13th United States/North American Mine Ventilation Symposium, 2010 – Hardcastle & McKinnon (Eds.

)
© 2010, MIRARCO - Mining Innovation
A study of leakage flow in a laboratory model and using CFD
M. Stephens & F. Calizaya
Mining Engineering Dept., University of Utah, Salt Lake City, Utah, USA

ABSTRACT: This paper presents the results of two models built to study the leakage flow in underground coal
mines. The first is a laboratory model designed to characterize leakage flow paths such as stoppings, overcasts,
doors, and regulators. The second is an analysis of air leakage in a 2-entry system using CFD. The results of the
lab model show that a direct application of the square law to the laboratory readings yields resistances that are
orders of magnitude greater than those obtained from field measurements. The results of both studies indicate that
the pressure drop across stoppings follow a decay function rather than a linear one. The results also indicate that
50-60% of total leakage occurs in the first 40% of the mine workings.

1 Introduction and maintenance of ventilation control structures and

appropriate planning. In order to minimize leakage, it is
Every underground mine contains harmful air necessary to have a better understanding of leakage flow
contaminants, such as toxic or flammable gases, dust, characteristics. In this paper, leakage flow has been further
fumes, smoke, heat or radiation. The fundamental purpose characterized in a laboratory model as well as using
of the underground ventilation system is to provide the computational fluid dynamics (CFD) software.
quantity and quality of air required to dilute these
contaminants to safe levels where personnel are required to 1.1 Problem Statement
work or travel. But the quantity of air required at the
workings is only a portion of the total quantity that must be In the western U.S., it has become common practice to
induced by the fan(s). Any quantity of air that passes develop longwall panels using the 2-entry system in order
through the main fan but is not usefully employed to create safer mining conditions under deep cover. But
somewhere in the mine is considered leakage. A few this mining method, along with increasing panel sizes and
problems associated with excessive leakage include coal beds containing high methane emission rates, have
uncontrolled recirculation, high fan pressures and high made it more challenging to properly ventilate the working
power costs which are often overlooked (McPherson, sections. In many mines, panels have reached lengths in
1993). Due to the vast number of stoppings required, excess of 3,000-m (2-mi). If gateroad pillars are developed
leakage is much more profound in coal mines than in at 30-m (100-ft) centers, at minimum, this requires 100
metal/non-metal. Currently, mine development in deep and stoppings in a single gateroad. Furthermore, as mining
extensive underground coal resources in the U.S. is progresses, the distance from the portals to the working
hindered due to leakage. These resources could be faces increases; in some cases this distance is in excess of
accessed and leakage in coal mines can be reduced perhaps 24-km (15-mi). This extensive development requires
most significantly with the use of booster fans. While thousands of ventilation control structures such as
leakage cannot be eliminated, it can be significantly stoppings, doors, regulators, overcasts and others.
reduced with good ventilation practices, proper installation

