=3.29kPa, P (1 hole)
max max
=3.24±0.15kPa, P (2 holes)=3.33kPa .CH = 9.5±0.3%. Operating conditions 294±1K,
max 4
83±1kPa. E =60±5mJ.
ign
The cases shown in Figure 4.48 through Figure 4.52 are for ignition in Port 2 with
varying the number of relief holes on the closed end of the reactor. Similar to previous findings
from other researchers, increasing the venting area can help reduce the overpressure (Bauwens,
Chaffee, & Corofeev, 2008; Bradley & Mitcheson, 1978; Cooper, Fairweather, & Tite, 1986;
Guo, Wang, Liu, & Chen, 2017; Solberg, Pappas, & Skramstad, 1981) and the flame front
propagation velocity towards the closed end of the reactor (Bauwens, Chaffee, & Corofeev,
96
|
Colorado School of Mines
|
2008). This is because increasing the number of relief holes, or venting area, allows for venting
of gases, lowering the pressures and temperatures, reducing combustion rates and the flame front
propagation velocity. Also, due to the combination of these effects, the shape of the flame also
changes as a function of relief holes. As can be seen in Figure 4.50, Figure 4.51, and Figure 4.52,
the flame propagating towards the open end has the same shape as a closed-end ignition flame
(Figure 2.7). However, as the number of relief holes increase, the shape of the flame propagating
towards the closed end looks more similar to an open-end ignition flame (Figure 2.6). For the
case of 2 relief holes, Figure 4.52, the increased venting was enough that the hot, buoyant
exhaust gases had time to rise to the top of the reactor and push over the propagating flame.
Again, this is due to increased venting, allowing for less pressure build up on the closed end of
the reactor (which is also reflected in the overpressure traces in Figure 4.47).
Additionally, it was found for this experimental setup that as venting area decreases,
pressure oscillations are sustained at a greater magnitude and for longer durations which is
important because pressure waves greater than 35kPa can severely harm human ear drums
(Owen-Smith, 1981; Institute of Medicine, 2014). Sustained high pressures can also reverse
airflow in a mine, destroy ventilation controls, and displace mine structures (Zhang & Ma, 2015).
This research demonstrates that even small changes to confinement relief openings directly
impact methane flame dynamics and overpressure.
Figure 4.48 Impact of number of relief holes on methane flame front propagation for ignition in
Port 2. Ignition location is 50cm from the open end. Dotted line indicates ignition location and
arrows indicate propagation direction of recorded flame fronts. CH = 9.5±0.3%. Operating
4
conditions 294±1K, 83±1kPa. E =60±5mJ. Each data point is the average of 5 data points.
ign
Standard deviation range is between 0-37% of the mean.
97
|
Colorado School of Mines
|
discussed in Chapter 3, Section 3.1.2, the main purpose of the experimental box is to begin to
understand how reactor shape and multiple pathways impacts methane flame propagation and
interaction with a simulated gob. The experimental box used has dimensions 51x34x15cm
(LxWxH) and a total volume of 18.4L which is only slightly larger in volume than the 12cm
diameter quartz reactor (16.9L). Experiments in the box were performed with and without a
simulated gob as described in Section 3.1.2, with ignition near the opening (unconfined) and
ignition near the closed end (confined).
Results in Table 4.3 show that for an ignition near the opening (open end), the simulated
gob enhanced the flame front propagation velocities, which was further enhanced when ignition
was located near the closed end in the top-right corner of the box. The methane flame front
propagation result trends were expected since previous research in this manuscript has shown
that 1) confined ignitions result in faster flames and 2) obstacles can enhance mixing and
combustion rates. However, looking further at the shape of the flame and the flame propagation
trends, video results from these deflagrations show that without a gob the flame expands in all
directions as shown in Figure 4.53 and Figure 4.55. Also, the confined, closed-end ignition flame
tends to travel faster towards the relief opening than towards the corners of the box which has
been observed by other researchers experimenting in rectangular enclosures (Solberg, Pappas, &
Skramstad, 1981). However, with a gob, the flame tends to travel through the gob faster than
around the entries as shown in Figure 4.54 and Figure 4.56. This was interesting because one
might hypothesize that the flame would tend to travel in the open spaces faster than through the
gob since these are the path of least resistance. Based on these results it was seen that the
enhanced turbulence by the gob increased the transport of unburned gases to the flame front,
accelerating combustion rates and flame speed.
Table 4.3 Average methane flame front propagation velocities and standard error of the mean
with and without a simulated gob (porous medium) for ignition near the open end (bottom-left
corner) and closed end (top-right corner). Averaged over 2 data sets.
Ignition location
Open end Closed end
Flame Front Propagation Velocity with No Medium 0.9±0.1m/s 7.7±1m/s
Flame Front Propagation Velocity with Porous Medium 1.8±0.2m/s 12±4m/s
100
|
Colorado School of Mines
|
Figure 4.55 Flame images of a closed-end ignition in the experimental box without a porous
medium. Ignition location is in the top-right corner. CH = 9.5±0.3%. Operating conditions
4
294±1K, 83±1kPa. E =60±5mJ.
ign
Figure 4.56 Flame images of a closed-end ignition in the experimental box with a porous
medium. Ignition location is in the top-right corner. CH = 9.5±0.3%. Operating conditions
4
294±1K, 83±1kPa. E =60±5mJ.
