University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Virginia Tech
|
cross-sectional dimensions are the same for all entries, as shown in Figure 26. The air velocity is
defined as 4 m/s at the inlet. The mine layout has been simplified, and obviously does not
represent the exact layout of an operating mine. Because CFD is such a computationally
expensive technique it is likely that when this methodology is used in the field a combination of
simplified mine layouts and less computationally expensive network simulation techniques will
be used. Dimension and air quantity in this model are representative of operating mines.
1188.7 m (3900 feet)
Intake
187.8 m (616 feet)
Regulator 2 Regulator 1
Exhaust
61.9 m (200 feet)
482.8 m (1584 feet) Active
Gob Panel
1214.3 m (3984 feet)
309.7 m (1016 feet) 309.7 m (1016 feet)
27.4 m (90 feet)
Figure 25. The layout of the full scale model mine
2.1 m (7 feet)
4.9 m (16 feet)
Figure 26. The cross section dimension of the model mine
5.4 CFD model setup
5.4.1 Assumptions
Approximations and simplifications of the actual problem are needed to construct the
CFD study, which allows for analyzing the problem with reasonable effort. The following
assumptions are made in this study:
1) No leakage in the mine;
2) Airflow is around the gob -- gob flow is not modeled at this stage;
3) Mine air is incompressible;
4) The flow in the mine is fully turbulent;
5) No heat transfer is considered, and wall and air temperatures are constant;
75
|
Virginia Tech
|
6) The gravity influence on SF is not considered;
6
7) Introduction of SF will not influence the final steady state air flow.
6
These assumptions are made based on our preliminary study focuses, which allows us to
have basic ideas and data with reasonable effort. For example, assumptions 1 and 2 are not
realistic, but this study does not focus on the air leakage and gob air flow, which are extensively
being studied by other researchers [8], [92], [141]; this study is a simple and preliminary
example of how to ascertain system damage remotely at the mine scale. However, some
assumptions could be eliminated as our study progresses by improving our experiment and CFD
model or incorporating the results of other study. Other assumptions may be constrained by the
CFD model. These can be improved through a better understanding and interpretation of the
actual problem and the result data. Take assumption 7 for example, the influence of releasing SF
6
to the air flow is very minor compare to the large space of the mine ventilation system.
5.4.2 Governing equations
CFD is based on the fundamental governing equations of fluid dynamics, including the
continuity equation, momentum equation, energy equation, and transport equation, which
express the fundamental physical principles of fluid dynamics [134]. The governing equations
can be expressed in the conservation form of transport equation [78]:
π
β ββ π Ξ gradπ ( 5-1 )
ππ‘ ππ‘
Where, π is general variable of interest, is air density, Ξ is diffusive coefficient, and is
ππ‘ ππ‘
source term [78].
5.4.3 Mesh and boundary conditions
Commercial drafting and meshing tools were used to generate an unstructured,
hexahedral mesh representation of the geometry of the model mine.
The inlet and the outlet of the model were specified as velocity inlet and pressure outlet,
respectively. The atmosphere pressure, which is 101.325 kilopascals, is applied to the pressure
outlet boundary. A 4 m/s velocity was assigned to the mine inlet to achieve realistic air quantities
to each panel. All of the other surfaces and regulators are treated as stationary walls with no slip.
76
|
Virginia Tech
|
5.4.4 Numerical details
The numerical simulations in this study were conducted using the commercial CFD
package, ANSYS FLUENT 12.1, to simulate the airflow and tracer gas dispersion. A standard
two equations k- turbulence model was employed to simulate the air flow and SF transport.
6
The standard k- model is the simplest complete turbulence model and widely used in the
modeling of mining turbulent flow in broad range of applications [8], [9], [22]. A second order
upwind scheme was used for variables including pressure, momentum, turbulent kinetic energy
and turbulent dissipation rate, which ensures the higher order of accuracy results. Discretized
airflow equations were solved with the SIMPLE algorithm in the CFD program to couple the
pressure, velocity, momentum and continuity equations.
5.5 Mesh independent study
5.5.1 Mesh quality and size
In order to achieve results that are independent of mesh size, three different mesh size
models were developed and analyzed, including coarse mesh, medium mesh, and fine mesh. The
mesh quality and node numbers are shown in
Table 8. The mesh quality is considered very high since the determinants are 1 and angles
are 90β°. The node numbers are approximately doubled progressively from coarse mesh to fine
mesh.
The meshes were generated to improve the node density both on the entriesβ cross
sections and along the roadways. The mesh density is high near the roof, ribs and floor in order
to resolve the rapid variation of flow variables near these regions. Mesh size is gradually
increased toward the center of the roadway where the flow variation gradient is relatively small.
The detail nodes number and size for each mesh are shown in Table 9. Figure 27 shows the cross
section mesh for coarse, medium, and fine mesh.
Table 8. Mesh quality and nodes number
Determinant 3Γ3Γ3 Angle Nodes Number
Coarse Mesh 1 90β° 6,692,976
Medium Mesh 1 90β° 13,160,889
Fine Mesh 1 90β° 22,944,064
Table 9. Cross section mesh parameters
With Height The first cell space near the wall (feet) Increase ratio
Coarse Mesh 38 16 0.12 1.05
77
|
Virginia Tech
|
Medium Mesh 50 21 0.10 1.05
Fine Mesh 65 26 0.08 1.05
.
Figure 27. Different mesh of cross section (from left: coarse mesh, medium mesh, and fine mesh)
5.5.2 Solution convergence
Some criteria need to be met to achieve converged numerical solutions. Two criteria are
used to check the solution convergence for each mesh [139]. The first criteria is residuals of each
conservation equation, which is a specified tolerance defined by Fluent. The solutions are
considered converged when the set residual tolerance has been reached. The criterion used in this
study is the continuity residual reaches or less than 10-6.
Sometimes the residual may reach the convergence criterion, but the solution still
changes with more iterations. Therefore, a second criterion is used, which monitors the solution
until it no longer changes with more iterations. In this study a point monitor was created and the
velocity at this point was monitored. The point monitor was set 304.8m away from the outlet, in
the middle of the roadway center line, and 0.15 m below the roof as shown in Figure 28. More
iterations are applied until the velocity is stable. Figure 29 shows the velocity change with
iterations for each mesh size model. The plot shows that the solution for each model at the
monitor point stabilized and reached a steady state with further iterations, so the results are
considered converged. One can notice that differences exist between different mesh size models.
The final velocity values for coarse, medium, and fine mesh size models at the monitor point are
3.8396 m/s, 3.9347 m/s, and 3.9871 m/s, respectively. The percentage change value can be used
to compare the results changes with mesh difference. The percentage change equation is given
by the following equation.
π΅
π | | % ( 5-2 )
Where A is fine mesh result and B is medium or coarse mesh result. Using equation 2 to compare
the velocity differences with mesh difference, we get:
π | | %
| | % %
78
|
Virginia Tech
|
π | | %
| | % %
Where V is the velocity result of each mesh. These calculations indicate that the velocity
differences at the monitor point between medium and fine mesh is less than 2%, which is
acceptable, and the conclusion can be made that mesh independence has been achieved, which
will be discussed in detail in the next section.
2.1 m (7 feet)
Velocity Monitor Point
4.9 m (16 feet)
Figure 28. Point monitor location
7
6.5
6 CoarseMesh
FineMesh
MediumMesh
ms)
/5.5
(
y
Vo ec
lit
5
4.5
4
3.5
0 200 400 600 800 1000
Iterations
Figure 29. Velocity convergence history for different mesh size
5.5.3 Mesh independence study
It is important to conduct the mesh independence study before using the CFD results
since the numerical solution may depend on the mesh size if mesh independence is not achieved
[48]. As the mesh becomes finer, the numerical solution will asymptotically approach the exact
solution of the governing equations [43]. Mesh independences are studied considering different
flow features and at different locations. The following section shows the details of the
comparison.
With the purpose of comparing the final result profile of different mesh size models, a
horizontal center line 45.7 m and 304.8 m away from the outlet was created for each case and
three velocity profiles across the center line were plotted on Figure 30 and Figure 31. From
79
|
Virginia Tech
|
Figure 30 we can see that as the mesh becomes finer, the results are changing but asymptotically
approaching a profile which is the actual results. Figure 31 also indicate that the results of
different mesh size are very close.
In this study, the solution is considered mesh independent since the result of medium
mesh and fine mesh are very close as can be seen in Figure 30 and Figure 31. The result
differences are within 2% as indicated by the calculation of previous section. This also indicates
that the medium mesh is sufficient for a robust solution. However, fine mesh will provide more
precise solution with longer computation time. In this study fine mesh is used since its
computation time is still acceptable.
5
4.5
4
3.5
CoarseMesh
MediumMesh
ms)
/
3 FineMesh
(
y2.5
cit
o
el
V 2
1.5
1
0.5
0
-186 -185 -184 -183 -182 -181 -180
Position(m)
Figure 30. Velocity profile at the line monitor 45.7 m away from outlet
5
4.5
4
3.5 CoarseMesh
MediumMesh
FineMesh
ms)
/
3
(
y2.5
cit
o
el
V 2
1.5
1
0.5
0
-186 -185 -184 -183 -182 -181 -180
Position(m)
Figure 31. Velocity profile at the line monitor 304.8 m away from outlet
80
|
Virginia Tech
|
quantities of SF . However, in turbulent flow, the diffusion, which is the third term of Equation
6
5-3, was programmed as a user defined function as follows [134]:
Ξ
ππ‘
Ξ
ππ‘
( 5-3 )
π‘
Where Ξ is diffusion coefficient of SF in air, is the turbulent viscosity, and is the
ππ‘ 6 π‘
turbulent Schmidt number. Tucker et al. [142] provided an equation to calculate Ξ of one gas in
ππ‘
another gas resulting in a diffusion coefficient of SF in air 8.96 Γ 10-6 m2/s; Bai et al. [136] used
6
a value of 9.7 Γ 10-6 m2/s in their study; while Ward et al. reported the diffusion coefficient of
SF in air is between 5.9 Γ 10-6 m2/s to 7.3 Γ 10-6 m2/s. These values do not differ substantially,
6
especially when considering , the second term in Equation 3, which is three orders of
magnitude larger than Ξ ; the final results is not significantly sensitive to the range of Ξ cited
ππ‘ ππ‘
in the literature. Therefore, a value of 5.9 Γ 10-6 m2/s is used in this study. The Schmidt number
is a dimensionless parameter which is the ratio of diffusion of momentum to the diffusion of
mass. For gases, it is approximately 0.7 [50].
5.6.3 Results
SF concentration was monitored at the outlet and plotted over time, as shown in Figure
6
34. Four possible tracer gas travel paths are plotted in Figure 35, which are numbered from 1 to
4. As can be seen from Figure 34, different ventilation scenarios have totally different SF
6
profiles at the outlet. Four peaks are observed for the normal ventilation status, which represents
the tracer gas going through all four paths showed in Figure 35 (from inlet directly to outlet,
from inlet to active panel then to outlet, from inlet to gob panel then to outlet, and from inlet to
active panel then to gob panel to outlet). Two peaks show up for active panel roof fall and gob
explosion scenarios, but the peak height and arrival times are different. In both scenarios, the
first peak represents flow path #1, in which SF flows from inlet directly to the outlet. In the
6
active panel roof fall case, the air path also goes from the inlet to gob panel then to outlet (flow
path #2); and in the gob explosion case, the second air path is from the inlet to active panel then
to outlet (flow path #3). The flow path #3, which going to gob panel is longer than flow path #2
that to active panel, therefore the second peak of active panel roof fall case shows up later than
that of gob explosion case.
The peaks can also be grouped according to the arrival time, which is directly related to
the flow paths. The arrival time of each flow path in the three different ventilation status are
82
|
Virginia Tech
|
approximately in the same time range due to the fact that the total quantity of mine air is not
changed. Figure 34 shows the simulated peaks for each scenario relative to each other, with the
scenario flow paths in Figure 35.
450 Flow path #1
Normal ventilation case
400 Flow path #2
)
b p350
Ventilation after gob explosion
p
( n o300 Flow path #3
it250 Ventilation after active panel roof fall
a
r
t
n e200 Flow path #4
c
n150
o
C
6100
F
S
50
0
0 10 20 30 40 50 60 70 80 90 100
Flow Time (min)
Figure 34. Simulated SF concentration at a point monitor on the outlet VS flow time
6
Intake
Regulator 2 Regulator 1
Exhaust
Active
Flow path #1 Gob Panel
Flow path #2
Flow path #3
Flow path #4
Roof fall
damage
Explosion
damage
Figure 35. Tracer gas flow paths
5.7 Conclusions and discussions
This study conducted a full scale simplified model mine CFD simulation. Solution
convergence and mesh independence studies were performed in order to obtain results
independent of the model mesh size. The mesh size used in this study can be utilized as a guide
when constructing field scale CFD models in underground mines. Additionally, this paper details
the methodology for conducting convergence and mesh independence studies which are
necessary for generating robust solutions with CFD.
83
|
Virginia Tech
|
Pulse injection of tracer gas (SF ) was simulated for three different ventilation scenarios,
6
which are the normal operating scenario, ventilation after roof fall in the active panel, and
ventilation after gob explosion. SF concentration was monitored at the outlet and plotted over
6
time. The SF concentration profile is obviously different for the three different ventilation
6
statuses, which indicated that this method can be used to analyze and predict the ventilation
status underground.
The simulated results are also valuable for the design of on-site experiments. For
example, the results can be used to determine how much tracer gas needs to be released in order
to achieve a concentration at the outlet that is practically detectable; how long the tracer gas
should be released in order to generate the peaks for each airflow path, which is a key factor in
identifying separate peaks for various ventilation scenarios; and the optimal time interval for
sampling in order to adequately resolve each peak (in this study the smallest peak width is 10
minutes, which requires 1 minute sample interval to capture 10 points on the peak).
Further field experiments studies are needed to validate the CFD mode. Nevertheless, this
study illustrates the potential of CFD to model tracer gas and its use in determining underground
ventilation status. The CFD model and the predicted results in this study provided valuable
guidance for further real mine CFD model and tracer gas field experiments design. Tracer gas
experiments may be time and resource consuming. Therefore, carefully field tracer experiment
design is very important in terms of efficiency and effectiveness. The CFD methods used in this
study allows the researchers to determine release rates and volumes, expected profile shape and
width allowing for design of best sample collection, and to anticipate profiles under various
scenarios; all essential information for the experimental design.
84
|
Virginia Tech
|
The following paper will be submitted to a peer reviewed journal. The experimental and CFD modeling
work and writing was primarily completed by Guang Xu with editorial and technical input from Dr. Kray
Luxbacher, Edmund Jong, Dr. Saad Ragab, and Dr. Michael E. Karmis.
6 Remote Characterization of Ventilation Systems using Tracer
Gas and CFD in an Underground Mine
6.1 Abstract
Following an unexpected event in an underground mine, it is important to know the state
of the mine immediately, even with limited information, to manage the situation effectively.
Especially when part or the whole mine is inaccessible, remotely and quickly ascertaining the
ventilation status is essential to mine personnel and rescue teams for making effective decisions.
This study developed a methodology that combines tracer gas and CFD modeling to remotely
analyze underground mine ventilation systems. The study was conducted in an underground
mine with four different ventilation scenarios created intentionally for this study. CFD models
were built not only to simulate various ventilation scenarios, but also to optimize tracer test
parameters to minimize the trial and error process. This minimization guarantees that the status
of a ventilation system can be identified more rapidly in an emergency situation. The
methodology was successful in identifying the experimental ventilation. This study showed that
this methodology was effective in the field. Limitations of this study are discussed at the end of
this paper.
6.2 Introduction
The remote collection of data is necessary when circumstances prevent people from
entering an underground mine. Such situations include roof falls, outbursts, water inundations, or
explosions. Communications between underground miners and rescuers on the surface may be
tenuous at best because very few commercially available communications systems are capable of
meeting basic requirements for emergency communications [117]. However, information
regarding the status of the mine must be gathered immediately to estimate the extent of damage
before determining rescue and recovery methods. Even with considerable improvement in
underground communications systems, it is still necessary to remotely ascertain the mineβs status
using other methods. Some alternate methods can be used to gather information safely, such as
collection of air samples from boreholes, insertion of video cameras into boreholes to visualize
85
|
Virginia Tech
|
underground status, and utilization of rescue robots if possible. However, none of these methods
are reliable enough to stand alone.
Remotely and quickly ascertain the ventilation status is essential to mine personnel and
rescue teams for making effective decisions. Especially in some incidents, such as explosions,
communication lines may be damaged, roof may be collapsed, and stoppings may be destroyed.
The post-incident status of these components are largely unknown. The airflow quantity, airflow
paths, and ventilation patterns will change according to the nature of the damage. Therefore, the
level of damage can be approximated by remote measurement of these ventilation parameters.
Due to the complexity of the ventilation system, employment of the tracer gas method is an
effective means of characterization ventilation systems where conventional techniques are
inadequate or cannot be effectively employed [1], [2]. Computational Fluid Dynamics (CFD) can
be used to model normal ventilation patterns as well as possible post-incident scenarios. By
comparing the actual tracer test results and the modeled results under different ventilation
conditions, the state of the ventilation system can be determined.
Tracer gas was first used in the building ventilation systems in the 1950s [61] and has
been widely used for ventilation analysis both in buildings and in underground mines [118].
Numerous studies have utilized tracer gas techniques as a means to evaluate ventilation systems
in underground metal/non-metal and coal mines. Sulfur hexafluoride (SF ) is widely accepted as
6
a standard mine ventilation tracer [118] because it can be detected in low concentrations, and it
,
is nontoxic, odorless, colorless, chemically and thermally stable, and does not exist naturally in
the environment [1]. Therefore, it was selected for use in this study. The applications for tracer
gases in underground mines include measurement of turbulent diffusion [107], methane control
[121], study of mine ventilation recirculation of return into intake air, transit flow times through
stoped areas, effectiveness of auxiliary fans, and estimation of volumetric flow rates [1], [122],
air leakage investigation, and evaluation of dust control measures [62]. However, most of these
studies did not use CFD to design the tracer test in advance and usually were based on
experience as well as trial and error.
CFD has become a powerful tool and has been commonly used to model underground
mine air flows [67], [100]. It has been used in a number of areas including ventilation airflow
patterns modeling [67], [98], study and control of coal spontaneous heating and underground fire
[9], [20], [83], optimization of gob inertization [95], dust control [98], and methane management
86
|
Virginia Tech
|
[91]. The combination of experimental measurement and CFD modeling of tracer gas has been
used to study airflow and contaminant transport in indoor environments and other industrial
applications [114], [123], but little research has been done to model underground tracer gas
applications, especially to optimize tracer test parameters and to remotely characterize
ventilation systems after an unexpected event.
Preliminary estimation of the damage
level and possible ventilation scenarios
using available information
CFD modeling of normal ventilation
Tracer on-site experiments
status and possible damage scenarios
Compare on-sit experimental
results and CFD simulated results
Identify the level of
ventilation damage
Figure 36. Flow chart of the methodology
The objective of the study presented in this paper is to develop a new methodology that
can identify the general level of ventilation damage using a tracer gas and CFD modeling. An
overview of the methodology can be seen in the flow chart shown in Figure 36. After an
unanticipated event that has changed the ventilation controls, the level of the damage and the
possible ventilation changes need to be estimated based on the available information. The CFD
model can then be built to model the normal ventilation status before the event and possible
ventilation damage scenarios. At the same time, tracer gas tests can be designed and performed
on-site. Tracer gas can then be released at a designated location with constant or transient release
techniques. Gas samples are collected at other locations and analyzed using a gas chromatograph
(GC). Finally, through comparing the CFD simulated results and the tracer on-site test results,
the general level of ventilation damage can be determined.
In this study, the tracer SF was released in an area of a limestone mine with varying
6
ventilation statuses intentionally created for this study. The scenarios included a stopping door
open and closed and a booster fan turned on and off in various combinations. For this study,
personnel were used inside the mine to conduct the experiments. However, in an actual event,
87
|
Virginia Tech
|
tracer gas release and sampling would be made through boreholes or other remote access. CFD
models were built based on detailed mine entry measurements and ventilation surveys under the
different scenarios. The CFD models were not only used for simulating tracer dispersion under
the different scenarios, but also for tracer test design to optimize essential parameters, such as
tracer release location, release rate and duration, and sampling location. The trial and error
procedure is thus reduced or even avoided to achieve the desired results. According to the
optimized parameters, SF was released in the mine using the pulse release method. Gas samples
6
were taken continuously using blood collection Vacutainers. The gas samples were analyzed by
GC and the tracer concentration profiles were plotted over time for the different ventilation
scenarios. The CFD model results agreed with the on-site tracer test results with reasonable
errors. The experimental results showed that this methodology can help determine the ventilation
system status. The developed methodology proved feasible in the laboratory in an earlier study
[143]. This methodology provides an alternate way to gather information that can be used by
mine personnel and rescuers to take safe and effective actions.