Figure 1 Schematic of the University of Utah’s coal mine model

485
 Each ventilation control device installed in a mine is a allowing a wide variety of stopping configurations. For
leakage path and its characteristics impact a mine’s example, the configuration that allows the least amount of
volumetric efficiency. These structures include permanent leakage in the system is to place gate #4 into all four
stoppings, overcasts, seals, bleeder entry systems, fan positions. This ability to vary the simulated stopping
housings, face curtains, tubing, etc. The quantity of air that resistance makes it possible to represent various types of
will leak through a device is dependent upon the pressure stoppings having differing resistance values. For instance,
differential across it, and the value of its resistance to gate #4, the closed stopping, can represent a permanent
airflow. Its resistance further depends on the construction seal or concrete stopping, whereas stopping #0 (fully open)
materials used as well as the quality of workmanship might represent a temporary stopping such as a brattice
employed (Calizaya & Stephens, 2006). cloth. The intermediate ones could represent a variety of
particular types of stoppings (steel panels, cementitious
blocks, concrete blocks).
2 Laboratory Model
A plan view of the coal mine model is shown in Figure 1. 2.2 Test Procedure
This model consists of 14.6-cm diameter ductwork One complete test consists of measuring static and/or
configured in a common U shaped ventilation system. The velocity pressures at each pre-determined station. A test is
intake and return drifts are joined by 5 crosscuts. The first initiated by inserting the desired gate valve configuration
four (A, B, C and D) are kept blocked by interchangeable, into the stopping slots, setting the fan motor frequency to a
perforated gate valves that form leakage paths (stoppings). predetermined level, and powering the fan. Once the
The last is kept open to represent an active mining section airflow in the system has stabilized, a calibrated
or face. The system is powered by a 2.5-kW centrifugal manometer and pitot static tube is used to measure static
blower fan equipped with a variable frequency drive and velocity pressures. A six-point, equal-area traversing
motor. This allows the motor to be set at any speed ranging method is used to measure velocity pressure readings from
from 0 to 60 Hz. When the fan motor is operated at 60 Hz which the average velocity and flow quantity is calculated.
(3600 rpm), the fan circulates 28.6 m3/min of air at This procedure is repeated at each selected station to
1500 Pa of static pressure. determine flow quantities throughout the system. Stopping
Table 1 Perforated gate valves used to represent stoppings resistances are calculated using Atkinson’s equation (P =
RQ2).
Holes
Gate Valve Open Area
Set Diameter (%) 2.3 Measuring Leakage
Number
(cm)
The total percent leakage in the system is the ratio of the
#0 (open) 1 5.6 100
quantity of air that is short circuited before reaching the
#1 37 0.64 27.2 working face to the total quantity circulated by the fan. It
#2 21 0.64 15.4 is calculated using the following equation:
#3 21 0.32 3.9 QT − QF
LT = (100) (%) (1)
#4 (closed) 0 0 0 QT
where: LT = total leakage (%),
#0 #1 #2 #3 #4 QT = airflow at the fan (m3/s),
QF = airflow at the face (m3/s),
Flow quantity through the duct is calculated by
multiplying the average velocity by the cross-sectional
area. Since the duct area in this model is assumed to
remain constant (0.0167 m2), for convenience, the average
velocity can represent air flow quantity and the total
leakage can be calculated by:
VT − V F
LT = (100) (%) (2)
VT
Figure 2 Photograph of the gate valves used in the lab. where: LT = total leakage (%),
VT = average velocity at the fan (m/s),
2.1 Leakage Flow Model VF = average velocity at the face (m/s),
In Figure 1, VT is measured at station 1 and VF is
There are 5 sets of perforated gate valve "stoppings" measured at station 5.
having holes of different diameter and number as shown in The leakage flow through an individual stopping is
Table 1 and Figure 2. Any one of the five gate valves can assumed to be the difference between flows measured on
be inserted into any stopping position (A, B, C and D),

486
the upstream and downstream sides of the split. For each set of gate valves was calculated using all tests in
example, in Figure 1, the leakage flow through stopping A which that type of gate was used. For instance, three
is the difference between the flows measured at stations 1 different tests were conducted using gate #1 (fan settings at
and 2, respectfully. 30, 45 and 60 Hz). So data from all three tests (shown in
Assuming that the quantity of air exiting the system is Table 4) were used to determine the average resistance for
equal to the amount entering the system, VT could this set of gate valves. Note in Table 3 that the average
alternatively be measured at station 10 and VF could be resistance values between each set of gate valves differ
measured at station 6. Similarly, leakage through considerably (by roughly an order of magnitude). In
individual stoppings could be measured using the stations addition, they are orders of magnitude higher than those
in either the inlet duct or the return duct. typically measured in the field (Bruce & Koenning, 1987,
Duckworth et. al., 1995).
2.4 Lab Results The primary purpose of the lab model is to obtain
measurements that can be used to calibrate a computer
A complete airflow study was conducted on the lab model
model so that more extensive and complicated
using various configurations. Table 2 shows an example of
configurations can be examined. In this case, the computer
the results of a test setup using stopping size #1 in
modeling was done using ANSYS Workbench 12.0, a
positions A, B, C and D. In this case, the fan motor was set
common, commercial CFD (computational fluid
to 30 Hz so that the fan supplied 14.2 m/s of air at about
dynamics) modeling software package.
400 Pa of static pressure, but only 8.55 m/s of air reached
the face. Using Equation 2, the total leakage for this Table 3 Average stopping resistance values calculated
configuration is 40%. Directly applying the square law to using Atkinson’s equation
the measured figures yields a resistance range from about LT, % Average
178,000 to 512,000 Ns2/m8, thus the average for this test Gate Valve Gate Valve
Resistance,
Set Position 30 Hz 45 Hz 60 Hz
configuration is 345,000 Ns2/m8. 2
Ns /m
8

Table 2 Perforated gate valves used to represent stoppings #0 (open) A, B, C, D 62.19 61.80 64,000
#1 A, B, C, D 39.82 39.90 41.86 358,000
2 8
Location V, m/s Leakage, % Δ p, Pa R, Ns /m #2 A, B, C, D 25.62 25.55 24.35 10,300,000
A 2.24 39.5% 255.0 183,294 #3 A, B, C, D 2.91* 35,900,000
#4 (Closed) A, B, C, D <3.00* 55,100,000
B 1.30 23.0% 207.0 440,717
C 1.09 19.2% 160.0 488,051
D 1.03 18.3% 137.5 462,871 3 CFD Modeling
Face 8.55 7.5 369
A 3D model of the lab was drawn to scale and meshed
Fan Duty: VT = 14.2 m/s, Pstatic = 400 Pa using the integrated ANSYS Workbench modules
(DesignModeler and Meshing). The Fluent module is used
for performing calculations and viewing results.