ign
4.7 71cm Diameter Reactor Experiments
As discussed in Section 3.1, this research performs experiments in both small-scale and
large-scale reactors in order to investigate scaling of methane flame front propagation velocity
and overpressure. The main purpose of the small-scale, laboratory experiments is to narrow-
down the necessary experiments to perform at the large-scale. Based on experiments performed
in the quartz reactor and experimental box, results show that methane gas explosions are
sensitive to confined ignitions, the amount of void space or blockage ratio, and obstacle surface
topology. Therefore, experiments performed at the large-scale in the 71cm diameter, 6.1m long
reactor are as follows:
102
|
Colorado School of Mines
|
• Closed-end ignition with no obstacles, CH = 9.5%
4
• Closed-end ignition with a rock pile at the open end of the reactor, CH = 9.5%, H =
4
0.24m, L = 1.8m
• Closed-end ignition with a rock pile at the closed end of the reactor, CH = 9.5%, H =
4
0.24m, L = 1.8m
One of the main challenges with performing the proposed experiments in the 71cm diameter
reactor is the weight of the rock pile and moving the rock pile towards the confined (closed end)
of the reactor. In order to have control over containing and moving the rock pile, a winch and
pulley system was set up and connected to half of a cut, metal barrel with diameter of 71cm as
shown in Figure 4.57. This setup allows a single operator to change the height, length, and
location of the rock pile with ease.
Figure 4.57 Image of a rock pile being inserted into the large-scale 71cm diameter, 6.1m long
steel reactor. Rocks are piled on top of a steel bed which is attached to a 2-ton winch and pulley
system allowing the rock pile to be inserted at different locations along the length of the reactor.
Experimental results with and without a rock pile in the 71cm diameter reactor are shown
in Figure 4.58 and Figure 4.59. As can be seen, when the rock pile is close to ignition, the flame
front propagation velocity is enhanced along the entire length of the reactor. When the rock pile
is further from ignition, at the open end of the reactor, the initial flame development is unaffected
by the rock pile. However, when the flame begins to interact with the rock pile, the pressure
wave in front of the flame induces turbulent motion in the gases above and within the void
spaces in the rock pile. This turbulence leads to increased flame speeds across the rock pile as
103
|
Colorado School of Mines
|
shown in Figure 4.58. Examining the pressure histories in Figure 4.59 shows that the rock pile at
the closed end produced the greatest overpressure while the flame was in the reactor. However,
when the rock pile was located at the open end of the reactor, a large, reflected pressure wave
was produced after the flame exited the reactor. This is likely due to a pressure increase in the
rock pile and density difference between products in the reactor and ambient air leading to a
reflected pressure wave. In general, however, the rock pile greatly enhanced flame speed and
overpressure which is important for understanding severity of explosions in an underground
mine. For example, an ignition closer to the gob may result in increased flame speed and
overpressure as compared to an ignition along the longwall face or near a cross-cut.
Figure 4.58 Impact of rock pile location on methane flame front propagation velocity for a
closed-end ignition in the 71cm diameter reactor. Obstacle: Rock pile, H=0.24m, L=1.8m.
Ignition location is at the closed end of the reactor. CH = 9.5±1%. Operating conditions
4
295±1K, 79±1kPa. E =60±5mJ. Each data point is the average of 2-4 data points. Maximum
ign
standard error of the mean is 5.5m/s. Figure credit: (Fig, 2019).
104
|
Colorado School of Mines
|
CHAPTER 5
COMPUTATIONAL FLUID DYNAMICS MODEL SETUP
To accurately solve high-speed methane gas deflagration physics in an underground
longwall coal mine requires computational fluid dynamics (CFD) modeling and complimentary
experiments to validate the model. Thus, the CSM research group has been developing and
validating a mine-scale, CFD model of the ventilation and movement of explosive gas zones in
and around the gob area (Gilmore, et al., 2016; Juganda, Brune, Bogin, Grubb, & Lolon, 2017;
Marts, et al., 2014). Complimentary to the mine-scale ventilation model, the Fig and Strebinger
have been performing methane-air combustion experiments and validating coupled, CFD
combustion models using ANSYS Fluent (Fig, Bogin, Brune, & Grubb, 2016; Fig, Strebinger,
Bogin, & Brune, 2018; Strebinger, Bogin, & Brune, 2019).
Previous research focused on developing 2D combustion models of the laboratory-scale
reactors for the 5cm, 9.5cm, 13.6cm and 30.5cm diameter reactors (Fig, 2019). These laboratory-
scale 5cm and 9.5cm models investigated the impact of humidity, radiation, and presence of a
rock pile on confined methane-gas deflagrations. Fig (2019) validated the models to accurately
capture the relative trends of humidity slowing down flame propagation and concluded that
including radiation into the model resulted in unrealistic flame shapes and propagation trends
(Fig, Bogin, Brune, & Grubb, 2017; Fig, 2019). Fig also found flame acceleration across a
modeled rock pile and trends matched experimental trends at a variety of scales (Fig, Strebinger,
Bogin, & Brune, 2018; Fig, 2019). Additionally, researchers investigated the impact of chemical
reaction models, 2-step, reduced mechanisms, and full mechanisms on methane gas deflagrations
and found that the reduced and full chemistry mechanisms resulted in faster flames and required
significantly more simulation time (Fig, 2019). Important to note in these models, researchers
found that modeling the turbulence using the k-ε turbulence model resulted in more accurately
flame front propagation velocities compared to the k-ω models. Also important to note, the flame
was initiated using a constant heat flux from an aluminum circle meant to represent the spark
electrodes. Although these modeling settings worked well in the 2D, small-scale CFD
combustion models, these assumptions will be revisited for these larger-scaled reactors in this
Chapter.
106
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.