6.3 Onsite Experiments Description
6.3.1 Location of the onsite experiments
The experiments were conducted in the mine section shown in Figure 37. It is a 250 m
belt entry, which is connected to a 40 m crosscut. The average cross sectional dimension of the
belt entry is 5 Γ 2.8 m and 5.87 Γ 3.82 m for the cross cut entry. However, the cross sectional
dimensions of both entries were measured every ten meters, and the results were used to
construct the CFD model so that the actual dimension of those entries could be more accurately
represented in the model. Three steel water pipes and one conveyer belt were present within the
entries. They are considered large enough to influence the air flow, so their dimensions and
positions were also measured and included in the CFD model as solid impermeable regions.
There are two velocity inlets, one is at the lower end of the belt entry, and the other one is the
stopping door at the end of the crosscut. Airflow directions are shown as red arrows in Figure 37.
88
|
Virginia Tech
|
Figure 37. Layout of the entry section (red arrows show the air flow)
There is a booster fan at the bottom of the belt entry (not within the measured and
modeled domain shown in Figure 37) that can be turned on or off causing total air flow change in
the belt entry. Therefore, the different ventilation scenarios could be created intentionally by
opening and closing the stopping door, and turning the booster fan on and off. A combination of
four different ventilation scenarios from manipulating these parameters is presented in Table 10.
Each scenario was assigned a case number and will be referred later in the paper for
convenience. The velocity boundary conditions at the inlets are also shown in Table 10. The goal
of this study was to design appropriate tracer gas tests and build CFD models to identify these
four ventilation scenarios.
Table 10. Different ventilation scenarios
Door velocity inlet (m/s) Belt entry velocity inlet (m/s)
Case #1: Stopping door open, booster fan on 4.21 2.54
Case #2: Stopping door close, booster fan on 0.06 2.54
Case #3: Stopping door open, booster fan off 4.21 2.10
Case #4: Stopping door close, booster fan off 0.06 2.10
6.4 Tracer test design using CFD
Tracer gas experiments are time and resource consuming. Therefore, it is important to
carefully design the experiments beforehand. Some essential parameters, such as the tracer
release location, release rate and duration, sampling location, and the expected results, can be
optimized using a CFD model before conducting the actual experiment. This makes it possible to
reduce or even avoid the trial and error procedure to achieve expected results. CFD models were
built in 2D to save computational time and in 3D to provide more accurate results.
89
|
Virginia Tech
|
6.4.1 CFD model setup
Approximations and simplifications of the actual problem are needed to construct the
CFD study, which allows for an analysis with reasonable effort. The following assumptions are
made in this study:
1) Mine air is incompressible;
2) The flow in the mine is steady and fully turbulent;
3) No heat transfer occurs during the study and the wall and air temperatures are constant;
4) The gravity is 9.81 m/s2;
5) Introduction of SF will not influence the existing steady state air flow.
6
The inlet and the outlet of the model were specified as velocity inlet and pressure outlet,
respectively. The averaged velocity values, which are shown in Table 10, were applied to both of
the velocity inlets in the belt entry and at the stopping door. These boundary conditions are based
on the average measured air velocity readings by a hot wire anemometer using the fixed point
traverse method. All of the other surfaces are treated as stationary walls with no slip. Both air
and wall temperatures are assumed constant. A realizable two equation k-Ξ΅ turbulence model was
employed to simulate the air flow. A second order upwind scheme was used for variables
including pressure, momentum, turbulent kinetic energy, turbulent dissipation rate and SF
6
transport, which ensures the higher order of accuracy results. Discretized airflow equations were
solved with the SIMPLEC algorithm in the CFD program to couple the pressure, velocity,
momentum, and continuity equations. A gravity of 9.81 m/s2 was used to establish gravitational
influence on the flow and SF distribution. Steady state flow was calculated first for all cases and
6
then SF was released for a certain period at a designated location using the two species (air and
6
SF ) transport model. Two solution convergence criteria are used: the continuity equation
6
residual reduced to 10-5 and the velocity at a pre-selected point achieved steady-state. A mesh
independency verification was conducted for the CFD models to ensure that the results were
independent of the mesh size. This verification is only presented for the 3D model in Section
6.4.3.
6.4.2 2D CFD model
A CFD model can be built in either 3D or 2D. A 3D model can provide more accurate
results, but it is much more computationally intensive compared to the 2D model, especially
when parameters need to be adjusted frequently during the optimization process. Therefore, a 2D
90
|
Virginia Tech
|
model was used first to provide results quickly, and after the optimized parameters were
determined, a 3D model was used to obtain more accurate results. The 2D model schematic is
shown in Figure 38.
The tracer gas release point was located in the crosscut entry in order to capture the flow
changes in both the belt entry caused by the booster fan and the crosscut entry caused by the
stopping door. Two locations were examined: one is 38 m from the door and is denoted as
Release Point 1, the other is 2.5 m from the door and is denoted as Release Point 2. Both points
are in the center of the entry. The sample points were chosen 10 m from the model outlet and 7
point monitors that are evenly distributed across the entry were created in the model to monitor
SF concentration change.
6
Outlet
Pressure 1 SSS0
aaa
m
mmm
pppfr lllo eeem
PPP
oooo iiiu nnnt tttl e 321t
Sample Point 4
Sample Point 5
Sample Point 6
Sample Point 7
Release point 1
38 m from the door
Release point 2
2.5 m from the door
Velocity
inlet
Door
Velocity inlet
Figure 38. Tracer gas release and sample points
To design a tracer experiment, the first parameter that needs to be defined is where to
release tracer. The two proposed release points in Figure 38 were examined under different
cases. Point 1 was chosen as the release point because tracer concentration profiles can be
captured within 10 min after release, while more than 20 min sampling time is needed to capture
the tracer profile when the door is closed if released at Point 2.
The second parameter is the amount of tracer to release so that gas samples can be
directly analyzed using GC at a concentration level within the detection limit. The amount of
tracer released to the mine is controlled by the combination of the release rate and the release
duration. The release rate is controlled by a flowmeter. After examining several different release
91
|
Virginia Tech
|
rates in the CFD model, it was determined 8.89 L/min was a reasonable rate to produce a
concentration profile within the GC detection range.
The release duration not only determines the maximum concentration of the tracer profile
but also the width of the tracer profile with respect to time. The profile time-width needs to be
wide enough to allow enough samples to be taken. As can be seen in Figure 39, under Case 1
(door open, fan on) and release at the rate of 8.89 L/min, different release durations result in
different profiles. If the release is too short, such as 10 s, the profile peak lasts only 30 s, which
makes it hard to adequately sample to capture the peak. The concentration profile becomes
broader as the release duration increases from 10 s to 2 min. Figure 39 also shows that the 30 s
release increased the maximum concentration level compared to the 10 s release. However,
increasing the release duration further, to 1 min and 2 min, did not increase the maximum
concentration level. That is because the release duration is long enough that the maximum
concentration leveled out under the fixed release rate. By comparing these profiles, releasing the
tracer for 1 min was determined to be adequate for our sampling and analyses convenience. The
expected concentration profile width is more than 80 s. This was wide enough to be captured by
a 5 s sampling interval that we use in the field.
14000
12000
) b Release 10 s
p
p 10000 Release 30 s
(
n Release 1 min
o
it a 8000 Release 2 min
r
t
n
e 6000
c
n
o
C 4000
6
F
S 2000
0
50 70 90 110 130 150 170 190 210 230 250
Time (s)
Figure 39. SF6 concentration profile under different release durations
The third parameter was where to take samples. Seven sample points, denoted as Sample
Point 1 to 7 in Figure 38, evenly distributed across the entry were monitored. The final sample
point was determined to be at Sample Point 7. The distribution of SF in the belt entry was
6
higher toward one side of the entry where Sample Point 7 was located and gradually lower
toward where Point 1 was located, which can be seen in Figure 41. The tracer gas concentration
profiles for these points were plotted in Figure 40. The maximum concentrations at the
92
|
Virginia Tech
|
monitored points were generally less than 12 ppm. At these concentration levels, a higher
concentration can be analyzed relatively accurately. Therefore, Sample Point 7 was chosen to
take air samples.
12000 Sample Point 1
)
Sample Point 2
b
p 10000 Sample Point 3
p
(
n
Sample Point 4
o 8000 Sample Point 5
it
a r Sample Point 6
t n 6000 Sample Point 7
e
c
n
o 4000
C
6
F
S 2000
0
0 20 40 60 80 100 120 140 160 180 200
Time (s)
Figure 40. SF6 concentration profile on different monitor points
Figure 41. SF6 concentration contour after 170s release (Case 1)
It is always better to know the expected results beforehand. Based on the studies shown
above, the optimized tracer test parameters are as follows: the tracer release location is located at
Point 2 in Figure 38, the release rate is 8.89 L/min for one minute, and the sampling location is at
Sample Point 7 in Figure 38. The SF concentration profiles under the four ventilation cases,
6
mentioned in Table 10, are shown in Figure 42. As can be seen, the profiles under the different
ventilation cases are separated with each other based either on the differences in the peak arrival
time or the peak height. The tracer profile shape is very different when the door is open and
closed. This is because the flow features in the crosscut entry, especially at Point 1, are different,
which is shown in Figure 43 as flow path lines. When the door is open, the tracer will go directly
to the belt entry and to the outlet because the flow direction around the release point directly
93
|
Virginia Tech
|
goes to the belt entry. However, when the door is closed, there are circular flow directions
around the release point causing the tracer to recirculate in the crosscut and flow to the belt entry
very slowly.
16000
14000
)
b
p 12000
p
(
n
Case 1: door open, fan on
o 10000 Case 2: door close, fan on
it
a r 8000 Case 3: door open, fan off
t
n Case 4: door close, fan off
e
c 6000
n
o
C
6
4000
F
S
2000
0
0 60 120 180 240 300 360 420 480 540 600 660 720
Flow Time after Tracer Released (s)
Figure 42. SF profile comparison under different ventilation status
6
Figure 43. Flow feature difference in the cross cut entry when the stopping door is open and close
6.4.3 3D CFD model
2D flow is different from 3D flow. A 3D model provides more accurate results.
Therefore, after the parameters were determined using a 2D model, a 3D model was used to
validate and provide more accurate results.
The geometry of the 3D model is complex, mainly because of the existence of the water
pipes in the entry. Meshes were generated using Ansys ICEM software and the cross section
meshes are shown in Figure 44. A tetra-dominated mesh with prism layers was generated at the
beginning of the modeling (Figure 44 (a)). However, the mesh quality is limited and difficult to
94
|
Virginia Tech
|
improve, which increase the difficulty for solution convergence. A hexa mesh (Figure 44 (b))
was generated later to improve the mesh quality, convergence behavior, and result accuracy. Due
to the complexity of the geometry, it was divided into four parts. The meshes were generated
separately in ICEM and combined together later in Fluent.
Figure 44. Mine entry mesh types
4.0
3.0
) m Medium Mesh
(
n Fine Mesh
o2.0
it
is
o
P
1.0
0.0
0.0 1.0 2.0 3.0 4.0
Velocity (m/s)
Figure 45. Medium and fine mesh velocity comparison
A mesh independence study was conducted by generating a medium and a fine mesh and
comparing the results. This step is essential in CFD modeling because the numerical solution,
such as velocity and tracer concentration in this study, may be affected by the mesh size if mesh
independence is not achieved [48]. As the mesh becomes finer, the numerical solution will
asymptotically approach the exact solution of the governing equations [43]. The total number of
nodes for the medium mesh is about 20 million, and 40 million for the fine mesh. Since the focus
95
|
Virginia Tech
|
of this study was not to find the coarsest mesh that can achieve acceptable accurate results, the
number of nodes chosen in this study was relatively large to guarantee a robust solution. The
velocity profiles on the vertical center line of the cross section 10 m from the outlet are plotted in
Figure 45. The profiles are nearly identical for medium mesh and fine mesh, which indicates that
the solution is mesh independent. The medium mesh was chosen for further modeling in this
study to save computational time.
For validation purposes, the modeled results were compared with the measured results.
Nine points on the cross section 10 m away from the outlet were monitored and compared with
the measured results. These points are numbered and shown on the contour in Figure 46(a). The
actual velocities were measured using a hot wire anemometer at each point for one min. The
average value was used as the measured value. Table 11 shows the velocity values at each point
and the error compared to the measured values. As can be seen, the error can be up to 12%, but
most of the errors are under 5%. The errors are acceptable since the flow in the mine entry was
constantly changing due to factors such as truck movement. The flow variation can cause large
measurement error as well. Thus, the CFD model is considered well agreed with the measured
data, especially in the high and the low velocity regions. The contour shown in Figure 46(a) also
indicates that the water pipes and conveyer belt have significant influence on the flow
distribution.
Figure 46. Velocity and SF contour at cross section 10 m from the outlet (used case 2 as an example)
6
Table 11. CFD and measured velocity comparison
Point Number P1 P2 P3 P4 P5 P6 P7 P8 P9
CFD Result 2.57573 2.97129 2.59509 3.36968 3.38794 2.92456 3.14654 3.32651 2.80662
Measured Result 2.63682 2.96578 2.34262 3.55174 3.22278 2.67142 3.14568 3.14054 2.50506
Error 2.3% 0.2% 10.8% 5.1% 5.1% 9.5% 0.0% 5.9% 12.0%
96
|
Virginia Tech
|
The validated model was used to model SF transport. As discussed previously in Section
6
6.4.2, SF was not evenly distributed over the cross section in the 2D model. Figure 46(b) shows
6
the SF contour from the 3D model 130 s after the tracer was released. Case 2 is used as an
6
example in this figure, but the tracer distribution is very similar for all four cases. As can be
seen, SF concentration is higher at the top right corner and lower at the bottom left corner. This
6
indicates that although SF is heavier than air, it does not necessarily concentrate on the bottom
6
since air flow features can overcome gravitational influences. This result also agrees with the 2D
model. Several monitor points, which are shown in the figure, were set up to monitor the SF
6
concentration change with time.
Because monitor point L2P7 is the corresponding point with Sample Point 7 in the 2D
model shown in Figure 38, the profile at this point was plotted as solid lines together with the 2D
results in Figure 47 for comparison purpose. As can be seen, the 3D result profiles are generally
lower than the 2D results, and the peak arrival times are also different. However, for different
ventilation scenarios, they provide similar profile trends regarding the relative tracer arrival time
and concentration level. As will be discussed later, the 3D results are more accurate compared to
the field test measured results. Sample point L2P7 in Figure 46(b) is about 2 m high in the entry,
which is not a convenient sampling point. The actual sampling point was chosen to be L3P7,
which is about 1 m from the floor. The SF concentration profile at this point is plotted in Figure
6
48 using dashed lines. Compared to the profiles at L2P7, they have a lower concentration level
because the distribution of SF is low toward the bottom part of the entry. The comparison to the
6
actual measured profiles will be discussed in the next section.
16000
)
14000
b
p12000 (2D) Case 1: door open, fan on (3D) Case 1: door open, fan on
p
( n10000 (2D) Case 2: door close, fan on (3D) Case 2: door close, fan on
o it 8000 (2D) Case 3: door open, fan off (3D) Case 3: door open, fan off
a
r (2D) Case 4: door close, fan off (3D) Case 4: door close, fan off
t n 6000
e
c n 4000
o
C 2000
6
F S 0
60 120 180 240 300 360 420 480 540 600 660 720
Flow Time after Tracer Released (s)
Figure 47. SF concentration change with time
6
97
|
Virginia Tech
|
6.5 Onsite experiment
6.5.1 Tracer gas release, sampling, and analysis
The pulse release technique was used to deploy the tracer. Air samples were taken at 5 s
intervals using the 10 ml blood collection Vacutainers. A Vacutainer is an evacuated glass tube-
shaped vessel and is capped with a self-sealing rubber septum. It has been used extensively for
sampling mine air and products of combustion because they are convenient and result in high
precision even after one to two weeks of storage [131], [132]. These evacuated containers are not
completely evacuated at the time purchased, so they were further evacuated in the laboratory to
improve the sampling accuracy.
A GC equipped with an electron capture detector (ECD) was used to analyze the
concentrations of SF in the collected samples. After gas samples were collected using
6
Vacutainers and brought to the GC laboratory, 20 ΞΌl of the gas sample were taken from the
Vacutainer and injected to the GC using a 100 ΞΌl gas-tight glass syringe.
6.5.2 Onsite experimental results
The onsite experiments were conducted according to the tracer test parameters produced
by the CFD model described in the previous section. These parameters are also summarized in
Table 12. The measured tracer profiles are plotted in Figure 48. It shows that the profile for the
different cases compared well with the CFD modeled results: the tracer arrival time is earlier and
the maximum concentration level is lower when the booster fan is on compared to when it is off;
the concentration profile is flatter and much broader when the door is closed compared to when it
is open. However, the modeled concentration levels are generally 30% lower than that of the
measured results, and the tracer arrival time differences are smaller for the measured results.
These errors can be caused by many factors, such as the model geometry not being
exactly the same as the actual geometry due to measurement limitations and simplifications; the
ventilation state was also constantly changing due to large vehicle movement or shift changes.
However, there are two major factors that may contribute the most to the error: the first one is
that the total flow quantity is about 13% less at the time of the field tracer test than the total flow
measured earlier to build the CFD model, and the second one is that the flow quantity difference
created by the booster fan is smaller than what was measured before. The first factor can lead to
98
|
Virginia Tech
|
the concentration level raising, and the second factor causes the tracer arrival time differences to
become smaller with the booster fan on and off.
There are a couple of possible explanations for the first factor. During the time of the
field experiments, a stopping door near another auxiliary fan located further away in the belt
entry was kept open due to maintenance. This caused recirculation that reduced the total flow
quantity. The CFD model was built in the summer of 2012, but the field experiments were
conducted 6 months later in the winter. The barometric pressure change from summer to winter
can reduce the total air flow quantity as well. The reason for the second factor is uncertain. It
could be because the velocity measurements contained errors in the summer, the barometric
pressure changed, or the mine configuration changed. The last two reasons could have caused
part of the air quantity created by the booster fan to go to other places instead of the location
where the experiments were conducted.
Table 12. Onsite tracer test parameters
Release Location Release Rate Sampling Location Sampling Frequency Total Sample Time
Point L3P7 in Figure
Point 1 in Figure 38 8.8 L/min 5 s 500 s
46(b)
Nevertheless, the CFD models were not adjusted according to the air flow quantities
measured in the winter since any mine ventilation system is dynamic and would continue to
change. The CFD model was also not updated because it is not practical to build the model and
conduct the tracer test in the same day to account for changes. In the cases studied in this paper,
although the total flow quantity was changed, the CFD model still successfully provided good
results that can be used to design the tracer test and predict expected ventilation statuses. As can
be seen in Figure 48, both the CFD and the field test results indicated that the tracer arrival time
is earlier but maximum concentration level is lower when the booster fan is on compared to
when it is off. The concentration profiles are much flatter and have a long tail when the door is
closed compared to when it is open.
99
|
Virginia Tech
|
(CFD) Case 1: door open, fan on (Field test) Case 1: door open, fan on
(CFD) Case 2: door close, fan on (Field test) Case 2: door close, fan on
(CFD) Case 3: door open, fan off (Field test) Case 3: door open, fan off
7,000
(CFD) Case 4: door close, fan off (Field test) Case 4: door close, fan off
6,000
)
b
p
p5,000
(
n
o
it4,000
a
r
t
n
e3,000
c
n
o
C2,000
6
F
S
1,000
0
30 60 90 120 150 180 210 240 270 300
Flow Time after Tracer Released (s)
Figure 48. Modeled and measured SF6 profile comparison
6.6 Conclusions and discussion
In conjunction with the laboratory experiments conducted earlier [143], on-site
experiments and CFD studies were conducted in this study to examine the methodology using
tracer gas and CFD modeling to remotely analyze underground ventilation systems in the field.
A mine section with a 250 m belt entry connected to a 40 m crosscut entry was chosen.
Four different test ventilation scenarios were created by opening and closing the stopping door,
and turning the booster fan on and off. Ventilation surveys and mine entry dimension
measurements were conducted before the tracer release experiments, which provided information
for the 2D and 3D CFD models. The CFD models were used to determine the optimal tracer gas
test parameters, such as the tracer release locations, rate, and duration, and the sampling
locations. The 2D model was used to provide preliminary tracer gas test parameters. After these
parameters were determined, the 3D model was used to obtain more accurate results.
The on-site tracer experiments were conducted according to the parameters determined
by the CFD model. A flow meter was used to control the tracer release rate. Air samples were
taken at 5 s intervals using the 10 ml evacuated containers. A GC equipped with an electron
capture detector (ECD) was used to analyze the concentrations of SF in the collected samples.
6
The CFD model results compared well with the experimental results with acceptable differences,
which are explained in the previous section.