3.1 CFD Model Calibration

The model was calibrated to the 30 Hz configuration
shown in Table 4. In Fluent, the gates are modeled as
porous media. The last open crosscut contains a wire-mesh
screen that provides some resistance, so it too was
modeled as a porous medium. By trial and error, the input
parameters were adjusted until the CFD model velocity
measurements were within 5% of the lab model's.
3.1.1 Gate Valves
The stopping gates may be modeled either as porous
media, or, if the plate is thin, as porous jumps. In this case,
Figure 3 Velocity trend of leakage flow through stopping
the plates 3.175 mm in thickness, so modeling each as a
size #1.
porous jump yields better convergence. Three factors
A graph of the flow quantities from Table 2 is shown in define a porous jump zone: face permeability, porous
Figure 3. The graph indicates that over 60% of the total medium thickness, and a pressure-jump coefficient. A thin
leakage occurs through the first two stoppings with nearly porous medium has a finite thickness over which the
40% leaking through the first stopping alone. It also pressure change is defined as a combination of Darcy's
indicates an exponential decay function. Law and an additional inertial loss term (User's Guide,
A summary of the laboratory results for various gate 2009):
valve settings is given in Table 3. The resulting leakage LT
was calculated using Equation 2. The resistance value for

487
 ⎛1 1 ⎞ 3.1.3 Final Model Calibration
ΔP = −⎜ μV + C 2 ρV 2 ⎟Δn (3) Once the porous jump coefficients are known, the model
⎝α 2 ⎠
can be calibrated. First the inlet velocity is set to the value
where: ΔP = pressure difference (Pa), measured in the lab. Then α and C2 are adjusted by trial
μ = fluid laminar viscosity (Pa·s), and error until the model is calibrated. Using this method
ρ = fluid density (kg/m3), there can be many different solutions. The CFD model is
α = permeability of the medium (m2), considered calibrated when the critical velocities are
C2 = pressure jump coefficient (m-1), within ±5% of the lab measurements as shown in Table 5.
V = velocity normal to the porous face (m/s), Table 5 Model correlation using average velocity (m/s)
Δn = thickness of the medium (m).
Sta. 1 Sta. 2 Sta. 3 Sta. 4 Sta. 5
Therefore when experimental data of the pressure drop
and velocity through the porous component are available, Lab 14.20 11.98 10.68 9.59 8.55
the porous media coefficients can be calculated. This CFD 14.20 12.40 10.90 9.56 8.54
process is described below using the data from the gate #1 % Diff. 0.0% 3.4% 2.0% 0.2% 0.3%
tests (Table 4). Table 6 Input parameters used in Fluent for gate #1
Table 4 Pressure and velocity measurements from three configuration at 30 Hz
lab configurations using gate #1 BOUNDARY BOUNDARY
NAME TYPE CONDITION DESCRIPTION
Calculated
Stopping Velocity Pressure Inlet Face Velocity inlet 14.2 m/s Constant velocity
Resistance
Position (m/s) Drop (Pa) Equivalent gauge
(N2s/m8) Inlet Face Pressure inlet 400 Pa
pressure
A 14.3 1017.5 337,856
Gauge pressure
B 13.1 857.5 340,985 Outlet Face Pressure outlet 0 Pa
60 Hz
C 12.4 700.0 306,597
α = 2.7e-8 m2 permeability
D 12.4 612.5 268,273
Pressure jump
A 10.8 575.0 331,754 A,B,C,D Porous jump C2 = 1000 m -1
coefficient
B 10.5 467.5 289,106 Δn = 0.003175 m
45 Hz Gate thickness
C 9.2 392.5 311,767
α = 2,000,000 m2 permeability
D 6.8 337.5 498,496
Open Pressure jump
A 9.7 250.0 178,978 Porous jump C2 = 134 m-1
Crosscut coefficient
B 5.7 212.5 448,700
30 Hz Δn = 0.003175 m screen thickness
C 4.8 170.0 499,382
D 4.6 142.5 465,077 3.1.4 Fluent Input Parameters
The input parameters used in the modeling process are
3.1.2 Calculating porous jump coefficients shown in Table 6. The velocity is set constant at 14.2 m/s
The velocity and pressure data in Table 4 are used to which is approximately equivalent to 400 Pa. For the open
generate a scatter plot and trend line through the points crosscut, α was set to 2E+6, providing essentially no
yielding a 2nd degree polynomial equation in the form: resistance to flow and C2 was set low enough to provide
only slight resistance.
ΔP = AV 2 + BV (Pa) (4) An example of the model output showing contours of
velocity (top) and pressure (bottom) are depicted in
where A and B are the resulting unit-less trend line
Figure 4. Figure 5 shows a closer view of velocity vectors
coefficients: 80.42 and 21.01, respectively. Substituting
near crosscut A. Eddies caused by turbulence can be seen
these values in Equation 3 yields the following
occurring on the inlet side of the crosscut. Also note the
relationships:
velocity variation through the crosscut due to the changes
⎛1 ⎞ in duct diameter.
21.01 = ⎜ μ ⎟Δn (5)
⎝α ⎠
3.2 CFD Model With 10 Crosscuts
⎛ 1 ⎞
80.42 = ⎜ C 2 ρ ⎟Δn (6) To further evaluate airflow patterns and pressure profiles
⎝ 2 ⎠
for various leakage flow conditions, the calibrated CFD
using: ρ = 1.01 (kg/m3), model was modified to have stopping gates in 10 crosscuts
μ = 1.7894 x 10-5(Pa-s ), instead of only 4. A schematic of the layout is shown in
Δn = 0.0031 (m), Figure 6. This layout closely resembles development of a
2-entry, longwall gateroad and the results can more easily
the calculated results are: C2 = 50,157 m-1 and α = 2.70E-9 be compared to a section having on the order of 100
m2. Although these coefficients are sometimes negative, in stoppings. Two conditions are evaluated: (1) at constant
Fluent they should always be entered as positive values inlet velocity, and (2) at constant face velocity.
(User's Guide, 2009).

488
Figure 4 Fluent output showing contours of velocity (top) and pressure (bottom).

Figure 5 Fluent output showing velocity vectors at stopping A (left) and B (right).

Figure 6 Schematic of modified CFD model having 10 stoppings plus an open face.

489
 stoppings and in case 2, each leakage path represents 10
3.2.1 Case 1-Constant inlet velocity.
stoppings. Figure 8 indicates that in case 2, nearly 60% of
The purpose of this case is to examine the changes in
the leakage occurs in the first 40 stoppings. Whereas in the
leakage during development while the velocity provided
calibrated model, 50% occurs in the first 40 stoppings.
by the fan remains constant. All of the input parameters
were the same as in the calibrated model (see Table 5).
In the calibrated model, VT (station 1) was 14.2 m/s
and VF (station 5) was 8.6 m/s, resulting in LT = 40%. In
this case, VT (station 1) is 14.2 m/s and VF (station 11) is
3.45 m/s, resulting in LT = 76%. The leakage nearly
doubles which is caused solely by increasing the number
of leakage paths.
3.2.2 Case 2-Constant face velocity.
In the real world, a minimum quantity is required in the
working section regardless of the extent of development.
So in this case, it was assumed that the face would have a
minimum quantity requirement of VF = 8.6 m/s. This
velocity is the same as in calibrated CFD model.
Therefore, all of the input parameters remain the same as
in case 1 except the inlet velocity which is increased
accordingly.
In order to achieve a minimum of VF = 8.6 m/s, VT
(station 1) had to be increased to 30 m/s, which results in Figure 8 Cumulative leakage % by number of stoppings.
about 71% leakage. This also increased the pressure
requirement nearly 4 times from 400 Pa to 1,430 Pa.
4 Discussions
4.1 Lab Model
The lab model was limited to only 5 crosscuts due to the
physical constraints of the space available as well as the
size of the fan. In practice it is recommended rule of thumb
is to make velocity measurements at a minimum distance
of four times the diameter of the opening on the upwind
side of a split and a minimum distance of ten times the
diameter (10-D rule) on the downwind side. Due to the
limited amount of space available in the lab, it was not
possible to follow the 10-D rule. The limited space
availability is also the reason why the crosscuts had to be
reduced in diameter for the gate valves.