100
|
Virginia Tech
|
This study indicates that different ventilation statuses will result in different tracer gas
distributions. The developed methodology uses CFD modeling and tracer gas test to remotely
analyze ventilation systems in subsurface excavations proved to be feasible both in the
laboratory in another study [143] and in the field. However, detailed ventilation surveys and
mine entry dimension parameters under normal conditions need to be available in order to
establish and calibrate the CFD model. Tracer test results may be more sensitive to certain
ventilation conditions than others depending on how the test is designed. CFD models played an
important role in this study, not only in providing the expected tracer concentration profiles, but
also in determining optimized parameters that would produce the best results and avoid the trial
and error processes. This is extremely important because if the tracer test is not well designed
and fails at first, more time will be squandered in subsequent tests.
This study also identified and incorporated errors that will occur as the result of the
dynamic nature of a mine ventilation system. It is only practical to develop a model of a mine
ahead of time and use it later, especially as applied to mine emergencies. This study
demonstrates that even with the incorporation of these errors the methodology is still valid.
In an actual situation where the ventilation status needs to be determined using this
methodology, only one tracer test could be conducted to produce one concentration profile which
could then be compared to the possible scenarios modeled by CFD. As can be seen from Figure
48, the profile shape was more sensitive to the door status but less sensitive to the booster fan
status. When the door was closed, the profiles have a very long tail that is obviously different
from the bell shaped profiles when the door is open. This is easy to identify. However, if the
door status is the same, the profiles when the booster fan was on or off were very similar. The
major difference is found in the earlier tracer arrival time when the booster fan is on. In this case,
it is hard to determine the booster fan status with the existing CFD model. It requires that the
CFD model be built based on the most recent ventilation survey data and that the modeled tracer
arrival time results are accurate enough to compare with the field tracer test results. Therefore, a
more accurate ventilation survey data under normal conditions is needed for such a case. The
tracer test could also be redesigned, for example by changing the release location, to capture the
fan status more accurately.
In an emergency situation rapid tracer deployment is essential. In this study, tracer tests
were designed using CFD, which eliminated the trial and error processes. The optimized
101
|
Virginia Tech
|
parameters obtained from the models proved to be very useful as tracer data was successfully
obtained after only one release. Tremendous time and resources were saved by reducing such
items as the number of trips to the mine and re-deployment of the tracer.
Estimating the level of damage and possible ventilation scenarios plays an important role
in successfully identifying the actual ventilation scenario. If actual ventilation status was not
accounted for in assumed scenarios, the methodology may fail to identify it. For example, in one
of our blind field tracer tests, the result is shown as the purple dashed line in Figure 49. The
result correctly indicated that the stopping door was open since the concentration profile does not
have a long tail. However, the booster fan status matched Case 3 result the most but with a
significantly lower tracer concentration level at the beginning of the profile. In this case, the best
conclusion that can be made is that the ventilation status follows Case 3, in which the door is
open and the fan is off. This conclusion was found to be inaccurate.
At the time of the blind test, the door was open and the booster fan was on. However,
another booster fan in the belt entry had been turned off and on for maintenance purposes. This
was not within the consideration of the modeled scenarios and therefore led to an inaccurate
prediction. Although this is one limitation of the methodology, it is usually possible to narrow
down the possibilities that are of the most concern. For example, certain booster fans being on or
off may be more important to rescue efforts than others; certain stopping door statuses may also
be more important than others to estimate the extent of an explosion. Therefore, it is still possible
for this methodology to cover major possible ventilation scenarios with the help of experienced
engineers.
6,000
)
b
p p 5,000 Blind Tracer Test
(
n
o 4,000 Case 3: Door Open Fan Off
it
a
r
t n 3,000
e
c
n
o 2,000
C
6
F 1,000
S
0
30 60 90 120 150 180 210 240 270 300
Flow Time after Tracer Released (s)
Figure 49. A blind tracer test result compared to case 3: door open fan off
102
|
Virginia Tech
|
The following paper will be submitted to a peer reviewed journal. The experimental and CFD modeling
work and writing was primarily completed by Guang Xu with editorial and technical input from Dr. Kray
Luxbacher, Edmund Jong, Gerrit Goodman, and Dr. Harold M. McNair.
7 Preliminary Guidelines and Recommendations for Use of Tracer
Gas in Characterization of Underground Mine Ventilation
Networks
7.1 Abstract
Tracer gases are an effective method for assessing mine ventilation systems, especially
when other techniques are impractical. Based on previously completed laboratory and field
experiments, this paper discusses some common problems encountered when using tracer gases
in underground mines. The discussion includes tracer release methods, sampling and analysis
techniques. Additionally, the use of CFD to optimize the design of tracer gas experiments is also
presented. Finally, guidelines and recommendations are provided on the use of tracer gases in the
characterization of underground mine ventilation networks.
7.2 Introduction
Ventilation is a fundamental to the engineering of underground mines because it has
considerable effects on health and safety. Measurements of airflow underground are usually
carried out using traditional instrumentation such as vane anemometers, hot-wire anemometers,
pitot tubes, and smoke tubes. However, these methods are not practical under certain
circumstances in an underground mine. Examples include the measurement of recirculation and
leakage, flow in inaccessible zones, and flow with very low velocities. Sometimes traditional
instrumentation fails to provide accurate results. For example, in a study that investigated jet fan
effectiveness in dead headings, tracer gas results were found to be more accurate than the results
of smoke tubes [144]. Therefore, tracer gas techniques are a valuable tool for accurately
measuring airflow in situations where traditional methods cannot be employed and providing
information to characterize underground mine ventilation.
The tracer gas technique is a useful and versatile tool for studying mine ventilation
systems with a long history of application. The Bureau of Mines [109] conducted a series of
tracer gas tests using sulfur hexafluoride SF and proved the usefulness of tracer gas techniques
6
in measuring recirculation, air leakage, airflow in large cross section, low flow velocity, and
transit air time. Grenier et al. [145] used tracer gases to analyze the spread of dust in a fluorspar
105
|
Virginia Tech
|
milling plant. The results indicated that tracer gases behave in the same physical manner as
respirable dust and can be used to find patterns in dust movement within a ventilation system.
Tracer gas has been accepted by the mining industry as a viable ventilation survey tool. More
examples that used tracer gas to investigate various ventilation problems can be found in Timko
and Thimons paper [144].
An ongoing research project that involves the selection of novel tracer gases for mine
ventilation, the development of a methodology to use tracer gases and computational fluid
dynamics (CFD) modeling to analyze, predict, and confirm the underground ventilation status
together with the location of the damage, and finally validate the developed methodology in the
laboratory and in the field are currently being conducted at Virginia Tech. Details of this work
have been published in several forums [140], [143], [146], [147].
The focus of this paper is to provide preliminary guidelines and recommendations for use
of tracer gas based on experience and practice. As this research progressed it was evident that
there are few resources in the literature that provide the practical aspects of conducting tracer gas
studies in mines. Some essential aspects of the tracer gas technique are discussed as well as new
findings and recommendations, and studies in the literature are referenced as well. Using CFD
modeling to design tracer gas experiments is also presented in this paper. Some modeling
examples are provided to illustrate how CFD can help to determine the optimized tracer release
and sampling locations, the release rate and duration, and eventually help to achieve desired
results. Tracer gas experiments are time and resource consuming in underground mines, the
guidelines and recommendations provided in this paper can be used by other researchers and
industries for the design of tracer gas experiments more efficiently with less trial and error.
7.3 Tracer gas techniques
7.3.1 Choices of tracers
Sulfur hexafluoride (SF ) is a widely accepted standard tracer gas that has been used in
6
mine ventilation studies. SF is non-toxic [148], and the Occupational Safety and Health
6
Administration (OSHA)βs Permissible Exposure Limit (PEL) and the American Conference of
Governmental Industrial Hygienists (ACGIH)βs Threshold Limit Value (TLV) for SF is 1,000
6
ppm [149]. The amount of SF released to a mine is generally much less than either the PEL or
6
TLV limit. SF can be detected accurately using gas chromatography (GC) in concentrations as
6
106
|
Virginia Tech
|
low as parts per trillion. It is also odorless, colorless, chemically and thermally stable, and not
found in the natural environment. It is not measurably adsorbed on sandstone and coal. These are
all desirable properties as a tracer gas [109]. Some alternative gases have also been used as mine
ventilation tracers, such as nitrous oxide and helium. However, because they are not easily
detected, large amounts need to be released thus causing transportation problems and difficulty
in achieving stable flows [150].
It has long been realized that multiple tracer gases can add flexibility to ventilation
surveys in many ways. Multiple gases not only allow the release of different tracers at different
points without increasing the number of collected gas samples, but provide more information
because the source of each tracer in one sample can be identified and air flows in different zones
can be investigated simultaneously with multiple tracers. Once a tracer is released to the mine, it
may take days to weeks for the tracer background to be reduced to a level that will not affect the
next test. However, if multiple tracers are available, a different tracer can be used to conduct
another test right after the previous test. Although the advantages are apparent, the use of the
multiple tracer technique is still not common in underground mines. One key requirement for
identifying other tracer gases is that the gas should be measurable by the same method being
used for SF , which is the most commonly used tracer gas. SF is commonly analyzed by GC, so
6 6
it would be better if the other tracers were able to be analyzed by the same GC method. Using the
same GC method, the additional tracers should have similar sensitivity to SF as well as be able
6
to be separated from SF . Kennedy et al. [150] investigated six Freons that are promising
6
candidate tracer gases. They found that only Freon-13B1 (CBrF ) and Freon-12 (CF Cl ) were
3 2 2
within two orders of magnitude of the sensitivity of SF when analyzed using a GC with an
6
electron capture detector. The other four Freon gases were several orders of magnitude less
sensitive. CBrF and CF Cl were tested in the field and it was concluded that they perform well
3 2 2
as mine ventilation tracer gases and are comparable to SF . Batterman et al. [151] used
6
hexafluorobenzene (HFB) and octafluorotoluene (OFT) for indoor ventilation tracers. However,
those two tracers were not simultaneously analyzed with SF . Patterson [152] researched the
6
selection of novel tracer gases that can be used in mines together with SF . Freon 14 (CF ), C F ,
6 4 3 8
and PMCH (C F ) were tested on different columns using various GC methods. CF and C F
7 14 4 3 8
were found to have much less sensitivity than SF on the tested columns, with similar retention
6
times to SF . PMCH was reported to be a appropriate tracer if used together with SF . A GC
6 6
107
|
Virginia Tech
|
protocol was developed that can be used to analyze SF and PMCH concurrently on an HP-AL/S
6
capillary column. One drawback of the protocol is the long 18 min analysis time. This process
can potentially be optimized and shortened.
In summary, as for the choice of tracers in underground mines, SF is no doubt the best
6
choice. CBrF ,CF Cl ,and PMCH (C F ) can be used together with SF . CBrF and CF Cl
3 2 2 7 14 6 3 2 2
have both been successfully used in a mine. PMCH is a novel underground mine tracer gas. The
authorβs research group is developing method to apply it as additional tracer that can be used
together with SF .
6
7.3.2 Tracer gas release technique
Accurate and precise release of tracer gas is critical to conducting a rigorous study; there
must be high confidence in the mass or rate of gas released to the system in order to achieve
meaningful analysis of results. There are two commonly used tracer gas release techniques:
pulse-injection, which is based on the injection of a short duration of tracer gas, and constant-
injection, which is based on the continuous injection of tracer gas.
Tracer gases can be released in a controlled manner using various methods. For pulse
injection, a known mass of tracer gas in a balloon or syringe can be used and injected to the
mine. Or directly release tracer gas from a pressurized container and determine the weight loss of
the container after release. However, it is hard for these methods to release in a controlled rate.
Flow meters and permeation tubes serve as more accurate controlled rate tracer gas release
methods, and they can be used for both pulse and constant injection.
Soap bubble flow meter (Figure 50 a), rotameter (Figure 50 b), and electrical flow meter
(Figure 50 c) are commonly used flow meters that can be used to measure and to control the flow
rate. The soap bubble flow meter measures the flow rate of tracer release, but accessory
instruments are needed to control the flow rate of gas. For gases contained in a compressed gas
bottle, a two stage regulator is often attached to the bottle. Capillary tubing may be used to
connect the regulator and the soap bubble flow meter adding more resistance to achieve better
gas flow control. The two stage regulator controls and maintains the delivery pressure to the flow
meter as long as gas tank pressure is greater than the delivery pressure. Selecting different
capillary tubing and adjusting the delivery pressure affects the flow rate. The described setup
provides flow stability unaffected by atmospheric pressure [145].
108
|
Virginia Tech
|
Variable area flow meters, also known as rotameters, provide a different set of controls.
The design is based on the variable area principle and indicates as well as controls the rate of
flow. Positioned vertically, the gas flow lifts the float in the flow tube. For a constant flow rate,
the float will be at a stationary position, which corresponds to a point on the measurement scale
that indicates the flow rate. The flow rate can be changed by adjusting the delivery pressure or
manipulating the valve on the flow meter. Although a specific meter can be purchased for a
certain gas, the air flow meters are the most commonly found in the market. The air flow meter
can be used for other gases with a correction factor provided by the manufacture.
An electrical mass flow controller can be used to accurately control the release quantity
of a tracer. Using a mass flow controller, the flow rate can be adjusted using the digital control
panel. It requires minimal adjustment to achieve the desired flow rate compared to the other
types of flow meters.
Figure 50. Different flow meters
The permeation device is a commonly used tracer release source for volatile compounds.
The basic concept is to seal a certain amount of tracer in an impermeable tube with permeable
material at one end. The emission rate can be determined by weighing the prepared tube at
intervals of several days until equilibrium is reached. The emission rate is relatively stable if the
temperature is constant. [153]. Dietz and Cote [154] described a perfluorocarbon (PFT) source in
which a known mass of PFT is injected into a fluoroelastomer plug and crimped in a metal shell.
PFT will diffuse from the end of the plug at a known rate that is inversely proportional to the
square root of time for the emission of the first 50%-60% of the original amount of PFT. Johnson
et al. [155] described the preparation of a permeation device they used for continuous and
constant release of SF . It was constructed from brass rod, Teflon, a frit, and a swaglok nut.
6
109
|
Virginia Tech
|
Although their design was not used in a mine tracer test, the method can be used for the release
of SF and other tracer gases in mines. Batterman et al. [151] described an updated method for a
6
constant injection technique with miniature PFT sources. The method was shown to be reliable
for measuring indoor air exchange rates. The source was a diffusion controlled release of
saturated vapor in the headspace of a PFT liquid container. A diffusion tube was inserted into the
septum that sealed a glass vial partially filled with PFT. The emission rate was evaluated using a
Fickian diffusion model and tested by experimentation. The described sources mentioned above
can be easily modified for a wide range of applications. In general, the permeation device release
methods are designed for very low emission rates over a long period of time, and this method has
not been used in underground mines.
7.3.3 Tracer gas sampling methods
There are two categories of gas sampling methods: collecting samples for laboratory
analysis and collecting samples for immediate analysis [156]. The first category is commonly
used in underground tracer gas studies, which includes gas sampling bags, hypodermic syringes,
and evacuated containers.
Gas sampling bags have been successfully used for a number of years to collect gas
samples and make gas standards. Gas sampling bags are available commercially and come in a
variety of sizes and shapes, and are made from a number of materials, such as PVC (polyvinyl
chloride) or Tedlar (polyvinyl fluoride). One needs to pay special attention to the material of the
sampling bag because the sampled gas may be reactive, adsorptive, absorptive, or diffusive with
the bag materials [156]. There are many applications of using gas sampling bags for collecting
gas and vapor samples [157]. It usually requires a pump to inject gas into the bag, so it is not a
very convenient tool for dynamic gas sampling. Kennedy [150] tested the tightness of TEDLAR
gas sampling bags and no detectable degradation of gas samples was found in a 24 hour period.
The storage time is typically no more than 24 hours, so analysis should be conducted as soon as
possible, unless storage experiments indicate a longer storage time.
Hypodermic syringes satisfy most underground sampling requirements. Syringes are
inexpensive, easy to carry, and hard to break. However, their short storage time restricts their
use. For example, in one study, gases such as carbon dioxide are lost rapidly due to permeation
[131]. It is standard practice to fill and evacuate the syringe twice before drawing so that the gas
110
|
Virginia Tech
|
left in the syringe does not contaminate the gas sample. The syringe can be sealed with a vinyl
cap [145]. The tightness of hypodermic syringes have been tested by Kennedy et al. [150]. In the
study, six syringes were filled with certain concentrations of SF , CBrF , and CF Cl , and left for
6 3 2 2
24 hours. The concentrations did show a significant reduction, with a loss of 0.5-1 percent for
SF , 2.5-3 percent for CBrF , and 6 percent for CF Cl . These syringes were emptied and refilled
6 3 2 2
with the same standard gas, left for another 24 hours, and tested again for the tracer gas content.
The loss of tracer was reduced to half of the original percentage after the first filling. This
indicated that the loss was not only due to permeation through the walls or leaks in the seals but
also due to adsorption of the tracer onto the syringes. Like sampling bags, samples generally
should be analyzed on the same day they are taken.
Vacutainers have long been used for sampling mine gases. A Vacutainer is an evacuated
glass or plastic tube-shaped vessel and is capped with a self-sealing rubber septum. Such
containers are commonly used to take blood samples. The advantages of Vacutainers are that
they are small, light-weight, economical, convenient, and simple to use [156]. The Bureau of
Mines officially adopted the Vacutainers for taking mine gas samples because they are
convenient and can obtain consistent results [131]. Freedman et al. [131] tested the 10 ml
Vacutainers for use in mines. In their study, a device was described which can evacuate up to 56
Vacutainers to a few millimeters of pressure. It was found that for stored gas samples, CO
2
shows substantial loss in concentration over the 41 days due to permeation. The CO
concentration level gradually increased up to 50 ppm over time in factory supplied or completely
evacuated Vacutainers. This phenomenon is unexplained. They recommended that the storage of
evacuated Vacutainers should not exceed 1 to 2 months if low levels of CO can cause
interference. If precise CO level is of interest in collected samples, the analysis should be done
2
within 1 week after samples are taken.
The 10 ml Vacutainer was also chosen as the sampling method for our experiments.
These evacuated containers are not completely evacuated at the time purchased and a small
amount of gas (air) is still present. However, the containers are evacuated to a fixed and
designated pressure [158]. The factory evacuation is designed to draw 10 ml of blood.
Preliminary tests indicate they draw 9.5 ml of water. The left three Vacutainers in Figure 51
illustrate this result. It was noted that the capability of drawing water to be reduce slowly with
time. The actual Vacutainer capacity measured by water displacement is 12.9 ml. Therefore, the
111
|
Virginia Tech
|
Vacutainers were re-evacuated in the laboratory to improve the sampling accuracy. The right
three Vacutainers in Figure 51 show the result of laboratory evacuation, which improved the
capability of Vacutainers of drawing 12.5 ml of water. This means 0.4 ml of air is left in each
Vacutainer, which is 3.1% of the actual capacity. This will make the measured gas concentration
3.1% less than the actual gas sample concentration, but it is acceptable for most tracer gas
analysis purposes.
Figure 51. Vacutainers factory evacuation and laboratory evacuation
The evacuation system used in the laboratory is shown in Figure 52. Basically it is a
vacuum pump connected to several needles so that several Vacutainers can be evacuated at a
time. Vacutainers are inerted onto each of the needles through the septum for about 30 s, and
need to be pulled off very slowly, which allows enough time for the septum to re-seal for a high
quality evacuation. Each needle connected to the system needs to be replaced after about 5
Vacutainer evacuations since it will dull and hard to penetrate the Vacutainer septa.
Vacuum Tube Ball Valve
PVC Pipe
Vacuum Pump Vacuum Gauge
Needle
Figure 52. Evacuation apparatus schematic
112
|
Virginia Tech
|
Although the Vacutainer storage time is promising, an official time has not been reported
in the literature. An attempt to test its storage time for SF was conducted during a month period
6
and no obvious SF concentration changes were found. However, the method used introduced
6
significant amounts of systematic errors that are difficult to control and quantify.
100.0%
n
o 99.0%
it
a
r
t
n 98.0%
e 97.9%
c
n
o 97.0%
C 96.8%
e 96.6% 97.1%
lp
m 96.0%
96.9%
96.5%
96.9%
a
S
la 95.0%
u
t Concentration in Vacutainer
c
A 94.0%
e 93.8%
h Expected Concentration
t
f 93.0%
o
e
g
a 92.0%
t
n
e
c r 91.0%
e
P
90.0%
0 1 2 3 4 5 6 7 8
Sample Time (s)
Figure 53. Vacutainer sample time
The Vacutainer sample time was also studied to determine how fast the gas sample will
fill the Vacutainers. The sample time is defined as beginning when the rubber septum is
punctured by a needle ending when the needle is removed from the septum. This is important,
especially for dynamic sampling, because it determines the minimum sampling interval. As in
the previous test, for each test, three samples were taken and the average concentration was used
to compare with other results. The test results are shown in Figure 53. Given adequate time for
the Vacutainer to draw sample, the expected concentration in the Vacutainer should be about
96.9% of the actual concentration, as mentioned before. The 96.9% concentration level is
marked in the dashed line in Figure 53. As can be seen, the concentration in the Vacutainer
reached the expected level with sampling times longer than 2 seconds. Variations exist which are
likely due to error during Vacutainer evacuation and GC manual injection. Therefore, for the 10
ml Vacutainer, the least sample time is recommended to be 2 seconds. Dynamic SF samples
6
were successfully taken every 5 seconds during 8 minutes period in our underground tracer tests.
In each 5 seconds, Vacutainer takes sample for 3 seconds and 2 seconds are needed for changing
113
|
Virginia Tech
|
to a new Vacutainer. This is the fast minimum sampling interval we can achieve if only one
people takes samples manually.
7.3.4 Analysis of tracer gas
There are two main techniques for determining gas concentrations: infrared spectroscopy
(IR) and gas chromatography (GC). This paper focuses on the GC method which is one of the
most widely used analytical techniques for gas samples due to its selectivity and sensitivity. A
wide range of compounds can be analyzed through the proper selection of columns and
detectors.
Gas chromatographs need to be calibrated over its operating range to quantitatively
measure a certain gas concentration. Thus, tracer standards are needed for the calibration. The
standards can be prepared in the laboratory or purchased as a certified mixture. To make the
standards in the laboratory, a gas-tight syringe can be used to inject a small, known quantity of
tracer into a known volume of air or ultra-pure nitrogen in a sealed container. An example of one
of these containers is shown in Figure 54. With both valves open, this container can be flushed
using ultra-pure nitrogen and filled with it after both valves closed. To prepare standards for
trace concentrations, a serial dilution technique can be used. Serial dilutions are completed by
withdrawing a measured quantity of the mixture prepared in the first step and injecting it into
another container. The procedure can be repeated until the desired concentration is achieved.
This method is time consuming and errors can be introduced at each stage of dilution. Although
this method can be used to produce useable calibration curves, the purchased certified gas
standards can provide better accuracy and repeatability [159].
Figure 54. Glass bulb for tracer gas standard preparation
Split injection is a widely used injection technique for GC analysis. It automatically
reduces the sample size to prevent the column from overloading by allowing only a fraction of
the sample enters the column. This fraction is defined as a split ratio. A split ratio of 20:1
indicates that one part of the sample enters the column and 20 parts exit the GC system through
the split vent [160]. Theoretically, this means 1/21 of the total sample is analyzed. Again,
114
|
Virginia Tech
|
theoretically, this makes it possible to compare two results with the same GC method but with
different split ratio. For example, a sample analyzed with a split ratio of 50:1 could be corrected
to 20:1 by multiplying a factor of 51/21. However, this split ratio calculation was found to be
inaccurate due to the mechanical nature of the splitting control. Figure 55 shows the measured
results of the same SF sample with different split ratio but all corrected to the split ratio of 20:1.
6
As can be seen, this correction method fails to provide accurate results. Although the accuracy of
the split on different GC machines may vary, it is recommended that a single split ratio is chosen
and that corrections to the data based on split ratio are avoided.
n 160%
154%
Measured Concentration Corrected to 20:1
o
it a r140% 135% Split Ratio
t n Actual Concentration
e
c n120%
o C 100%
la100%
u
t c 77%
A 80%
e 63%
h
t f 60% 53%
o
e
g a 40%
t
n
e
c 20%
r
e
P
0%
SR 10:1 SR 15:1 SR 20:1 SR 30:1 SR 50:1 SR 100:1
Split Ratio
Figure 55. Same sample measured with different split ratio comparison
Since this correction method is not always reliable, the split ratio should be held constant
for all calibration standards and gas samples. A calibration curve is used to quantify the
concentration in collected samples. To create a calibration curve, the analysis concentration
range needs to be determined first, and then start from a reasonable injection amount and split
ratio for the calibration. It is better to inject the highest concentration standard first to make sure
the column will not be overloaded. If it is overloaded, increase the split ratio or reduce the
amount of injected sample. Afterward, inject the lowest concentration standard to ensure that the
concentration is above the detection limit; otherwise, the split ratio needs to be decreased or a
larger amount of sample needs to be injected. After the injection volume and split ratio are
determined, these parameters need to be the same for the rest of the standards and samples that
need the generated calibration curve.
115
|
Virginia Tech
|
7.4 The use of CFD model for tracer gas design
Tracer gas experiments are time consuming, especially if they are conducted on a trial
and error basis. Careful planning will ensure meaningful data collected efficiently. CFD
modeling is a powerful tool that can be used for experimental design. This tool can substantially
reduce the effort that is expended in the field. In CFD modeling, tracers can be modeled in
several ways. The two most common methods are the two species transport model or definition
of a user defined scalar to model tracers. Parameters can be easily changed in the CFD model,
such as the location and the release amount. This provides detailed information to achieve the
desired results, which can help with the tracer gas experiment design. This section focuses on the
discussion of the determination of some parameters using CFD modeling in the characterization
of underground mine ventilation networks.
One important parameter is that the expected concentration in collected samples needs to
be within the detection range of the GC. It is known that the concentration decreases as the
distance between the sampling and release location increases. If these locations have been
determined, the tracer gas release rate and duration can be used to determine the concentration
level.
The release rate is the most sensitive parameter that influences the concentration. Figure
56 shows the SF concentration profile results provided by one of the studies conducted by the
6
authors. The scenario modeled was a belt entry connected with a crosscut entry shown in Figure
59. The release location is in the crosscut and is denoted as βRelease Point 1.β The sampling
location is 190 m away from the crosscut and is downstream from the velocity inlet in the belt
entry. SF was released under different rates for one minute in the CFD model. As can be seen in
6
Figure 56, the basic shape of the concentration profile will not change by changing the release
rate, but the profile is taller under larger release rate. The result can be used to find an optimized
release rate that can produce a tracer concentration profile, which can be reasonably analyzed by
GC or other instruments.
116
|
Virginia Tech
|
300
250
)m
9 L/min
p 30 L/min
p 200
( 200 L/min
n
o
it
a r 150
t
n
e
c
n
o 100
c
6
F
S
50
0
0 50 100 150 200
Time after tracer release
Figure 56. Tracer concentration profile under different release rate
The release duration also influences the concentration levels in collected samples but is
not as sensitive as the release rate. Under certain circumstances, increasing the release duration
will not necessarily increase the concentration level. In the same CFD model mentioned above,
different tracer release durations were modeled under the same release rate (9 L/min). As can be
seen in Figure 57, the 30 s release increased the concentration level when compared to the 10 s
release. However, increasing the release duration further, to 1 min and 2 min, did not increase the
concentration level. That is because the concentration reached its maximum under the fixed
release rate and the maximum concentration leveled out. Although this is not always the case, the
CFD results can provide the data that can be used to pre-determine the concentration level under
certain release duration.
14
)
m12 Release 10 s
p
p Release 30 s
(10
n Release 1 min
o
it 8 Release 2 min
a
r
t
n 6
e
c
n
o 4
C
6
F 2
S
0
50 70 90 110 130 150 170 190 210 230 250
Time after release (s)
Figure 57. SF6 concentration profile under different release duration
117
|
Virginia Tech
|
Another important parameter that CFD can help to determine is the shape of the expected
concentration profile. The farther the sampling location is from the release location, the broader
and flatter the profile becomes. If the release and sampling locations are determined, longer
release durations can produce broader profiles as well. Figure 57 shows that the concentration
profile becomes broader as the release duration increased from 10 s to 2 min. Generally, a
broader profile is better, especially when the sampling method is not continuous, such as syringe
or Vacutainer sampling. A broader profile allows enough samples to be taken to accurately
resolve the profile. If the profile is too narrow, for example it is only 10 s, but the sampling
interval is 5 seconds, only 3 samples maximum can be taken within the 10 seconds time. This
makes it very hard to accurately depict the concentration profile.
250 Release 1 min 1800
n
im
1
e
Release 30 min 1600
n
im
0
3
s a200
1400 e
e s
le a
e
R
r o
f )b
p
p150
11 02 00 00 le
R
r o f
)b
( n o it a r tn100 68 00 00 p p ( n o it
a
r
e
c
tn
n 400 e
o C
6 F
S
50
200
c n o
C
6
F
S
0 0
0 10 20 30 40 50 60 70 80 90 100 110 120
Flow Time (min)
Figure 58. Tracer profile under different release duration
However, in some cases, too broad of a profile will not benefit the tracer experiment not
only because it takes longer for sampling, but also because it weakens the profile
characterization. As indicated from a CFD study presented in [140], under the normal ventilation
scenario, there are four flow paths. If the tracer is released for 1 min, the profile is the solid line
displayed in Figure 58, which has four peaks and each of them represent one flow path.
However, if the release duration changes to 30 min, the profile is the dashed line shown in Figure
58. It not only requires 30 min more sampling time to capture the entire profile, but the
characterization is also not as obvious as before. It is hard to tell that there are four flow paths
from this kind of tracer profile. In summary, too short of a release duration may produce tracer
profiles that are hard to capture with certain sampling intervals; too long of a release duration
118
|
Virginia Tech
|
may weaken the tracer profiles characterization. CFD modeling can be used to check if the
release duration is good enough for certain sampling intervals and to characterize certain
ventilation scenarios.
Changing the location of tracer release point will not only change the profile peak arrival
time but also change the profile shape when the flow feature is complex. For example, in the 2D
CFD model mentioned before and shown in Figure 59. Two tracer release points were examined
in the model: point 1 is 38 m from the door, and point 2 is 2.5 m from the door. As can be seen in
Figure 60, due to the recirculating flow featured in the zone at point 2, the downstream tracer
profile not only appears later and broader compared to release at point 1, but the basic shape is
also changed. The profile is much broader and has a longer tail. In this case, the basic tracer
profile shape is different when released in the same entry but at different locations. The CFD
model can identify these complex flow zones, which provides information that helps to decide
where to release the tracer and get a profile that is suitable for certain purposes.
Figure 59. Tracer release at different locations
119
|
Virginia Tech
|
14000
) b 12000
p
p Release 38 m from the door
(
n
10000
o Release 2.5 m from the door
it 8000
a
r
t
n 6000
e
c
n
o 4000
C
6
F 2000
S
0
60 120 180 240 300 360 420
Time (s)
Figure 60. Tracer profile under different release location
The examples shown above demonstrate that CFD is a very powerful tool as an aid for
tracer gas experimental design. The modeling results can provide information for determining
some of the key experimental parameters mentioned above. Using these optimized parameters to
perform the actual on-site experiment can help to avoid the trial and error process and obtain the
desired results efficiently. However, the CFD model requires much more computational power
compared to other modeling methods, such as network modeling. Using a high performance
computer for the modeling is a solution to reduce computational time. In addition, starting from a
2D model can also save time tremendously. Although 2D flow is totally different from 3D flow,
and the 2D model cannot provide as accurate a result as 3D, a 2D model can still provide enough
information to determine most of the influencing factors. After those factors have been
determined, using a 3D model to validate the 2D model can increase the confidence and the
accuracy of the results. Finally, in cases where the flow regime is not complex, and the sampling
and collection points relative to the entry cross-section are not considered critical to results,
network modeling can even be used to understand expected results. The advantage of network
modeling is much lower computational requirements along with less user skill.
7.5 Conclusions and discussion
The tracer gas technique is a precise and reliable methodology for characterizing
underground mine ventilation networks. However, it is time consuming and resource heavy,
especially when it is based on a trial and error basis. The guidelines and recommendations
provided in this paper are based on the experiences and practices of the tracer gas research
120
|
Virginia Tech
|
conducted over the last few years. These recommendations can help other researchers and
industries to reduce the effort in conducting tracer gas experiments.
Several common topics in the tracer gas techniques are discussed. As for the choice of
tracers, CBrF and CF Cl are found to have been successfully used together with SF using the
3 2 2 6
same GC method. PMCH was also found to be a good tracer for use with SF . However, PMCH
6
has not yet been proven in underground mines.
As for the tracer release techniques, commonly used flow meters are discussed and
compared. Additionally, tracer release by permeation tube is introduced although has yet to be
used in mines. It may prove to be a promising technique for tracer release in mines. These
release methods control the flow rate differently and can be used to serve specific purposes.
There are several tracer gas sampling methods that have been used in mines. Both gas
sampling bags and hypodermic syringes have a short storage time. Samples taken using these
methods should be analyzed within 24 hours. The material of the sampling bags could affect its
sampling capability for certain gas samples. It was reported in the literature that the sample loss
in the hypodermic syringes are due to permeation through the walls, leaks in the seals, and
adsorption of the tracer onto the syringes. Vacutainers have the longest storage time compared to
other sampling methods. However, substantial loss of CO in stored samples was reported due to
2
permeation. CO concentration levels were found to be gradually increased over time in
evacuated Vacutainers in one instance. Storage time and sample time for SF have not been
6
officially reported in the literature. An attempt to test its storage time for SF was conducted
6
during a month period and no obvious SF concentration changes were found. However, the
6
method used introduced significant amounts of systematic errors that are difficult to control and
quantify. The minimum sample time of the Vacutainers is found to be 2 seconds, however, the
minimum practical sampling interval for one person sampling manually is 5 seconds.
GC is a commonly used method for tracer analysis. The preparation of gas standards in
the laboratory was described, which can be used to calibrate a GC when a certified gas mixture is
not available. The correction method from one injection split ratio to another is discussed and
found to be inaccurate, thus it should be avoided unless experiments show it is acceptably
accurate. A procedure was proposed for generation of a calibration curve. These guidelines can
be used to save time and guarantee an accurate GC result.
121
|
Virginia Tech
|
CFD modeling can help with the tracer experiment design. Examples are presented which
illustrate how CFD modeling helps to determine the important factors in a tracer experiment,
such as the release rate and duration, the expected concentration profile, and the release location.
After these parameters are optimized in the CFD model, the trial and error process can be
reduced and the desired results can be obtained more efficiently. However, for a large scale 3D
CFD model, the computational time is considerable. Many of the parameters can be studied
using a 2D model to save time. But the 2D model cannot provide as accurate result as the 3D
model because 2D flow is totally different from 3D flow.
The aim of this work is to provide a practical guide for people with technical expertise
who want to apply tracer gas techniques to mine ventilation. Although many impressive
applications of tracer gas techniques are available in the literature, few, if any sources are
available that provide a guide use of the method. This method can allow for characterization of
the ventilation system from an assessment of leakage to the efficiency of designs for gas and dust
dilution, and provide engineers with additional tools for improving health and safety.
122
|
Virginia Tech
|
8 Conclusions and Discussions
8.1 Conclusions
The objective of this research was to develop a new methodology that can characterize
the underground mine ventilation systems remotely using tracer gas techniques and the CFD
modeling method. This provides an alternate way to gather information that can be used for mine
personnel and rescuers to take safe and effective actions after an unexpected event.
Ultimately, this work demonstrated that general determination of changes to a mine
ventilation system is achievable through examination of tracer gas profiles, both at the lab and
field scale, for transient and steady state release. Also, this work has informed the practical use
of tracer gases in mines, and this body of knowledge is expected to contribute to more efficient
and more common use of tracer gases by mine engineers, which will allow for better
characterization of mine ventilation system and improved safety.
Experiments were conducted in the laboratory first before going to the field. A simplified
conceptual mine model built with PVC pipes was used for tracer gas experiments. Different
ventilation scenarios were simulated by opening or closing different valves. Instead of using
pulse release, tracer gas was released constantly at designated location of the model mine. This is
because the small size of the model mine caused that the tracer gas concentration profile lasts a
short period of time, and could not be monitored frequently enough to be resolved. CFD models
were built for assumed ventilation scenarios, and the results compared well with those were
measured with reasonable errors. Ventilation scenarios can be predicted by comparing the
experimental data and the CFD results. This laboratory study prepared for the on-site
experiments, such as developing a proper GC method and sampling method for the field
experiments. It also indicated that tracer gas parameters need to be optimized in order to obtain
substantially different tracer gas profiles for different ventilation scenarios.
Based on the conceptual model mine laboratory experiments, a full scale model mine
CFD simulation was conducted. This full scale mine CFD model allows for pulse release of
tracer gas at the inlet and monitoring of tracer gas at the exhaust. Tracer gas concentration
profiles are different because different ventilation scenarios have different air flow paths, which
allows for analysis and prediction of the ventilation status. However, this work shows that tracer
gas test parameters need to be optimized to successfully characterize each scenario. For example,
123
|
Virginia Tech
|
the amount of tracer gas released determines whether or not its concentration at the sampling
location can be practically detected; the release duration determines whether or not the
concentration profiles can be easily separated for different ventilation scenarios.
Following the previous studies, field experiments were conducted to examine the
developed methodology in the field. A 2D CFD model was used to determine the optimal tracer
gas test parameters, such as the tracer release locations, rate, and duration, and the sampling
locations. After these parameters were determined, the 3D model was used to obtain more
accurate results. The optimized parameters obtained from the models proved to be very useful as
tracer data was successfully obtained after only one release. This is essential for rapid
deployment of tracer in an emergency situation. However, detailed ventilation surveys and mine
entry dimension parameters under normal conditions need to be available in order to establish
and calibrate the CFD model. Computational resources required for this work illustrate that it is
only practical to develop a model of a mine ahead of time and use it later, especially when
applied to mine emergencies. Errors will occur as the result of the dynamic nature of a mine
ventilation system, but this study demonstrated that the methodology is still valid even with these
errors if the tracer gas test is carefully designed.
Finally, based on the laboratory and field scale work, a practical guide for people who
want to apply tracer gas techniques to mine ventilation was provided. Some common topics of
tracer gas techniques were discussed as well as new findings, such as the Vacutainer sample
retention time and the inaccuracy of GC split ratio conversions. These recommendations can help
other researchers and industries to reduce the effort in conducting tracer gas experiments and
make them more attractive to operators. Examples are presented on the use of CFD modeling to
determine the important factors in a tracer experiment, such as the release rate and duration, the
expected concentration profile, and the release location. The trial and error process can be
reduced and the desired results can be obtained more efficiently with the help of CFD modeling.
Although emergency situations need to be considered case by case, some rapid
suggestions on the use of this methodology can always be provided based on available
information. Most of the mines do not have an established CFD model, but network modeling
can help estimate tracer gas arrival time, the number of expected peaks, and a reasonable
sampling interval. The possible extent of the damage is also important to know to quickly
determine the tracer gas release and sampling location, and the release amount. CFD modeling
124
|
Virginia Tech
|
may not be needed in some simple scenarios. However, CFD modeling can definitely assist in
designing an effective tracer gas test and increasing the confidence of the results.
8.2 Highlights of the research
The methodology developed in this study proved to be successful in remotely
characterizing underground mine ventilation systems both in the laboratory and in the field. It
can be potentially used after unexpected event, such as explosion and roof fall, to collect
information that will help decision makers to manage a mine emergency more effectively and
increase safety for rescue personnel. The application can also be extended to circumstances other
than unexpected events, such as to study the air flow in inaccessible and to understand
complicated ventilation networks, although these are not the focus of this work.
A recommended general CFD modeling procedure was discussed which can be used as a
guideline for a more reliable simulation. The mesh independence study is one of the most
essential procedures that was emphasized. This is because the modeling results can be
misleading if mesh independence is not achieved, especially when modeling multiple gas
species.
Instead of using trial and error, this study used CFD modeling to design effective tracer
tests. The optimized parameters obtained from the models proved to be very useful, and each
field tracer test result was successfully obtained after only one tracer release. This is important
because in emergency situations, underground information needs to be gathered quickly and the
trial and error process is usually not allowed. Tremendous time and resources were saved by
reducing such items as the number of trips to the mine and re-deployment of the tracer.
Finally, preliminary guidelines and recommendations for use of tracer gas were provided.
Because tracer gas experiments are time and resource consuming in underground mines, the
provided guidelines and recommendations can be used by other researchers and industries to
design tracer gas experiments more efficiently.
8.3 Limitations and future work
The first step of the developed methodology is to estimate the level of damage and
possible ventilation scenarios. This plays an important role in successfully identifying the actual
ventilation scenario. This is because if ventilation damage occurs in a manner other than the
assumed scenarios, the methodology may fail to identify it. Although this is one limitation of the
125
|
Virginia Tech
|
methodology and there could be many possible different scenarios after an unexpected event, it
is usually possible to narrow down the possibilities that are of the most concern. For example,
certain booster fan being on or off may be more important than others to rescue efforts than
others; certain stopping door statuses may be more important than others to estimate the extent of
an explosion. Therefore, it is still possible for this methodology to cover major possible
ventilation scenarios with the help of experienced engineers.