Figure 7 Pressure drop across individual stoppings. 4.2 Porous Jump Parameters
A sensitivity analysis was conducted on the 3 parameters
3.3 Results used to define a porous jump (Δn, C2and α). The pressure
jump coefficient (C2) had the most profound effect. When
A comparison between the two cases is shown in Figure 7. C2 is large (> 10,000 m-1), the effect of α is not
This figure shows the differential pressure across the noticeable. For the CDF model, the combined effect of
stoppings. In both cases, the trend is that of an exponential these values is equivalent to the resistance of the stopping.
decay function.
A crucial factor in considering leakage flow is
4.3 CFD Model
determining where the majority of the leakage occurs. A
meaningful method for doing this is to examine the Although the lab model was accurately replicated in the
cumulative leakage beginning with the first stopping. The CFD model, there are some notable differences. Firstly, the
cumulative leakage curves for cases 1 and 2 are very physical model contains joints located throughout the
similar because the percent leakage rates are on the same ductwork that would seemingly cause some additional
order (70%). A comparison is made between case 2 and turbulence. There was no attempt to replicate this.
the calibrated model which only has 4 leakage paths. In Secondly, the CFD model showed good velocity profiles
this comparison, both conditions are plotted as if the throughout the inlet side of the ductwork, but velocity
number of leakage paths is equivalent to 100 stoppings. So profiles on the outlet side of the ductwork fluctuated
in the calibrated model, each leakage path represents 25 significantly. In the lab model, measurements in the outlet

490
ductwork seemed to give more consistent velocity profiles also indicated that with increasing distance from the fan,
than those in the inlet stations. both pressure drop and leakage flow through the stoppings
exhibits an exponential decay function. If similar stoppings
are used throughout the system and the fan is kept at
5 Conclusions constant velocity/pressure, the results indicate that when
Stoppings and other control devices used in underground the number of leakage paths was increased from 4 to 10,
mines can be viewed as air paths with high resistance. The the leakage increased from 40% to 70%. The only ways of
amount of air that leaks through these devices depends on avoiding this increase in leakage is to either increase the
various factors including the type of construction materials resistance by improving the tightness of the stoppings or
used, workmanship, and stopping inspection and by decreasing the main fan pressure by utilizing a booster
maintenance. Improperly constructed and poorly fan.
maintained structures will significantly lower this
resistance and cause undue leakage.
Acknowledgements
The circumstances of a real-world mine rarely exhibit
the ideal conditions needed to obtain the most accurate The authors gratefully acknowledge the financial support
measurements. The airflows and pressure drops in an and cooperation of the National Institute for Occupations
underground mine are subject to considerable variation due Safety and Health (NIOSH) under contract No. 200-2009-
to movement of equipment, opening vent doors, and other 30328. Acknowledgment is also extended to Dr. Michael
changes. In addition, mine layouts are often extremely G. Nelson for his support on this project.
complex and may be such that the airflow profile at the
location where a measurement is required is not fully
developed. This can make so that fluid-flow laws are not References
truly applicable. Nevertheless, a practical effort must be Calizaya, F. and Stephens, M. (2006), Studies of leakage flow in
made. the U.S. underground coal mines. In J.M. Mutmansky &
In ventilation planning, resistance values are often R.V. Ramani (Eds.), Proc. 11th U.S./North American Mine
measured in the field or in an experimental lab. These Ventilation Symposium 2006, pp 599-606, Taylor & Francis
measurements are used in a simulated model which is Group Plc., ISBN: 0415401488.
McPherson M.J. (1993), Subsurface Ventilation and
calibrated to match the field conditions. This calibrated
Environmental Engineering. Chapman & Hall, London,
model is then used to further evaluate the pressure/quantity ISBN: 0412353008.
requirements for future mining scenarios. Bruce W.E. & Koenning T.H. (1987), Computer Modeling of
Experimental tests conducted at the University of Underground Coal Mine Ventilation Circuits, Selection and
Utah’s coal mine model indicate that for a set of similar Application of Airway Resistance Values. Proceedings of
stoppings, the trend of pressure drop across the stoppings the 3rd US Mine Ventilation Symposium, SME, Littleton,
decreases with distance from the main fan. The trend CO, pp. 519-524.
resembles an exponential decay curve more than a linear Duckworth I.J., Wallace G.K. & Wise R. (1995), Ventilation
one. A direct application of the square law to the lab model Planning and Design of the Skyline Mines. Proceedings of
the 7th US Mine Ventilation Symposium, SME, Littleton,
gave resistance values several orders of magnitude greater
CO, pp. 9-14.
than those typically measured in real mine settings. Fluent User's Guide (2009), ANSYS FLUENT 12.0 USER'S
A CFD model was calibrated to within 5% of the lab GUIDE. ANSYS, Inc, ISO 9001:2008
measurements. Additional analyses with the CFD model

491