The computational time for a large 3D CFD model is considerable. It is almost
impossible to model a large portion or an entire mine. Sometimes 2D CFD models can be used,
but their accuracy is limited compared to 3D models. Ventilation network modeling is more
practical to simulate a full scale mine, but it cannot resolve the details of tracer gas behavior at
the micro scale, such as how a tracer concentration is distributed over entry cross sections.
Therefore, CFD is an important supplemental component of accurate tracer gas modeling and
experimental design at the field scale. A hybrid scheme that combines the benefit of CFD and
network modeling should be investigated in future work. This allows that CFD be used in critical
areas, while most parts of a mine will be modeled using network modeling to save computational
time with equally effective results.
Only one tracer (SF6) was used in this study, however, it has long been realized that
multiple tracer gases can add flexibility to ventilation surveys in many ways. This not only
allows the release of different tracers at different points without increasing the number of
collected gas samples, but also provides more information because the source of each tracer in
one sample can be identified and air flows in different zones can be investigated simultaneously.
Although the advantages are apparent, the use of the multiple tracer technique is still less
common in underground mines. Therefore, the application of multiple tracer gases in the field
using the developed methodology is future application.
126
|
Virginia Tech
|
[40] A. M. Wala, S. Vytla, C. D. Taylor, and G. Huang, βMine face ventilation : a comparison
of CFD results against benchmark experiments for the CFD code validation,β Mining
Engineering, vol. 59, no. 10, pp. 49β55, 2007.
[41] M. Parra, J. Villafruela, F. Castro, and C. Mendez, βNumerical and experimental analysis
of different ventilation systems in deep mines,β Building and Environment, vol. 41, no. 2,
pp. 87β93, Feb. 2006.
[42] A. D. Gosman, βDevelopments in CFD for industrial and environmental applications in
wind engineering,β Journal of Wind Engineering and Industrial Aerodynamics, vol. 81, no.
1β3, pp. 21β39, May 1999.
[43] D. N. SΓΈrensen and P. V. Nielsen, βQuality control of computational fluid dynamics in
indoor environments.,β Indoor air, vol. 13, no. 1, pp. 2β17, Mar. 2003.
[44] J. C. Edwards and C. C. Hwang, βCFD modeling of fire spread along combustibles in a
mine entry,β in SME Annual Meeting, 2006, p. Preprint 06β027.
[45] H. Lomax, T. H. Pulliam, and David W. Zingg, Fundamentals of computational fluid
dynamics. New York: Springer, 2001.
[46] J. M. Berkoe and D. M. Lane, βPutting computational fluid dynamics to work on mining
and metals projects,β in SME Annual Meeting, 2000, p. Preprint 00β115.
[47] S. M. Salim and S. C. Cheah, βWall y + strategy for dealing with wall-bounded turbulent
flows,β in Proceedins of International MultiConference of Engineers and Computer
Scientists, 2009, vol. II.
[48] T. T. Bui, βCFD analysis of nozzle jet plume effects on sonic boom signature,β in 47th
AIAA Aerospace Sciences Meeting, 2009, pp. 1β28.
[49] P. J. Roache, βPerspective: A method for uniform reporting of grid refinement studies,β
Journal of Fluid Engineering, vol. 116, 1994.
[50] Fluent, βUserβs guide.β 2006.
[51] M. Ariff, S. M. Salim, and S. C. Cheah, βWall Y+ approach for dealing with turbulent
flow over a surface mounted cube: Part 1-low reynolds number,β in Seventh International
Conference on CFD in the Minerals and Process Industries, 2009, no. December, pp. 1β6.
[52] M. Ariff, S. M. Salim, and S. C. Cheah, βWall Y+ approach for dealing with turbulent
flow over a surface mounted cube: Part 2-high reynolds number,β in Seventh International
Conference on CFD in the Minerals and Process Industries, 2009, no. December, pp. 1β6.
[53] R. Bhaskaran and L. Collins, βIntroduction to CFD Basics,β Direct. pp. 1β17, 2005.
130
|
Virginia Tech
|
[54] P. Wesseling, Principles of computational fluid dynamics. New York: Apringer, 2001.
[55] W. L. Oberkampf and C. J. Roy, Verification and Validation in Scientific Computing.
Cambridge University Press, 2010, p. 784.
[56] R. K. Jade and B. S. Sastry, βAn experimental and numerical study of two-way splits and
junctions in mine airways,β in 12th U.S./North American Mine Ventilation Symposium,
2008, pp. 293β298.
[57] F. F. Peng and Y. K. Xia, βAnalysis of a dense-medium separator for course coal
separation using computational fluid dynamics,β Minerals and Metallurgical Processing,
vol. 24, no. 1, pp. 1β12, 2007.
[58] F. F. Peng and Y. Xia, βFluid dynamic modeling of fine particle separation in hindered-
settling bed separators by CFD,β in SME Annual Meeting, 2004, p. Preprint 04β033.
[59] J. ToraΓ±o, S. Torno, M. Menendez, M. Gent, and J. Velasco, βModels of methane
behaviour in auxiliary ventilation of underground coal mining,β International Journal of
Coal Geology, vol. 80, no. 1, pp. 35β43, Oct. 2009.
[60] A. M. Wala, J. C. Yingling, J. Zhang, and R. Ray, βValidation study of computational
fluid dynamics as a tool for mine ventilation design,β in Proceedings of the 6th
International Mine Ventilation Congress, 1997.
[61] J. B. Dick, βMeasurement of ventilation using tracer gas,β Heating, Piping and Air
Conditioning, vol. 22, no. 5, pp. 131β137, 1950.
[62] R. J. Timko and E. D. Thimons, βSulfur hexafluoride as a mine ventilation research tool-
Recent field applications,β U.S. Bureau of Mines, Washington, DC (USA), 1982.
[63] R. B. Krog, S. J. Schatzel, and H. N. Dougherty, βAirflow distribution patterns at a
longwall mine depicted by CFD analysis and calibrated by a tracer gas field study,β in
SME Annual Meeting, 2011, p. Preprint 11β067.
[64] I. M. Konduri, M. J. McPherson, and E. Topuz, βExperimental and numerical modeling of
jet fans for auxiliary ventilation in mines,β in Proceedings of the 6th International Mine
Ventilation Symposium, 1997, pp. 505β510.
[65] F. H. Post and T. van Walsum, βFluid flow visualization,β Focus on Scientific
Visualization, vol. 40, pp. 1β37, 1993.
[66] A. Wala, D. Turner, and J. Jacob, βExperimental study of mine face ventilation system for
validation of numerical models,β in Proceeding of the 9th North American U.S. Mine
Ventilation Symposium, 2002, pp. 191β196.
131
|
Virginia Tech
|
[91] C. Γ. Karacan, T. Ren, and R. Balusu, βAdvances in grid-based numerical modeling
techniques for improving gas management in coal mines,β in 12th US/North American
Mine Ventilation Symposium, 2006.
[92] G. Esterhuizen and C. Karacan, βA methodology for determining gob permeability
distributions and its application to reservoir modeling of coal mine longwalls,β in SME
Annual Meeting, 2007.
[93] T. X. Ren, R. Balusu, and P. Humphries, βDevelopment of innovative goaf inertisation
practices to improve coal mine safety,β in Coal Operatorsβ Conference, 2005, pp. 315β
322.
[94] R. Mossad, A. Vella, and R. Balusu, βInertisation of highwall mining to control methane
concentrations at the Moura Mine,β in Seventh International Conference on CFD in the
Minerals and Process Industries, 2009, no. December, pp. 1β6.
[95] R. Balusu, P. Humpries, P. Harrington, M. Wendt, and S. Xue, βOptimum inertisation
strategies,β in Queensland Mining Industry Health and Safety Conference, 2002, pp. 133β
144.
[96] M. A. Trevits, L. Yuan, M. Thibou, and G. Hatch, βUse of CFD modeling to study inert
gas injection into a sealed mine area,β in SME Annual Meeting, 2010, p. Preprint 10β207.
[97] A. Vella, βVentilation of highwall mining to control methane concentration at the Moura
Mine,β The University of Southern Queensland, 2006.
[98] J. Van Heerden and P. Sullivan, βThe application of CFD for evaluation of dust
suppression and auxiliary ventilating systems used with continuous miners,β in
Proceedings of the 6th US Mine Ventilation Symposium, 1993, pp. 293β297.
[99] R. B. Srinivase, E. Y. Baafi, N. I. Aziz, and R. N. Singh, βThree dimensional numerical
modelling of air velocities and dust control techniques in a longwall face,β in Proceeding
of the 6th U.S. Mine Ventilation Symposium, 1993.
[100] T. Ren and R. Balusu, βThe use of CDF modelling as a tool for solving mining health and
safety problems,β in Coal Operatorsβ Conference, 2010, pp. 339β349.
[101] D. M. Hargreaves and I. S. Lowndes, βThe evaluation of the design and operation of
alternative auxiliary ventilation layouts for use with continuous miner systems in rapid
development drivages,β Final Report 7220-AC-857, Commission of the European
Communities, Directorate-General Energy & Transport, 2001.
[102] S. A. Silvester, I. S. Lowndes, and D. M. Hargreaves, βA computational study of
particulate emissions from an open pit quarry under neutral atmospheric conditions,β
Atmospheric Environment, vol. 43, no. 40, pp. 6415β6424, Dec. 2009.
134
|
Virginia Tech
|
[103] J. Lichter, A. V Potapov, and R. Peaker, βThe use of computational fluid dynamics and
discrete element modeling to understand the effect of cell size and inflow rate on flotation
bank retention time distribution and mechanism performance,β in Proceedings 39th AGM
of Can. Min. Proc., 2009, pp. 473β496.
[104] F. D. Ardejani, E. Baafi, K. S. Panahi, R. N. Singh, and B. J. Shokri, βApplication of
Computational Fluid Dynamics ( CFD ) for simulation of acid mine drainage generation
and subsequent pollutants transportation through groundwater flow systems and rivers,β in
Computational Fluid Dynamics Technologies and Applications, 2011, pp. 123β160.
[105] A. Mazzoldi, T. Hill, and J. Colls, βA Consideration of the jet-mixing effect when
modelling CO2 emissions from high pressure CO2 transportation facilities,β Energy
Procedia, vol. 1, no. 1, pp. 1571β1578, Feb. 2009.
[106] C. Dixon, O. Heynes, and M. Hasson, βAssessing the hazards associated with release and
dispersion of liquid carbon dioxide on offshore platforms.β MMI Engineering Ltd.,
Warrington, UK, 2008.
[107] G. Arpa, A. Widiatmojo, N. P. Widodo, and K. Sasaki, βTracer gas measurement and
simulation of turbulent diffusion in mine ventilation airways,β Journal of Coal Science
and Engineering (China), vol. 14, no. 4, pp. 523β529, Nov. 2008.
[108] R. Suglo and S. Frimpong, βAccuracy of tracer gas surveys in auxiliary ventilation
systems in mines,β in The 9th U.S./North American Mine Ventilation Symposium, 2002.
[109] E. D. Thimons and F. N. Kissell, βTracer gas as an aid in mine ventilation analysis,β
Bureau of Mines Report of Investigations, 1974.
[110] M. G. Grenier, S. G. Hardcastle, G. Kunchur, and K. Butler, βThe use of tracer gases to
determine dust dispersion patterns and ventilation parameters in a mineral processing
plant,β American Industrial Hygiene Association Journal, vol. 53, no. 6, pp. 387β394,
1992.
[111] D. a Kirchgessner, S. D. Piccot, and S. S. Masemore, βAn improved inventory of methane
emissions from coal mining in the United States.,β Journal of the Air & Waste
Management Association (1995), vol. 50, no. 11, pp. 1904β19, Nov. 2000.
[112] Z. Zhang, X. Chen, S. Mazumdar, T. Zhang, and Q. Chen, βExperimental and numerical
investigation of airflow and contaminant transport in an airliner cabin mockup,β Building
and Environment, vol. 44, no. 1, pp. 85β94, Jan. 2009.
[113] Z. Zhang and Q. Chen, βExperimental measurements and numerical simulations of
particle transport and distribution in ventilated rooms,β Atmospheric Environment, vol. 40,
no. 18, pp. 3396β3408, Jun. 2006.
135
|
Virginia Tech
|
Approaches to Simulation of an Underground Longwall Mine and Implications for
Ventilation System Analysis
Hongbin Zhang
ABSTRACT
Carefully engineered mine ventilation is critical to the safe operation of underground
longwall mines. Currently, there are several options for simulation of mine ventilation.
This research was conducted to rapidly simulate an underground longwall mine,
especially for the use of tracer gas in an emergency situation. In an emergency situation,
limited information about the state of mine ventilation system is known, and it is difficult
to make informed decisions about safety of the mine for rescue personnel. With careful
planning, tracer gases can be used to remotely ascertain changes in the ventilation
system. In the meantime, simulation of the tracer gas can be conducted to understand the
airflow behavior for improvements during normal operation.
Better informed decisions can be made with the help of both tracer gas technique and
different modeling approaches. This research was made up of two main parts. One was a
field study conducted in an underground longwall mine in the western U.S. The other one
was a simulation of the underground longwall mine with different approaches, such as
network modeling and Computational Fluid Dynamics (CFD) models. Networking
modeling is the most prevalent modeling technique in the mining industry. However, a
gob area, which is a void zone filled with broken rocks after the longwall mining, cannot
be simulated in an accurate way with networking modeling. CFD is a powerful tool for
modeling different kinds of flows under various situations. However, it requires a
significant time investment for the expert user as well as considerable computing power.
To take advantage of both network modeling and CFD, the hybrid approach, which is a
combination of network modeling and CFD was established. Since tracer gas was
released and collected in the field study, the tracer gas concentration profile was
separately simulated in network modeling, CFD model, and hybrid model in this study.
The simulated results of airflow and tracer gas flow were analyzed and compared with
the experimental results from the field study.
|
Virginia Tech
|
Two commercial network modeling software packages were analyzed in this study. One
of the network modeling software also has the capability to couple with CFD. A two-
dimensional (2D) CFD model without gob was built to first analyze the accuracy of CFD.
More 2D CFD models with gob were generated to determine how much detail was
necessary for the gob model. Several three-dimensional (3D) CFD models with gob were
then created. A mesh independence study and a sensitivity study for the porosity and
permeability values were created to determine the optimal mesh size, porosity and
permeability values for the 3D CFD model, and steady-state simulation and transient
simulations were conducted in the 3D CFD models. In the steady-state simulation, a
comparison was made between the 3D CFD models with and without taking the
diffusivity of SF in air into account.
6
Finally, the different simulation techniques were compared to measured field data, and
assessed to determine if the hybrid approach was considerably simpler, while also
providing results superior to a simple network model.
iii
|
Virginia Tech
|
In addition, I appreciate the help and suggestions from my group members: Dr. Guang
Xu and Dr. Edmund Jong. Dr. Xu introduced Ansys Icem to me and Edmund Jong taught
me the injection techniques of GC.
I would like to thank Dr. Steven Schafrik for teaching me the knowledge of high
performance computer and Linux systems.
Besides, I appreciate the generous help provided by Stephen Theron and Tyler Smith
from PADT Inc. Both of them helped me on the hybrid model.
Additionally, I want to thank the National Institute for Occupational Safety and Health
for providing me the opportunity to conduct this research.
Finally, I want to thank my family, especially, my wife, Ting Du, for her constant love,
support, and trust. She did take a good care of everything at home and it allowed me to
focus all of myself on my work. I owe half of my success to her. Even though my parents
were in China, they are always my patrons.
This publication was developed under Contract No. 200-2009-31933, awarded by the
National Institute for Occupational Safety and Health (NIOSH). The findings and
conclusions in this report are those of the authors and do not reflect the official policies of
the Department of Health and Human Services; nor does mention of trade names,
commercial practices, or organizations imply endorsement by the U.S. Government.
v
|
Virginia Tech
|
1 Introduction
This study is important because it stands to contribute to more rapid and accurate
modeling of mine systems. Computational Fluid Dynamics (CFD) modeling, network
modeling, and a combination of CFD and network modeling were analyzed in this thesis.
This thesis is made up of five chapters. Chapter 1 gives an overview of the thesis.
Chapter 2 reviews the literature of CFD, network modeling, and tracer gas applications.
Chapter 3 and 4 consist of two manuscripts. The paper in Chapter 3 was published in the
SME annual meeting in 2014. The paper in Chapter 4 is planned for publication in the
2015 Mine Ventilation Symposium. Conclusions for the work are in Chapter 5.
CFD modeling has become more and more popular in mining recently. It is critical in
areas where resolving flow patterns is important. Flow patterns can change considerably
after an accident, like explosion that damages the ventilation infrastructure. However, by
using CFD, the flow patterns and distributions in both accessible and inaccessible areas
can be visualized. Additionally, CFD is useful understanding airflow behavior in low
velocity areas not well represented by network modeling. However, CFD is not an easy
technique you can learn in a short period. It requires understanding of both mathematics
and fluid dynamics, and some familiarity with complex software. Another limitation of
CFD is that it requires computing power, as well as expert knowledge, and it is time
consuming for the user to build the numerical model.
Network modeling has been the most prevalent technique for simulation of mine
ventilation systems applied in the mining industry for many years. Compared with CFD
modeling, network modeling is much easier to learn and does not require as much
computing power. There are two key software packages used in this study. One is
VnetPC, developed by Mine Ventilation Services, which is based on the Square Law and
the Hardy Cross process. The other one is Flownex, developed by M-Tech Industrial
(Pty) Ltd., which is based on partial differential equations of mass conservation,
momentum conservation, and energy conservation. VnetPC is discussed in detail in
Chapter 3 and Flownex is analyzed in Chapter 4.
1
|
Virginia Tech
|
A combination of CFD and network modeling was established to save computing time
and ensure high accuracy results at the same time. Flownex was used as the network
component and Ansys Fluent was used as the CFD component in the combination. The
integration capability of Flownex makes it possible to connect a one-dimensional (1D)
network model with a 3D CFD component. Details about this hybrid model are shown in
Chapter 4.
The field experiment in this study was conducted in an underground coal mine in the
western U.S. Four students from Virginia Tech and four workers from the mine
conducted the experiment. Detailed field study information can be found in Jongβs
dissertation (Jong 2013). The gob in the underground mine was designed for airflow to
go around it instead of flowing through it. Tracer gas technique was utilized to ascertain
airflow information in a two phase field study. Sulfur hexafluoride (SF ) was selected as
6
the tracer gas. With the help of GC and ventilation survey data, the volumetric flow rates
of SF at different sample points were obtained.
6
2
|
Virginia Tech
|
2 Literature Review
2.1 CFD
CFD is now widely applied in the mining industry. Due to the high level of accuracy and
flexibility, CFD has been successfully applied in spontaneous combustion control, dust
control, methane movement simulation, fire spread simulation, ventilation airflow
simulation, and gob inertisation, just to give a few examples in mine ventilation. At the
same time, many studies have been published on CFD applications in other areas of
mining. However, few studies have been done on a hybrid CFD-network model. Since
the purpose of this study is to figure out a hybrid CFD-networking model for the
simulation of an underground longwall mine, CFD applications in longwall gob areas and
other fields are briefly reviewed. An exhaustive literature of CFD applications in mining
can be found in Xuβs dissertation (Xu, 2013).
CFD is powerful for analyzing flow patterns and solving fluid dynamics problems,
especially in the areas people cannot access, like a gob. A gob area, which is created as a
longwall advances, is complex in terms of geometry and quantifying the properties of
porous media, as it is subject to dynamic geomechanical conditions. Since it is made up
of broken rocks falling from the roof, the gob is also not accessible. In addition,
ventilation surveys and experiments around the gob are also difficult to conduct. CFD
makes it possible to visualize the flow in the gob. Many CFD studies have been done on
gob areas in the past. Some of the studies are reviewed here to show the advantages of
using CFD in the gob areas.
CFD plays a significant role for realizing the changes of flow patterns and behavior in the
gob. Yuan and others (Yuan, Smith, and Brune 2006) successfully analyzed the airflow
patterns in a gob under various ventilation systems by using CFD. The gob in one panel
was divided into five zones with constant permeability values in the study. This
conclusion was based on Fast Lagrangian Analysis of Continua (FLAC) modeling of
longwall mining used in their paper. Ren and others (Ren, Balusu, and Claassen 2011)
published another paper with a CFD study and they found that the gas flow pattern in the
3
|
Virginia Tech
|
gob was related with the retreat direction of longwall. Karacan and others (Karacan, Ren,
and Balusu 2008) summarized the techniques of numerical modeling used in the mining
industry and issues faced for gas management in another study, discussing CFD and
numerical reservoir modeling. They noted that CFD was better but more complicated
than the network models. Besides, CFD was more effective in the areas where network
models were not suitable and capable.
CFD is also helpful on figuring out the relationship between permeability and gob
properties. Karacan and Esterhuizen (Esterhuizen and Karacan 2007) found that
permeability in the vertical direction had a relationship with a caving and block rotation
model. Both a CFD model and a FLAC3D model were used in the study, linking
geomechanics with changes in the porous media. Another CFD study on the spontaneous
heating in the gob was conducted (Yuan and Smith 2008a) to determine the possibility of
having spontaneous heating in the gob areas. Yuan and Smith conducted another CFD
study in 2008 (Yuan and Smith 2008b). The purpose of this study was to learn how the
gob characteristics will change spontaneous heating in the gob. Oxidation of coal was
treated as spontaneous heating in the model. They concluded that permeability and
inducing time had an inverse relationship. A bleederless ventilation system and nitrogen
injection in the gob was also successfully studied in a similar CFD study (Smith and
Yuan 2008).
In addition, CFD models are used to develop optimum gob inertisation strategies to
improve coal mine safety. Gob inertisation can help decrease the chance of potential
explosion during longwall sealing operations. Ren and others demonstrated with CFD
that gob inertisation can be achieved within a few hours of sealing a panel (Ren, Balusu,
and Humphries 2005). Proactive inertisation strategies were also developed to suppress
spontaneous heating in the gob. Results showed that the inertisation was more effective at
two hundred meters behind the face than that right behind the face line. In another study,
Ren and Balusu (Ren and Balusu 2005) developed gob inertisation strategies by using
CFD models. The purpose of their study is to understand flow migration dynamics in the
gob. One of the key techniques used in the study was linking User Defined Functions
4
|
Virginia Tech
|
(UDFs) to Fluent solver. With the help of the UDFs, a momentum sink was added to the
momentum equations to model the airflow through the gob.
CFD can be used to improve the control of spontaneous combustion in longwall gobs
(Ren and Balusu 2005). Gob permeability distributions and gas emissions are the most
important part of the CFD models. With the help of pressure, flow rate, and gas
distribution in a longwall gob area, initial models were calibrated and gob permeability
distribution was refined. Due to CFD modeling work, innovative gob gas control
strategies for spontaneous combustion have developed rapidly and efficiently. There are
several major factors affecting spontaneous heating and CO production underground,
such as, seam structure, condition of gateroads behind the working face, caving pattern
behind the face, location of faults, and length of the back return. In addition, CFD
modeling was used to study the gas flow mechanics and distribution in longwall gobs.
CFD models have been useful to develop control strategies for gas and spontaneous
heating, such as reducing air velocity and increasing gob drainage flow rate (Ren and
Balusu 2005).
Dust control is another significant application of CFD. Dust is generated during the
excavation process and it is a major concern in terms of underground minersβ health and
safety. CFD is a very effective tool to evaluate and improve dust suppression in
continuous miner and roadheader sections as described by Heerden and Sullican in their
paper (Heerden and Sullivan 1993). After the process of establishing correct geometry,
defining properties of fluids and boundary conditions, velocity vectors were plotted and
velocity contours were made. Respirable dust particles were assumed to follow the gas
flow in the underground and flow lines were used to quantify the movement of dust
particles. The model was used to investigate dust suppression with various parameters,
such as the positions of a force ventilation column, a brattice, and an exhaust column,
drum rotation, water sprays and air movers. In addition, methane concentration and
emission rates were simulated through CFD.
5
|
Virginia Tech
|
CFD is a flexible modeling technique that can also evaluate methane movement in mines.
During the production, methane will be released from the working face and enter the gob
areas. CFD can help define the flow distribution in geometrically complex conditions.
Also, CFD makes it possible to test the effects of modifications under the same ambient
conditions. Hence, the control of methane emissions system can be optimized. In
Kelseyβs paper (Kelsey et al. 2003), CFD simulations of methane drainage were
performed. The quantity of methane removed reflected the effect of drainage. In addition,
methane drainage effect on flow through strata was visualized. Based on CFD and
information from geotechnical modeling, numbers and spacing of methane drainage
boreholes were examined and drainage was optimized.
CFD can be applied to simulate fire spread along combustibles in the underground, a
serious safety issue in underground mines. Edwards and Hwang (Edwards and Hwang
2006) developed a CFD fire spread model to control fire spread and reduce CO and
smoke emissions. Flame spread rate was examined for the ribs and roof of a coal mine
entry, timer sets, and a conveyor belt. Char forming materials with thermal properties
were also modeled in the CFD program. Fire Dynamics Simulator (FDS), which is based
on CFD, was applied to simulate the fire spread in a mine entry. Navier-Stokes equations
are solved numerically in the simulator. Basically, fame spread has a relationship with
pyrolysis gases emission and flame front is defined by the leading edge of the fuel
surface at the pyrolysis temperature. Afterwards, the flame propagation is determined by
temporal movement of the pyrolysis temperature. In the paper (Edwards and Hwang
2006), the model was used to simulate the 1990 fire at Mathies Coal Mine. The coal lined
tunnel flame spread rate was analyzed and it was turned out to be insensitive to the heat
of pyrolysis, but significantly sensitive to the coal moisture content. Moreover,
predictions made by CFD model of the dependence of flame spread along conveyor belt
on air speed met the results obtained by Lazzara and Perzak (Lazzara and Perzak 1987).
Overall, CFD is a flexible, grid-based numerical technique. It has been successfully
applied in various fields of mine ventilation improve underground mine safety. CFD
makes it possible to conduct all kinds of simulations, even in the geometrically complex
6
|
Virginia Tech
|
conditions, like the gob areas. However, CFD is rarely combined with other software to
produce quicker and highly accurate results, which is why this work examines a hybrid
approach.
2.2 Network Modeling
Network modeling software used in this paper are VnetPC and Flownex. Since VnetPC
was already reviewed in (Zhang et al. 2014), Flownex is mainly discussed in this part.
Flownex is a thermodynamic modeling software. It uses both steady state and transient
simulation to compute temperature, flow rates, and other flow properties. It has been
widely used in fields like fluid system design and optimization, but less widely applied to
mine ventilation. An important reason for using Flownex in this study is that it is capable
of integration with other software, like Ansys Fluent. VnetPC does not have this
integration capability. Since Flownex is not as popular as VnetPC in the mining industry,
several studies are reviewed to give a general idea about its useful features.
Flownex is able to perform both steady state simulation and transient simulation for
complex geometries. Slabbert (Slabbert 2011) used Flownex to simulate a typical Pool
Type Research Reactor. Both steady state and transient simulations were performed in
Flownex to check the capability and accuracy of the Flownex model. Results from the
Flownex model were compared with that from the Engineering Equation Solver (EES).
Moreover, the results from the Flownex model matched that EES very well. Slabbert
concluded that Flownex could get the results very fast and it was a good tool for both
steady state and transient simulations. Flownex was used as one of the two software to
model the performance of transient heat exchanger in another study (Olivier 2005). Two
methods for analyzing network problems were studied. One was the Implicit Pressure
Correction Method (IPCM). The network solver used in this method was Flownex. The
other one was the Runge Kutta method with Trapezoidal Damping (RKTD), which was
an explicit method. Xnet was applied as the network solver. Both the two solvers were
used to simulate the transient heat exchange performance. Results from the Xnet were
7
|
Virginia Tech
|
highly accurate but Xnet required much computational time. In examining the results
from these two solvers, Olivier found that the results from Flownex matched that from
Xnet very well. She concluded that Flownex was an important tool to simulate the
transient heat exchanger and solve complex networks.
Flownex is also powerful due to its coupling capability with external software, such as
WKIND, TINTE, and CFD. Walter and others (Walter, Schulz, and Lohnert 2004)
simulated a Pebble Bed Modular Reactor (PBMR) plant and its power conversion unit
(PCU) with a coupled Flownex-WKIND model. WKIND is a solver for one group
neutron diffusion equation and it is able to simulate one-dimensional neutronics behavior
(Walter, Schulz, and Lohnert 2004). The integration capability of Flownex was enabled
with a memory map file provided by the Windows application programing interface
(API). Results such as, pressure drop, inlet temperature, and mass flow rate were
computed in the Flownex and transferred to WKIND as the inputs. The coupled
simulation stopped when both the Flownex and WKIND model reached their steady-state
solutions. In addition, Marais (Marais 2007) developed a new method, which was a
combination of TINTE (TIme dependant Neutronics and TEmperatures) and Flownex, to
validate the point kinetic neutronic model of the PBMR. TINTE is a solver for a two-
dimensional neutron diffusion equation and it is able to solve neutronic models (Gerwin,
Scherer, and Teuchert 1989). The coupling feature of Flownex played an important role
in the method. Results from the hybrid model showed that rough boundary conditions
could be obtained thought the indirect coupling method.
According to aforementioned studies, Flownex is good at steady state and transient
simulations. The integration capability of Flownex makes it possible to solve complex
networks with an external software.
2.3 Hybrid Modeling
As abovementioned, Flownex has the capability to be combined with an external
software. Since the hybrid approach in this study was completed based on both Flownex
and CFD, the combination of Flownex and CFD is individually reviewed in this section.
8
|
Virginia Tech
|
Several hybrid Flownex-CFD models were successfully made with various purposes in
the past.
A study on the hybrid Flownex-CFD model for simulating the flow distribution in the
PBMR was investigated by Huang (Huang 2008). The hybrid model was created to save
the computing time by take advantages of both Flownex and CFD. Flow paths in the
reactor was simulated in the 3D CFD model. The core structure of the PBMR was
simulated in the Flownex model. Flownex created a complex network structure by
combining all the flow paths. Results from these two models had a good agreement on a
global scale. Huang concluded that the hybrid model made it quick and accurate to get
the results by taking advantages of both the Flownex and CFD models. Based on Huangβs
work, a similar study for PBMR was conducted by Janse Van Rensburg and Kleingeld
(Janse Van Rensburg and Kleingeld 2010). Results from the Flownex model were passed
to the CFD model as boundary conditions. With the help of the hybrid method, leakage
flows were successfully identified and ranked in a High Temperature Reactor (HTR),
which was a predictor for bypass flows and leakages in a rector. Reasons of the bypass
flows and its effects on temperatures of the fuel and component were analyzed in the
paper. Later on, Janse Van Rensburg and Kleingeld (Janse Van Rensburg & Kleingeld,
2011) did another study on the leakage and bypass flows in an HTR by using the same
hybrid approach. They found that changes of the pressure drop in a pebble bed did not
lead to same changes in the leakage flows. Janse Van Rensburg and Kleingeld have done
many studies on the leakage and bypass flows in a HTR and the rest were not reviewed in
this thesis due to the limited space.
Gouws (Gouws, 2007) conducted a similar Flownex-CFD study to find possible
geometrical changes for the dome of a combustion chamber. The reason for making the
changes was that there were crack formation on the dome. Flownex was used to simulate
the combustion process while CFD was utilized to figure out the changes to the dome
geometry. Pressure losses, temperature and flow distributions were obtained from the
Flownex model. These data were applied as the boundary conditions in the CFD model to
simulate the flow distribution inside the combustor. According to the hybrid approach, it
9
|
Virginia Tech
|
This paper was presented at the 2014 SME annual meeting in Salt Lake City, and is
included in the meeting preprints (Feb. 23-26, Salt Lake City, UT, Preprint 14-148).
Hongbin Zhang conducted all the CFD and the network modeling work and wrote the
paper with technical and editorial input from coauthors: Hassan El-Hady Fayed, Dr.
Kray D. Luxbacher, Edmund Jong, and Dr. Saad Ragab. Please cite this article as:
Zhang, H., Fayed, H. E.-H., Luxbacher, K. D., & Ragab, S. (2014). The feasibility of
hybrid network and CFD modeling for mine ventilation applications. 2014 SME Annual
Meeting (Preprint 14-148). Salt Lake City, UT (USA).
3 The Feasibility of Hybrid Network and CFD
Modeling for Mine Ventilation Applications
3.1 Abstract
This paper examines the feasibility of combining network modeling and computational
fluid dynamics for modeling of underground mine ventilation systems. Both simulation
methods have specific advantages and disadvantages for analysis of mine systems.
Network modeling is widely utilized by many operations and allows for assessment of
current systems and simulation for planning purposes. Alternatively, CFD has been
utilized only marginally by operations and is typically a research tool, requiring
considerable computational power, complex models, and careful analysis and calibration
of results. Network modeling allows for a holistic approach to the ventilation system,
giving quantity, velocity and pressure in every branch, but CFD modeling can resolve the
flow regime in 2D or 3D, which is idea when examination of an area on a more detailed
basis is useful, such as dust and gas control. Integration of the two can allow for more
flexible systems analysis. The feasibility of integration, along with application to limited
underground mine data are examined.
3.2 Introduction
Mine ventilation plays a significant role in underground mining by not allowing
contaminated air to enter the working face. Generally, mining companies use network
modeling to simulate the ventilation system for analysis and planning purposes.
However, network modeling is not detailed enough to simulate complex airflow, nor does
it resolve flow patterns in a given cross section. Computational Fluid Dynamics (CFD)
11
|
Virginia Tech
|
can be applied in this case to assist the network model. In this study, VnetPC was used as
the network modeling tool and compared to a two-dimensional CFD model of an
underground coal mine in the western US. The goal of this study was to analyze the
feasibility of hybrid network and CFD modeling for mine ventilation applications.
3.3 Background
3.3.1 CFD Applications in Mining
Computational fluid dynamics (CFD), a grid-based numerical technique, is widely used
in mine ventilation. Due to the high level of accuracy and flexibility, CFD is applied in
spontaneous combustion control, dust control, gob inertisation, methane movement
simulation, fire spread simulation, and ventilation airflow simulation. CFD was used to
improve the control of spontaneous combustion in longwalls (Ren and Balusu, 2005).
Innovative gob gas control strategies for spontaneous combustion have been developed
by Ren and Balusu using CFD. In addition, Ren and Balusu used CFD modeling to study
the gas flow mechanics and distribution in longwall gobs. These models were used to
develop control strategies for gas and spontaneous heating, such as ensuring gas quality
in gob degasification systems, reducing air velocity and increasing gob drainage flow
rate, and immediate sealing of active gob.
CFD is a flexible modeling technique that can also evaluate methane movement in mines.
During production, methane is released from the working face and enters gob areas. CFD
can help characterize the flow distribution in geometrically complex conditions, and also
makes it possible to test the effects of modifications under varying conditions, allowing
for optimization of methane control systems. For example in Kelseyβs work, CFD
simulations of methane drainage were performed, the quantity of methane removed
reflected the effects of drainage, and methane drainage effects on flow through strata
were visualized (Kelsey et al., 2003). Based on CFD and information from geotechnical
modeling, numbers and spacing of methane drainage boreholes were examined and
drainage was optimized.
12
|
Virginia Tech
|
CFD is especially useful for the simulation of underground ventilation airflow.
Aminossadati and Hooman built a two-dimensional CFD model to simulate the airflow
behavior in underground crosscut regions (Aminossadati and Hooman, 2008). Airflow
was directed into these regions by using brattice sails and a CFD model examined the
effects of brattice length on airflow behavior. This CFD model made it possible to
determine the optimum size of brattice sails. CFD is a useful numeric modeling tool that
has been successfully applied in various fields of mine ventilation to ensure miner safety.
3.3.2 Network Modeling Applications in Mine Ventilation
VnetPC is a popular network ventilation simulation program designed to help the mine
ventilation workers monitor and design underground ventilation layouts. Based on the
data obtained from ventilation surveys and airway dimensions, the program is able to
provide various ventilation parameters such as air quantity, air velocity, airway
resistance, and pressure drops using computations based on Kirchhoffβs Laws and (Mine
Ventilation Services, 2013) the Hardy Cross iterative technique.
There are advantages and disadvantages to using network modeling. On one hand, it is
relatively simple to build a network model, which is why it is so popular with most
mining companies. Wallace and others (Wallace et al., 1990) and Banik (Banik et al.,
1995) determined that network modeling was capable to simulate the longwall gob
leakage. Besides, Mcpherson (McPherson, 1988) used network modeling to analyze
ventilation networks in a block caving mine in Chile. Moreover, network modeling was
successfully applied to evaluate the ventilation efficiency and cost by using auxiliary fans
in coal mines (Wallace et al., 1990).
On the other hand, network modeling results do not characterize or visualize ventilation
properties on a small scale; for example, such as velocity, pressure drop, and volume
flow rate distribution in the cross-sectional area of a mine entry. This can be limiting in
certain applications, including release and monitoring points for tracer gases, where
13
|
Virginia Tech
|
smaller scale behavior can affect flow and mixing. In addition, because of one-
dimensional and branched nature, it is difficult for network modeling to run simulation in
the inaccessible regions, like gobs (Karacan et al., 2008).
3.4 Detailed Field Study
This preliminary examination of hybrid CFD and network modeling simulation is part of
a larger field study, examining characterization of a longwall ventilation system. A CFD
model was developed for comparison to the network model. The purpose is to build an
efficient, accurate model for detailed ventilation studies. The real world data was
collected in an underground coal mine by using a multiple tracer gas technique. Sulphur
hexafluoride (SF ) and perfluoromethylcyclohexane (PMCH) were chosen as the tracer
6
gases in this study. SF and PMCH were released at release points (RP) 1 and 2
6
separately in the underground coal mine as shown in Figure 1. Locations of five sample
points are also shown in the figure. With the help of tracer gas technique and a ventilation
survey, air velocity and volume flow rate for the site are known.
3.5 Two-Dimensional CFD Model Setup
3.5.1 Geometry
The mine geometry was imported directly from the mine map into a meshing program.
The original geometry for the mine was smoothed. Since characterization of the gob area
is complex, the widths of four air paths around the gob were decreased to half of the
original width to match the real world situation. The simplified geometry and dimensions
for the geometry was shown in Figure 1. Sample points locations were also shown in
Figure 1.
14
|
Virginia Tech
|
Figure 1. Plan view of the model mine. (Note: this figure has been changed for the
consistency of labeling throughout the thesis.)
3.5.2 Hypothesis
Although CFD can simulate the real situation, some assumptions should be made to
simplify the CFD model for grid generation and make CFD simulation feasible in terms
of memory and CPU time. Several simplifications were made. There is no leakage
between air passages in the model mine. Leakage can significantly influence ventilation
in an underground coal mine and this model will be updated in the future. In addition, Air
flow is incompressible and fully turbulent. The gravity influence on mine air is neglected.
Besides, the gob area is treated as solid wall in the model. Moreover, there is no heat
transfer during the procedure and the wall and air temperatures are constant.
3.5.3 Governing Equations
In general, the continuity equation, momentum equation, and energy equation are the
three governing equations used in CFD. Since the fluid in the model was assumed to be
incompressible and there is no heat transfer, the energy equation was not applied in this
model. Reynolds Averaged Navier-Stokes (RANS) equations (Ansys, 2009), containing
continuity equations and momentum equations, were used in the CFD model. The
standard k-epsilon turbulence model was used to compute eddy viscosity and Reynolds
stress term.
Continuity equation:
15
|
Virginia Tech
|
ββVββ = 0 (3-1)
where β is the mean velocity vector.
Momentum equation:
βP T
β(Vββ βVββ ) = β +ββ(ΞΌ (βVββ +(βVββ ) )
Ο eff (3-2)
where = ΞΌ+ , Β΅ is the molecular viscosity of air, is the eddy viscosity and it is
π‘ π‘
computed from k-Ι model, is the pressure gradient.
3.5.4 Meshing
A two-dimensional geometry and a quadrilateral, structured mesh were created. Node
density was designed to be high close to the walls and low in the middle as shown in
Figure 3. The reason was that airflow velocity close to the wall changes considerably and
it is necessary to have a good resolution in this region. Since the model was complex, the
whole geometry was divided into two parts before meshing and combined using an
interface boundary condition. The mesh for the entire model mine and for a small part of
the mine was presented in Figure 2 and Figure 3, respectively.
Figure 2. Meshing picture for the model mine (plan view).
16
|
Virginia Tech
|
Figure 3. Mesh distribution in a small part of the model (plan view).
There were three intakes and three returns in this mine model as shown in Figure 1. All
the intake air were defined as inlets in the CFD model. All the outlets represented the
return air. The neutral air was treated as return 2 in the figure. Three inlets were specified
as velocity inlets. Three outlets were defined as outflow. The field measured airflow
quantities were applied to the outflow boundary condition. At the same time, the
measured airflow velocities were applied to the velocity-inlet boundary condition. Details
are found in Table 1. The remainder of the geometry was treated as no-slip, stationary
walls. In this study, the gob area was also treated as wall boundary condition because of
the fact that only very small amount of air would be able to flow through the whole gob
area, and because this is a preliminary study. The gob area will be treated as porous
media in the next study.
Table 1. Intakes, returns, and neutral boundary conditions (Note: this table has been
changed for the consistency of labeling throughout the thesis.).
Airway Velocity (ft/s) Airway Airflow Quantity Percentage
Intake 1 6.480 Return 1 0.192
Intake 2 2.300 Neutral 2 0.196
Intake 3 0.719 Return 3 0.630
3.5.5 Solution Setup
In this study, the numerical solutions were processed using the Ansys Fluent 14.5. The
type of the solver is defined as pressure-based. Time was steady and 2D space type was
17
|
Virginia Tech
|
planar. Gravity was neglected in this study since it was not expect to significantly affect
the airflow in the underground mine. A standard k-epsilon turbulence model was used to
simulate the airflow. Fluid in the model was defined as air.
3.6 Mesh Independence Study
The mesh independence study was conducted to ensure that mesh size is not influencing
results. The optimum mesh size was defined after this study. Three kinds of mesh; fine
mesh, medium mesh, and coarse mesh, were analyzed. Node numbers of the three mesh
size models were about 800,000, 650,000, and 500,000, respectively. The coarse mesh
was generated first. Then the medium mesh fine mesh were created by increasing the
nodes density on both entry direction and cross-section direction. There were two criteria
to qualify the mesh. One criteria was that a determinant value should be at least 0.3 to be
acceptable for a solver. The other criteria was that the minimum angle value must be
greater than to be acceptable for Fluent (Ansys, 2007). The mesh quality in this study
matched the two criteria.
The finer the mesh, the closer the numerical solution to the exact solution (SΓΈrensen &
Nielsen, 2003). Airflow rate results of the sample points were collected from the three
mesh models as shown in Figure 4. It is apparent that the solution is not changing
significantly, which indicates that the solution is mesh independent. This also means that
the medium mesh is sufficient for a good solution. Then only the results from medium
mesh will be compared with the experimental measurements in the next section.
18
|
Virginia Tech
|
Air Quantity (kcfm) Comparison for Different Mesh
Size Models
100.00
90.00
80.00
)
m
70.00
f
c
k ( 60.00
y
titn 50.00 Fine Mesh
a u 40.00 Medium Mesh
q
r iA 30.00 Coarse Mesh
20.00
10.00
0.00
1 2 3 7 9
Sample points
Figure 4. Air quantity profile of the three mesh size models.
3.7 Results
3.7.1 CFD Results
Air velocity and air quantity were the two main results obtained from CFD simulation.
Because of the complexity of underground geometry, air velocity measured in the
underground did not always match the average velocity simulated. Otherwise, air
quantity, which was not affected by complex situation in the underground, was used as
the indicator to make comparison between simulated results and real results. Velocity
contours were generated to provide a visualization of the three mesh results. Details could
be found in Figure 5. Because of the limited space, only the velocity contour of fine mesh
model was shown in Figure 5.
In this study, air quantity results obtained from experimental measurements were treated
as known values, although error in their measurements could have occurred, they were
measured by experienced ventilation personnel. The sample point locations can be found
in Figure 1. Air quantity results obtained from the medium mesh model were compared
to the experimental measurements. Details are shown in
Table 2. Error from CFD simulation was calculated based on Equation (3-3):
19
|
Virginia Tech
|
Error comparison between CFD model and VnetPC model
100.00
80.00
60.00
40.00
20.00
0.00
SP1 SP2 SP3 SP7 SP9
CFD Error (%) Vnet Error (%)
Figure 7. Error comparison between CFD model and network model.
From Figure 7, there is no doubt that the CFD model is more accurate than the network
model. However, CFD model is not perfect since it still has a 16% error at sample point
1. There are several possible reasons for the error at sample point 1 in CFD model. First,
the CFD model is a two-dimensional model, which is not accurate enough to represent a
volume flow rate. Second, there may be error associated with the experimental
measurement taken at sample point 1. Last, the gob area was treated as solid wall instead
of porous media in the model. If the gob area is defined as porous media, some air will
flow into the gob decreasing the air quantity at sample point 1.
From Figure 5 and Figure 6, the two-dimensional CFD model is more detailed than the
network model. The airflow distribution in the cross-sectional area can be seen clearly in
the CFD model. However, network modeling only uses simple lines and numbers
indicating the airflow. Considerable differences for the airflow quantity can be found in
the two figures. There are two kinds of places in Figure 5 where CFD is especially useful
to analyze the airflow. First, CFD is necessary to resolve where the intake air comes up
through the crosscut. Because when the airflow changes direction, a small part of the
intake air circulates next to the wall. The airflow then becomes disordered and velocity is
22
|
Virginia Tech
|
no longer uniformly distributed in the cross-sectional area. Secondly, CFD is potentially
needed where there is an obstacle in the airway. The open door in Figure 5 is a good
example. The intake air circulates before entering the open door and is divided into
several air groups with various velocity magnitudes after passing through the open door.
It is obvious that a tracer released in these two kinds of places may not behave uniformly.
However, with the help of CFD, the right release location of a tracer can be determined
and the data collected will be more representative and accurate. Network modeling is
more appropriate at a global scale, like the airways around the gob and the intake air at
the bottom in Figure 6. It is also not effective to use network modeling to visualize or
understand airflow at a small scale, such as, passing through an obstacle, like the open
door in Figure 5.
Based on Table 3, network modeling does not represent low velocity flow well. The
maximum error, 98.35%, appears at sample points 5 where the real airflow quantity is
only 17 kcfm. As the airflow quantity increases, the error becomes much smaller.
In conclusion, the CFD model is accurate but requires more expertise, time to develop a
model, and computational power. Network modeling, such as VnetPC, is not as complex
as CFD model and it practical for assessment of ventilation systems in large mines at a
global scale. It is also more appropriate to model the large value branches.
Therefore, it is time-saving to simulate the whole mine by using network modeling. After
comparing results from VnetPC with experimental measurements, CFD model should be
built in certain areas where there is a considerable error in a network model, and where
resolution of complex flow regimes are important. A hybrid model has great potential to
allow for practical modeling of large mines while still looking at detailed flow regimes as
necessary. To make this hybrid network and CFD method more accurate, the gob area
will be defined as porous media in the next study.
23
|
Virginia Tech
|
The following paper will be presented at the 15th North American Mine Ventilation
Symposium in Blacksburg, VA. Hongbin Zhang conducted all the CFD, the network
modeling, and the hybrid model work, and wrote the paper with technical and editorial
input from coauthors: Kareem Akhtar, Dr. Kray D. Luxbacher, Dr. Saad Ragab, Edmund
Jong, and Tyler Smith. Please cite this article as: Zhang, H., Akhtar, K., Luxbacher, K.
D., Ragab, S., Jong, E., & Smith, T. (2015). A comparison of simulation methods for mine
ventilation systems. 2015 Mine Ventilation Symposium. Blacksburg, VA (USA).
4 A Comparison of Simulation Methods for Mine
Ventilation Systems
4.1 Abstract
This paper introduces an approach for simulation of airflow and tracer gas distribution in
an underground longwall mine located in the western U.S. The approach takes advantage
of both computational fluid dynamics (CFD) and network modeling for a mine
ventilation system. Network modeling is popular with mining companies because it is
relatively simple and easily updated. Network modeling is usually one-dimensional (1D)
while CFD can solve both two-dimensional (2D) and three-dimensional (3D) domains. A
hybrid model of CFD and network modeling was developed in this paper to demonstrate
the approach. Furthermore, a network model, a 2D CFD model, and a 3D CFD model
were conducted separately. The gob area was simulated as porous media in both the 2D
and 3D CFD models. Because there were no accurate porosity and permeability data
provided for the gob, a sensitivity study on the porosity and permeability data was
created in the 3D CFD model to eliminate the effects from these data. Other than the
modeling work, a field study was conducted in the mine and the results from the field
study were considered as right results when compared with the results from the three
models. Tracer gas technique was used in the field study. One reason for releasing the
tracer gas into the underground was that airflow information in complex ventilation
situations could be quickly and remotely obtained, especially for an emergency, like an
explosion. Besides, with the help of tracer gas technique, the CFD models could be
created to simulate the areas where the airflow was complex instead of simulating the
whole mine. Simulations of tracer gas were also conducted in the network model, the 3D
CFD model, and the hybrid model. In the 3D CFD model, two kinds of simulations,
which are steady-state simulation and transient simulation, were completed. All the
25
|
Virginia Tech
|
results from these models were analyzed and compared with the results from the
experiment.
4.2 Introduction
The primary purpose of mine ventilation in an underground mine is to provide enough
oxygen to the personnel working in the underground mine, and to dilute methane and
dust concentrations. Network modeling is a popular and important method used in the
mining industry to predict and analyze airflow distribution in the underground system.
However, it is not a good tool for the simulation of a gob area. Furthermore, airflow
distribution in a cross sectional area cannot be visualized with a network model.
Computational fluid dynamics (CFD) is a good fit in these areas as demonstrated in the
preliminary study (Zhang et al. 2014). The main disadvantage of CFD is that it is time-
consuming and requires high computational power. Additionally, CFD is a difficult
technology to acquire and individuals utilizing this time of modeling should have a
background in fluid dynamics and computational methods. A hybrid model was used to
demonstrate advantages of both CFD and network modeling in this paper. A preliminary
2D CFD model of this study was created (Zhang et al. 2014). However, the gob area was
not simulated in the 2D CFD model. In this paper, a three-dimensional (3D) CFD model
was created for an underground coal mine in the western U.S. The gob was simulated as
porous media in the 3D CFD model. A network model was also created for the same
underground mine and Flownex was used as the network modeling software. Moreover, a
hybrid model with a 3D CFD component and a Flownex was also created. Results from
all the three models were compared with the experimental data.
4.3 Detailed Field Study
The field experiment of this study was done in the underground longwall mine in the
western U.S. Four students from Virginia Tech and four workers from the mining
company did the experiment together. The experiment followed a release-collect process.
Two trace gases were released at two release points (RP1 and RP2) and collected at five
sample points (SP1, SP2, SP3, SP7, and SP9). SP1 was located at the entry inby the last
open crosscut, where all the airflow from three inlets join together. SP2 was located at the
26
|
Virginia Tech
|
entry outby in the beltline. SP3 was located next to the gob. SP7 was located at the entry
outside the airlock in the bleeders. SP9 was located at the tailgate entry. Locations for the
release and sample points are shown in Figure 8.
The tracer gases were sulphur hexafluoride (SF ) and perfluoromethylcyclohexane
6
(PMCH). SF and PMCH were released at RP1 and RP2, respectively. Since PMCH was
6
released in the gob, only SF was analyzed in this study. SF was released at a constant
6 6
mass flow rate of 200 standard cubic centimeters per minute (SCCM). Airflow quantity,
SF flow quantity, and SF concentration were then obtained by using trace gas technique
6 6
and conducting a ventilation survey. The results from the field experiment were treated as
the exact results and they were compared with the modeling results in this study.
4.4 CFD Model
4.4.1 Geometry of the Model Mine
An overview of the model mine geometry with three monitor planes and three monitor
lines is shown in Figure 8. It is obvious that the mine geometry was simplified in this
paper compared to the mine geometry in the preliminary study (Zhang et al., 2014).
Pathways with stoppings installed were deleted in this geometry, neglecting leakage.
Comparison between these two geometries can be found in Figure 65 and Figure 66 in
the Appendix. Since it is assumed that there is no leakage between air passages (in
Section 4.4.3), this simplification does not have an effect on the final results. In the
meantime, the simplified geometry saves the time on the grid generation for both the 2D
and 3D CFD models. An enlarged view for the locations of the monitor planes and lines
is shown in Figure 9.
27
|
Virginia Tech
|
3D CFD model were simulated as a stationary wall. RP1 was simulated as a small cube
(shown in Figure 118) in the model. All the surfaces of the cube were assigned interface
boundary conditions, which meant the upstream airflow could flow through the cube and
carry on the SF released from the cube. Volume inside the cube was treated as a flow
6
domain and a source term was added to match the 200 SCCM release rate of the SF . All
6
the intakes in both the models were assigned velocity-inlet boundary conditions. Two
returns and a neutral in both the models were assigned outflow boundary conditions. The
velocity and the airflow quantity percentage values were obtained from the field study in
the underground mine. Details about the boundary conditions were shown in Table 4.
Meshes for both the 2D and 3D CFD models were divided into several parts before being
solved in Ansys Fluent 14.5. Interface boundary conditions were used to combining these
meshes together when they were ready to be solved. Eight meshing parts for the 3D CFD
model were shown from Figure 111 to Figure 118 in Appendix.
Table 4. Intakes, returns, and neutral boundary conditions for both the 2D and 3D CFD
model.
Airway Velocity (ft/s) Airway Airflow Quantity Percentage
Intake 1 5.2340 Return 1 0.0586
Intake 2 2.2998 Neutral 2 0.2234
Intake 3 0.7840 Return 3 0.7180
4.4.3 Assumptions
Several assumptions are made to save the time on establishing the model without
affecting the results significantly. There is no leakage between air passages. Heat transfer
is not taken into account and temperature of wall and air are considered to be constant.
The gravity of the air in the underground mine is neglected. Airflow in the underground
is incompressible and fully turbulent. The injection of the tracer gas does not affect the
airflow.
29
|
Virginia Tech
|
4.4.4 Governing Equations
Governing equations used in both the 2D and 3D CFD models are the continuity equation
(4-1) and momentum equation (4-2). The energy equation was not used because the
airflow is incompressible and there is no heat transfer in the model mine. User-Defined
Scalar (UDS) transport equation (4-3) was used because SF was simulated as a user
6
defined scalar in the 3D CFD model. A user defined function (UDF) was interpreted to
calculate the diffusivity in Equation (4-3). All the properties of SF were obtained by
6
solving the UDS transport equation. The standard k-epsilon model was utilized to
simulate the turbulent flow. SIMPLE scheme was chosen as the solution method. All the
equations shown below were obtained directly from the theory guide of Ansys Fluent
(Ansys 2009a).
Continuity equation:
ββVββ = 0 (4-1)
Momentum equation:
βP T
β(Vββ βVββ ) = β +ββ(ΞΌ (βVββ +(βVββ ) ) (4-2)
Ο eff
where Vββ is the mean velocity vector, ΞΌ = ΞΌ+ΞΌ , Β΅ is the molecular viscosity of air, ΞΌ is
eff
the eddy viscosity and it is computed from k-Ι model, βP is the pressure gradient.
User-Defined Scalar (UDS) Transport equation:
βΟΟ β βΟ
k + (Οu Ο βΞ k) = S , k=1,β¦,N (4-3)
i k k Ο
β βx βx k
i i
where and are the diffusion coefficient and source term, is an arbitrary scalar,
Ο is air density, u is velocity, and represent temporal and spatial derivative
π‘
separately, ( ) represents the convection term in the equation, and (Ξ )
represents the diffusion term in the equation.
A User defined function (UDF) was utilized for calculating the diffusivity of SF () in air
6
(Ansys 2009b):
30
|
Virginia Tech
|
ΞΌ
Ξ Ο = ΟΞ + (4-4)
eff Ο S
C
where is turbulent viscosity, is the turbulent Schmidt number, Ξ is the effective
π‘ π‘ eff
diffusion coefficient of SF in air, is the diffusion coefficient of SF in air in this
6 π‘ 6
study, Ξ Ο represents the effective diffusivity, ΟΞ represents the molecular diffusivity,
Ο Ο
and represents the turbulent diffusivity. The Schmidt number is set to 0.7 (Ansys
2006) because all the fluids modeled in the CFD models are gases. (Ansys Fluent 12.0
Theory Guide, 2009). The diffusion coefficient is set to 5.9Γ 0β6 m2/s in this study. Ward
(Ward & William, 1997) reported the range for the diffusion coefficient of SF in air was
6
from 5.9Γ 0β6 m2/s to 7.3Γ 0β6 m2/s. Besides, in Equation (4-4), the diffusion
coefficient is three orders of magnitude smaller than turbulent viscosity (Xu 2013). It
means the diffusion coefficient does not affect the results very much and 5.9Γ 0β6m2/s is
appropriate for this study.
4.4.5 2D CFD Model
4.4.5.1 2D CFD Model Setup
In the preliminary study (Zhang et al. 2014), the gob area was not taken into account. It
was added to the 2D CFD model in this paper. Due to the complicated geometry of the
model mine, the geometry was divided into seven parts in the 2D CFD model while the
3D CFD model was divided into eight parts. The reason for the difference on the number
of parts was that 2D CFD model did not have a thickness and the gob had the same height
as the coal seam. Since the geometry was divided into various parts in the 2D and 3D
CFD models, meshes for the CFD models were created based on the parts. However, the
parts of meshes for either the 2D CFD model or the 3D CFD model were combined to
one mesh before being solved in the Ansys Fluent. Additionally, the original geometry
was simplified by deleting the airways with no flow. The leakages of stoppings were not
modeled in the 2D CFD model. In addition, tracer gas was not introduced in the 2D CFD
model but it was simulated in the 3D CFD model. In order to know how the flow behaves
in the gob area, two cases were generated based on the porosity and permeability values
of the gob. In the first case, the gob area was treated as one zone. In the second case, the
31
|
Virginia Tech
|
gob area was divided into five zones (Yuan, Smith, and Brune 2006). The purpose of
creating the 2D CFD model was to eliminate the concerns on the one-zone case and five-
zone case. If the results from these two cases were almost the same, five-zone case would
not be built up in the 3D CFD model for this study.
Icem 14.5 was used to build up both casesβ meshes and Ansys Fluent 14.5 was applied to
obtain the solution. Since the porosity and permeability values of the mine were
unknown, assumptions of these values have been made based on a longwall study done
by other people (Yuan, Smith, and Brune 2006). The porosity and permeability values for
the five-zone case can be found in Table 5. Based on another longwall study (Lolon
2008), the porosity and permeability values were set to 12800 md (millidarcy) and 0.24
separately for the one-zone case.
Table 5. Assumptions for porosity and permeability values (five-zone case).
Permeability (md) Porosity
1000000 0.25
200000 0.24
70000 0.23
10000 0.22
5000 0.21
4.4.5.2 One-zone Case
The gob area in the one-zone case was treated as one part with the same porosity and
permeability values across the gob. Theoretically, the one-zone case was not true because
falling rocks had a size distribution across this area and porosity and permeability values
should also had a range in this area. The purpose for creating this case was to determine if
a zonal, more realistic model has a considerable effect on the results. An overview of the
one-zone case in the 2D CFD model is shown in Figure 10. CFD results from this case
were compared with the experimental results in Table 6.
32
|
Virginia Tech
|
Figure 10. Plan view of one-zone case.
Table 6. Results from one-zone case.
Sample One-zone Case CFD Results
Air quantity (kcfm) Error (%)
points (kcfm)
1 81.00 90.33 11.51
2 19.60 18.76 4.28
3 17.00 15.72 7.51
7 53.74 53.74 0.01
9 63.00 59.09 6.20
Form Table 6, it is clear that all the errors from the 2D CFD model are under 12% for all
the sample points. Sample point 1 has the largest error, which is 11.51%. This is not
surprising because the airflow at sample point 1 becomes very complicated after the three
intake airflows come together.
4.4.5.3 Five-zone Case
In this case, the gob was divided into five parts with different porosity and permeability
values. The reason for dividing the gob was because the compaction of caved rock was
not constant in the gob. It was compacted more close to the center of the gob than the
boundary of the gob since the loading of the overburden decreased from the center to the
boundary of the gob as the working face moved forward. Therefore, the porosity and
permeability values were not distributed evenly in the gob. An overview for this case was
shown in Figure 11. Results from the five-zone case were compared with the
experimental results in Table 7.
33
|
Virginia Tech
|
Figure 11. Plan view of five-zone case.
Table 7. Results from five-zone case.
Sample Five-zone Case CFD Results
Air quantity (kcfm) Error (%)
points (kcfm)
1 81.00 89.82 10.89
2 19.60 18.75 4.32
3 17.00 18.27 7.49
7 53.74 53.74 0.01
9 63.00 59.10 6.20
According to Table 7, the largest error from the five-zone case is about 11% compared
with the experimental results. Since the airflow at sample point 1 is complex, it was with
expectation that the largest error appears at sample point 1. At the same time, the results
from the CFD model verifies the accuracy of the model.
4.4.5.4 Results Comparison
Table 8. Results comparison among the 2D CFD models.
Sample Error (%)
points One-zone Case Five-zone Case
1 11.51 10.89
2 4.28 4.32
3 7.51 7.49
7 0.01 0.01
9 6.20 6.20
All the results from the one-zone case and the five-zone case summarized in Table 8. The
largest error is under 12% and appears at sample point 1 for both the two cases. Results
from the one-zone case and five-zone case are almost the same. In terms of the CFD
34
|
Virginia Tech
|
models, these results confirms that these results are accurate. It is not necessary to build
up the five-zone case in the 3D CFD model for this paper. There are several reasons for
the errors at the sample points. The experimental results may not be 100% accurate
because of the complex mine geometry. Additionally, several assumptions were made
when establishing the CFD model. The 2D CFD model is just a numerical model and it
cannot simulate everything as what it is in the underground mine. However, with such
complex mine geometry, the fact that errors of the results from the two 2D CFD models
are under 12% is acceptable.
4.4.6 3D CFD Model with and without Turbulent Viscosity by
Using UDF Approach
4.4.6.1 3D CFD Model Setup
Figure 12. Plan view of the 3D CFD model.
Based on the conclusions from the 2D CFD model, only the one-zone case was discussed
in the 3D CFD model. A plan view of the 3D CFD model is shown in Figure 12. The
caving height of the longwall was between 35ft to 40ft and the gob height was averaged
in the 3D CFD model. As a result, the height of the gob was 37.5ft in the 3D CFD model.
The height of the coal seam was 10.5ft according to the mine survey. The gob area was
still simulated as porous media and airflow in the gob was modeled as laminar flow.
Since the standard k-epsilon turbulence model was widely applied on turbulent flows in
35
|
Virginia Tech
|
the mining industry (Xu 2013), it was used to simulate the flow turbulence in the model
mine.
4.4.6.2 Mesh Independence Study
The purpose of mesh independence study is to ensure that various kinds of meshes will
not affect the results obtained from the 3D CFD models. At the same time, it shows the
results does not change as the mesh goes finer. In addition, based on the results from the
mesh independence study, optimal mesh size can be determined and errors from the mesh
can be minimized. Three kinds of meshes were created in this study and details were
shown in Table 9. Fine mesh has the most number of nodes and coarse mesh has the least
number of nodes. The fine, medium, and coarse mesh were shown in Figure 13. The 3D
CFD results from these three meshes were compared with the experiment data and were
shown in Table 10, Table 11, and Table 12 separately. Detailed comparisons of airflow
quantity and SF flow quantity with experiment data for the three meshes were shown in
6
Figure 14 and Figure 15 separately. Equation (4-5) was used to calculate the errors
between the results from CFD model and experiment at certain sample point or release
point.
|Result from one CFD modelβResult from experiment|
Error (%)= β 00% (4-5)
|Result from experiment|
Table 9. Summary of nodes numbers for three kinds of meshes.
Mesh Nodes Number (million)
Fine 22.10
Medium 12.46
Coarse 7.19
Figure 13. Plan view of the three meshes (from left: fine mesh, medium mesh, and coarse
mesh) in a cross sectional area.
36
|
Virginia Tech
|
SF Flow Quantity Error (%) Comparison
6
90
80
70
60
)
%
50
(
r
o40
r
r
E
30
20
10
0
SP1 SP2 SP3 SP9 RP1
Points
Fine Mesh Medium Mesh Coarse Mesh
Figure 15. SF flow quantity error (%) comparison among three meshes.
6
From Figure 15 and Figure 16, the CFD model with coarse mesh has a large error for
both the airflow quantity and SF flow quantity at sample point 3. For all the other
6
sample points, the coarse mesh CFD model has almost the same results as the fine mesh
and coarse mesh CFD models. It reflects that sample point 3 is considerably sensitive to
the mesh size. It also proves that airflow at sample point 3 is complicated and has a
relatively larger range of quantity than the other sample points.
Obviously, the coarse mesh is not a good choice to build up the 3D CFD model.
Theoretically, the medium mesh should be used as the standard mesh because of its less
computing time compared with the fine mesh. However, after considering the complex
mine geometry, the fine mesh was selected as the final mesh for the steady-state
simulation and transient simulation.
4.4.6.3 Sensitivity Study for Porosity and Permeability Values
No accurate data about porosity and permeability of the gob were provided. The reason
was that there were no porosity and permeability tests done in the mine. Then a
39
|
Virginia Tech
|
sensitivity study for the porosity and permeability data were conducted to analyze the
effects of these data on the CFD results. In a CFD model, permeability is represented by
two properties, which are inertial resistance and viscous resistance.
Porosity and permeability values for a gob in anther underground mine in the western
U.S. were shown in Table 13 and they were reported in Lolonβs Ph.D dissertation (Lolon
2008). Since the porosity and permeability values were obtained based on experiment and
field data from the underground mine (Lolon 2008), these results were reliable to be
utilized in this paper. Additionally, the underground coal mine in this study is also
located in the western U.S., which means the two mines have some geologic similarities.
The porosity and permeability data were shown in Table 13 and were applied to the 3D
CFD model in this paper.
Table 13. Porosity and permeability values used in the CFD models.
Porosity 0.24
Viscous Resistance (πβ2) 7.91Γ 07
Inertial Resistance (πβ1) 14700
The relationship between the porosity (n) and permeability (k) were demonstrated by
Kozeny-Carmen equation (Scheidegger 1957) as shown in Equation (4-6). The equations
for computing viscous resistance and inertial resistance were shown in Equation (4-7) and
Equation (4-8) (Lolon 2008) separately.
π2 π3
k = π β (4-6)
0 ( βπ)2
where π is the mean particle size (m), k is the theoretical specific permeability (π2),
π
and n is the porosity.
(4-7)
C =
1 k
3.5β( βn)
(4-8)
C =
2 d βn3
m
where C is the viscous resistance and C is the inertial resistance for the gob.
1 2
40
|
Virginia Tech
|
Table 14 shows the porosity and permeability values used in the sensitivity study. The
porosity values were set as what they were in the table. The viscous resistance and
inertial resistance were then computed by using Equation (4-7) and Equation (4-8).
Again, the purpose of the sensitivity study was to see the effect of the various porosity
and permeability data on the results of the CFD models.
Table 14. All the porosity and permeability values used in the sensitivity study.
Porosity C (πβ2) C (πβ1)
1 2
0.15 2.25Γ 08 67341.47
0.2 8.40Γ 07 26738.53
0.24 4.39Γ 07 14700.00
0.26 3.27Γ 07 11257.69
0.3 1.91Γ 07 6932.21
0.35 1.04Γ 07 4053.65
0.4 5.91Γ 06 2506.74
All the detailed results from the sensitivity study are shown from Table 22 through Table
28 in the Appendix. Error comparisons made for airflow quantity and SF flow quantity
6
among these cases were presented in Figure 16 and Figure 17, respectively.
Airflow Quantity Error (%) Comparison
12
10
) 8
%
(
r
6
o
r
r E 4
2
0
SP1 SP2 SP3 SP7 SP9 RP1
Points
0.15 0.2 0.24 0.26 0.3 0.35 0.4
Figure 16. Airflow quantity error (%) comparison with different porosity and
permeability data.
41
|
Virginia Tech
|
SF Flow Quantity Error (%) Comparison
6
50
40
)
% 30
(
r
o
r 20
r
E
10
0
SP1 SP2 SP3 SP9 RP1
Points
0.15 0.2 0.24 0.26 0.3 0.35 0.4
Figure 17. SF flow quantity error (%) comparison with different porosity and
6
permeability data.
According to Figure 16, the largest error of airflow quantity from the CFD models is 10%
and it appears at sample point 3. Except sample point 3, all the other sample points have
almost the same results from the CFD models. The airflow errors at sample point 3 range
from 2% to 10% in the CFD model with different porosity and permeability data. It is
understandable because sample point 3 is located right next to the gob and the airflow is
significantly sensitive to the change of porosity and permeability in the gob.
From Figure 17, all the SF flow quantity results from the cases with different porosity
6
and permeability values are pretty close. It means the porosity and permeability values
does not have a significant effect on the CFD results.
Based on Figure 16 and Figure 17, the porosity and permeability values does not affect
the results very much. It eliminates the concerns on not having field data on the porosity
and permeability for the mine gob. Therefore, the porosity and permeability values
(shown in Table 13) prove fine to be utilized for this mine geometry.
42
|
Virginia Tech
|
4.4.6.4 Steady-State Simulation
4.4.6.4.1 Two Kinds of Simulations and Results
According to the userβs guide of Ansys Fluent (Ansys 2006), the diffusivity term in the
UDS equation is a constant. Due to the flow turbulence, the effective diffusivity changes
as shown in Equation (4-4). The diffusivity of SF in air was calculated with the turbulent
6
Schmidt number, which meant the influence of turbulence was considered in the
equation. In order to realize the effects of the turbulence on the diffusivity of SF in air
6
for this mine geometry, the steady-state simulation in the 3D CFD model was completed
under two conditions. One was without taking the turbulent diffusivity (Equation (4-4))
into account. Results from this case were shown in Table 15. The other one was with the
turbulent diffusivity (Equation (4-4)) taken into account., which was much closer to the
actual situation in the underground. In the underground, the SF flow was affected by not
6
only the molecular diffusivity but also the turbulent diffusivity. Results were shown in
Table 16. The airflow quantity and SF quantity results from the two cases were
6
compared with each other in Figure 18 and Figure 19 separately.
Table 15. Results from the 3D CFD model without turbulent diffusivity.
Fluent Fluent
Air Air SF
Sample SF SF Results Results 6
quantity 6 6 Error Error
points (cfm) (kcfm) Air SF
(kcfm) 6 (%) (%)
(kcfm) (kcfm)
SP1 81.00 0.0129 1.29Γ 0β5 86.89 8.99Γ 0β6 7.27 30.32
SP2 19.60 0.0033 3.30Γ 0β6 19.60 1.99Γ 0β6 0.00 39.82
SP3 17.00 0.0028 2.80Γ 0β6 15.51 1.57Γ 0β6 8.76 43.86
SP7 53.74 53.74 0.00
SP9 63.00 0.0089 8.90Γ 0β6 62.99 6.35Γ 0β6 0.01 28.63
RP1 53.60 0.0088 8.80Γ 0β6 52.88 8.82Γ 0β6 1.34 0.23
43
|
Virginia Tech
|
SF Flow Quantity (in 10^(-6) kcfm) Comparison
6
10
)
m 9
f
c
k 8
)
6 7
-
(
^ 6
0
1 Without Turbulent
5
n
i Diffusivity
( 4
y
t i 3 With Turbulent
t
n
a 2 Diffusivity
u
Q
1
w
o 0
l
F
SP1 SP2 SP3 SP9 RP1
6
F
S Points
Figure 19. SF quantity comparison between two cases.
6
According to Figure 18 and Figure 19, the results from these two cases does not have a
big difference. Therefore, only the CFD model without turbulent diffusivity interpreted is
examined. From the Table 15, on one hand, all the errors for airflow are below 9%. The
largest error for air quantity appearing at sample point 3 is 8.76%. This error is
acceptable due to the complexity of the geometry and the low airflow quantity at the
sample point. On the other hand, the errors for the SF flow ranges from 28% to 44% at
6
the sample points. For SF flow, the largest error, which is 43.86%, appears at sample
6
points 5. It is reasonable that sample point 3 has the largest error because it is located
next to the gob and this area is difficult to be accessed to do good experiments.
Additionally, based on the experimental data, the quantity of SF flow at sample point 1,
6
which is 0.0129 cfm, is greater than that at the release point 1, which is 0.0088 cfm. The
difference between these two results is about 47%. It proves the fact that error exists in
the experiment data. Similarly, based on the 3D CFD model, the quantity of SF flow at
6
sample point 1, which is 8.99Γ 0β6 kcfm, is greater than that at the RP1, which is
8.82Γ 0β6 kcfm. However, the difference between these two results is less than 2%.
Since all the boundary conditions were balanced in the 3D CFD model, the 2% difference
45
|
Virginia Tech
|
on the results between these two points comes from the numerical error (discretization
error) of Ansys Fluent itself.
It is obvious that all the SF concentration results from the 3D CFD models are smaller
6
than that from the experiment. It is because the experiment is a transient simulation. But
the results were compared with both the results from steady-state simulation and transient
simulation in the CFD models. The researcher, who did the experiment, took an average
value of the experiment results with both high and low concentration values (as shown in
Figure 42). The averaged results were reported as the steady-state simulation results of
the experiment by the researcher. However, it actually increased the values of the
experimental results since there were higher values than lower values in the experimental
results.
4.4.6.4.2 Contour Comparisons between the Two Cases (With Turbulent
Diffusivity and Without Turbulent Diffusivity)
To show the complex airflow behavior in this complicated mine geometry, some figures
obtained from both the two cases of 3D CFD models were shown below. Contours for
both the SF concentration and velocity magnitude at different locations, such as areas
6
near the working face and the release point, were analyzed. More contours are presented
in the Appendix.
The overall SF concentration distribution for the two cases can be seen in Figure 20 and
6
Figure 21. SF concentration distribution near the working face for the two cases are
6
shown in Figure 22 and Figure 23. According to the four figures, the distribution of SF
6
concentration in the 3D CFD models are the same for both the two cases.
46
|
Virginia Tech
|
Figure 35. Contour (YZ plane) of SF mass concentration right after sample point 1 in the
6
CFD model with turbulent diffusivity.
According to the contours from Figure 20 to Figure 35, it is clear that contours for both
SF concentration and airflow velocity magnitude at different locations are the same in
6
the 3D CFD models with turbulent diffusivity or without turbulent diffusivity. It means
that turbulent diffusivity does not affect the results very much for this specific mine
geometry.
4.4.6.4.3 Monitor Lines for the Two Cases (With and Without Turbulent
Diffusivity)
To better understand these two cases, three monitor lines were created for both of the two
CFD models. Locations of the monitor lines can be found in Figure 9. All the monitor
lines were created in the center of the cross sectional areas and an example is shown in
Figure 36. Since the injection of SF does not affect the airflow distribution, only the
6
distribution of SF concentration across the monitor lines were examined. SF
6 6
concentration comparisons between the two cases at the three monitor lines are presented
from Figure 37 to Figure 39.
56
|
Virginia Tech
|
Based on Figure 37, the SF mass concentration stays the same across the monitor line 1
6
for both of the two cases. The reason is that the monitor line 1 is located a little bit far
from the release point and no other flows can disturb the airflow at monitor line 1.
Moreover, it shows the SF is evenly distributed at monitor line 1 for the two cases.
6
From Figure 38, the SF mass concentration differs across the monitor line 2 for both of
6
the two cases. At the same time, the trend of SF concentration in the two cases are the
6
same. The SF concentration is high at the beginning of the monitor line 2 and then it
6
drops to about 305 PPB (part per billion) at the end of the monitor line. The SF
6
concentration trend at monitor line 2 can also be verified in Figure 34 and Figure 35. It is
because monitor line 2 is located right after the location where all the airflows from three
intakes join.
The distribution of the SF mass concentration at monitor line 3 is like a normal
6
distribution chart as shown in Figure 39. It is within expectation since the monitor line 3
is located right after the release point of SF . The location of monitor line 3 also explains
6
the high SF mass concentration value (about 1.05 PPM (parts per million)) in the figure.
6
The results from Figure 39 matches the contours of SF mass concentration very well
6
from Figure 24 to Figure 27.
As a result, the turbulent diffusivity does not contribute a lot to the 3D CFD model. It
may be caused by this mine geometry and other uncertainties.
4.4.6.5 Transient Simulation
The field study in this paper was completed after a six-hour experiment (Jong 2013) in
the underground mine. Since research group was curious about the difference between
the results from the 3D CFD model and the results from the experiment. The 3D CFD
model with transient simulation was then built up to make a comparison to the
experiment. Since the turbulent diffusivity does not affect the results in the steady state
simulation, the 3D CFD model with transient simulation was simulated without
interpreting the turbulent diffusivity.
59
|
Virginia Tech
|
Figure 41. SF mass concentration over time at sample points and monitor planes.
6
Since the model mine has a very large scale (about 4000ft in length and 1400ft in width)
with a relatively small velocity magnitude (about 5.2ft/s at intake1), it takes a long time
for the SF to reach the three outlets (as shown in Figure 40), especially return 1. Since
6
monitor plane A is very close to the release point 1, SF flow reaches this plane very fast
6
and a peak shows up after about five minutes (shown in Figure 41). Additionally, mass
concentration of SF in monitor plane A is about 850 PPB, which is much larger than that
6
in other monitor planes and sample points. For the rest monitor planes and sample points,
there is an increasing trend shown in each of their profiles (shown in Figure 41). SF flow
6
reaches steady state in about six hours. That SF are detected proves the airflow actually
6
follows the path from the release point 1 to the sample points.
4.4.6.5.2 Results Compared with Experimental Results
Results from the 3D CFD model were compared with the experimental results at four
sample points. Locations for the sample points could be found in Figure 8 and Figure 9.
An overall comparison for the results at the four sample points between the 3D CFD
model and the experiment was shown in Figure 42. Results comparison at a single sample
point (sample point 1, 2, 3, and 9) were shown from Figure 43 to Figure 46, respectively.
61
|
Virginia Tech
|
Figure 46. Comparison for transient simulation results at SP9.
According to Figure 42, it is clear that experimental results fluctuate a lot while the CFD
results are more stable. In examining Figure 43, Figure 44, Figure 45, and Figure 46,
results from the experiment have a higher SF concentration than that from the 3D CFD
6
model on average. SF mass concentration increases at first and then it has a decreasing
6
trend for all the four sample points in the experiment. Oppositely, it takes at most half an
hour for SP1, SP2, and SP3 to reach a steady state in the CFD model. SP9 reaches its
steady state at around six hours because it is farthest away from the SF release point.
6
There are several reasons for the higher SF mass concentration results in the experiment.
6
The k-epsilon turbulence model is not perfect and does not represent turbulent flow in the
model very well due to the complex mine geometry. Additionally, there were people and
mining equipment (a huge chuck) moving when the experiment was conducted.
However, these factors were not simulated in the model. Besides, the injection of tracer
gas was not modeled exactly as what it was in the underground. In the 3D CFD mode,
tracer gas acts more like a "marker" flowing with the airflow. In the injection point for
the tracer gas, air passes the injection point and carries the tracer gas with it in the CFD
model. But in reality, the tracer gas was released from a container held by an individual.
64
|
Virginia Tech
|
Moreover, the airflow and SF flow might not be well-mixed when the sample points
6
were taken in the experiment. Oppositely, the conditions set in the 3D CFD model are
perfect and flows of air and SF are mixed very well in the model.
6
4.5 3D CFD Model with Species Transport Model
As aforementioned, the turbulent diffusivity in the previous 3D CFD models was taken
into account by interpreting the UDF into the model. However, the results from the
steady-state simulations showed that there was no difference between the cases with and
without turbulent diffusivity. Additionally, according to Figure 24 and Figure 25, back
diffusion appeared before the release point for both the two cases with and without
turbulent diffusivity. Moreover, SF diffused quickly after it was released for both two
6
cases. However, in theory, SF should concentrate in the center of the duct and gradually
6
diffuse according to Equation (4-3). In addition, results from the CFD model with
turbulent diffusivity taken into account should be different from that from the model
without turbulent diffusivity. The reason is that in Equation (4-4), the molecular
diffusivity is three orders of magnitude smaller than turbulent diffusivity, which means
the effective diffusivity of SF in air depends on the turbulent diffusivity. Furthermore,
6
the back diffusion before the release point is not understandable.
To conclude, the 3D CFD models built with and without UDF did not show the right
behavior of the tracer gas in the underground. As a result, the new CFD model with
species transport model was created to correctly simulate the tracer gas behavior in the
underground.
4.5.1 Geometry of the Model Mine
The geometry for the 3D species transport model is shown in Figure 47. There are four
intakes, two returns, one neutral, four sample points, two release points, and three
monitor planes presented in the figure. An enlarged view, which is shown in Figure 48, is
made to see the locations of the sample points and monitor planes more clearly.
65
|
Virginia Tech
|
intake 4 and a mixture of air and SF was released from it. In reality, only SF was release
6 6
from the RP1, the reason why the mixture instead of only SF was released was discussed
6
later on. The rest five surfaces of the cube were treated as stationary wall. The volume
inside the cube was not treated as a flow domain so there was no flow inside the cube.
All the intakes in the model were assigned velocity-inlet boundary conditions. Two
returns and a neutral in both the models were assigned outflow boundary conditions. The
velocity and the airflow quantity percentage values were obtained from the field study in
the underground mine. Details about the boundary conditions were shown in Table 17
and Table 18.
The mesh, which was used in the species transport model, was divided into several parts
before being solved in Ansys Fluent 14.5. Interface boundary conditions were used to
combining these meshes together when they were ready to be solved.
Table 17 Intake boundary conditions in the species transport model.
Airway Mixture Velocity (ft/s) SF Mass Fraction
6
Intake 1 5.234 0
Intake 2 2.2998 0
Intake 3 0.784 0
Intake 4 5.5 0.0001078
Table 18 Boundary conditions for the two returns and one neutral in the species transport
model.
Airway Flow Rate Weighting
Return 1 0.0586
Neutral 2 0.2234
Return 3 0.718
As presented in Table 17, the boundary conditions for intake 1, 2, 3, return 1, 3, and
neutral 2 in the species transport model were the same as that in the 3D CFD models
(with and without UDF). SF mass fraction in the intake 1, 2, 3 were all set to zero, which
6
meant that no SF was released from these three intakes.
6
68
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.