University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Colorado School of Mines
|
Figure 2.18a shows the soil conditioning injection ratios (FIR, BIR, PIR, WIR) as well as
foam expansion ratio (FER) and foam agent concentration ( during mining of ring 77 (see
figures 2.21b and 2.21c). These plots are calculated based on P CLf)C data readings per five seconds
and advance rate of the EPBM. Note that time is presented in h:mm format throughout with t =
0:00 as the start of excavation for the presented ring.
(a)
(b)
(c)
Figure 2.18: Ring 77 (a) Soil conditioning Injection ratio; (b) FER; (c) foam agent concentration
Soil conditioning as major ground treatment during mining is controlled either by the
operator or automatically with soil conditioning system default setting. Air flow rates are
regulated with injection pressure to maintain the flow rates. Moving average per six data point of
air flow rates are shown in figures 2.19a and 2.20a versus mining time ring 75 (PLC machine
records data per 5 seconds). As it is illustrated in this figure, each foam generator received a
specific flow rate either by default setting of the flow control system or by the operator. Moving
21
|
Colorado School of Mines
|
average per six data point of solution pump flow rates are shown in figures 2.19b and 2.20b
versus mining time ring 75. It is evident that solution flow rate among the 20 pumps is very
consistent; however, air flow rate is fluctuating significantly and is varying greatly among the 20
foam generators (FG). Interestingly, air flow rates are oscillating per 30 seconds. This oscillation
can be due to cutterhead rotation. All of the ports on the bulkhead and screw conveyors are
closed except port 28 that is located on top of the chamber. This suggests that foam is delivered
100% to the ports on the cutterhead in all FGs other than FG8. Cutterhead with 1.0rpm rotation
is traveling per 30 seconds and for half of its circumference from top to bottom hydrostatic
pressure is changing. Moving average per six data point of foam pressure is depicted versus
180°
mining time ring 75 in figures 2.19c and 2.20c. The same behavior is evident in these Figures.
Pressure oscillating per 30 seconds for about 1bar in few pumps and less than 0.5bar at other
others. This can suggest that foam pressure is changing with port location on the cutterhead.
Figures 2.21a and 2.21b are depicting the foam flow rates during mining of ring 77 to
demonstrate that soil conditioning system is controlling the air flow based on ground pressure.
When an increase in chamber pressure in sensed, the air control valve is operated automatically
to increase the air flow rate. By increasing air flow rate, FER is also increasing.
Foam is expected to expand due to a pressure difference (ΔP) between the foam gun and
face pressure (Boyd’s law). The foam volume is the combination of air flow rate under specific
pressure and solution flow rate under solution pump pressure. The volume of foam (bubbles) in
front of the face when mixed and travelled with muck through the CH opening and the chamber
may change significantly. The CH outer rim circumference is 55.0 m and injection ports are
installed in different diameters from cutterhead center to perimeter. For instance, a foam port
near the outer rim of the CH travels around 0.9 m/s during mining. In addition, with rotation rate
of 1.0 rpm this port will travel from top to bottom of the face in 30 sec. considering average
foam flow of around 180 lit/min, this port will inject 3.0 lit/sec. This rate is changing among
different port locations and foam gun foam flow rates.
22
|
Colorado School of Mines
|
CHAPTER 3 - ANALYSIS OF EPBM ALONG THE ALIGNMENT
3.1. Engineering Soil Units and Geotechnical Parameters
The complex sediment distribution of the Seattle area is due to the advance of several
glaciers that eroded previous sediments and overlaid new sediments. This includes glacier clay
and silt, outwash sand and gravel, till and till like material such as glaciomarine drift. This will
also increase the probability of boulders and drop stones. The geotechnical data report (GDR)
and baseline report (GBR) provide information about the ground geology and geotechnical
characteristics. Here, ground conditions and soil units are explained briefly. Eight major soil
units are defined in the GDR as follows:
Engineered and Non-engineered Fill (ENF): Predominately consists of very loose to very
dense Sand with varying amounts of silt and Gravel. It also consists of Silt, Clay and organic
Silt.
Recent Granular Deposits (RGD): Predominately consists of loose to dense or locally
very dense Sand, and Sandy Silt. It also contains localized zones of Silt, Sandy Gravel, and
Gravelly Sand with varying lateral extent and thickness.
Recent Clay and Silt (RCS): Predominately consists of soft to stiff, Silty Clay and Clayey
Silt with variable amounts of Sand and Gravel and localized zones of medium dense to dense
Clayey Sand. It also contains layers, lenses, and dikes of cohesionless Sand with varying lateral
extent and thickness.
Till Deposits (TD): Predominately consists of a very dense or hard cohesive mixture of
Gravel, Sand, Silt, and Clay. It contains fractured cohesive Clay and Silt with varying lateral
extent and thickness, and also contains interbeds, dikes, and lenses of saturated cohesionless Silt,
Sand, and Gravel with varying lateral extent and thickness.
Cohesionless Sand and Gravel (CSG): Predominately consists of dense to very dense
Silty Sand to Sandy Gravel. CSG also contains lenses and layers of Clay and Clayey Silt that
26
|
Colorado School of Mines
|
provide cohesion within soil layers and impede downward/ upward and lateral movement of
groundwater.
Cohesionless Silt and Fine Sand (CSF): Predominately consists of very dense Silt, fine
Sandy Silt, and Silty fine Sand. CSF also consists of interbeds and lenses of silt and fine Sand
with minor Clay content, which still behave like cohesive soils and have varying lateral extent
and thickness and contain fractures.
Cohesive Clay and Silt (CCS): Predominately consists of hard, interbedded Silt and Clay.
It also consists of multiple layers, lenses, and dikes of cohesionless Silt, Sand, and Gravel, with
varying lateral extent and thickness.
Till-Like Deposits (TL): Consists of materials with high spatial variability and can grade
from an unsorted mixture of Silt, Sand, and Gravel to clean or relatively clean Sand, in a relative
short distance. TL predominately consists of a heterogeneous mixture of dense to very dense
Gravel, Sand, and fines, and exhibits little to no cohesion. TL also consists of layers and lenses
of glacial till and layers and lenses and dikes of saturated cohesionless Gravel. Table 3.1
provides the geotechnical parameters of the introduced soil units.
The aquifer system in Puget Sound is mostly in alluvial deposits of coarse grained
sediments, including sand and gravel. Puget Sound, due to the complexity of glacier stratigraphy,
has a complex ground water flow and hydrogeological condition. The permeability varies in
adjacent soil layers and even within a single stratigraphic unit (see table 3.1). As a result,
frequent perched groundwater-bearing layers and multiple piezometric surfaces are expected.
Table 3.2 provides information about water and ground depth to the crown of the tunnel
as well as hydrostatic pressure, cover to diameter ratio (C/D) and coefficient of lateral earth
pressure at rest (k ). These data only include early mining (10% of the project) and are based on
0
the approximate station of the each ring. The source of data is from both the GDR and GBR.
Tunneling before ring 67 included the pre mining ground treatment (jet grout) and dewatering
grout treatment. As Table 3.2 shows, ground water and ground depth increase from ring 67 to
ring 145. Hydrostatic pressure increases from 0.9 bar to 1.7 bar.
27
|
Colorado School of Mines
|
In order to estimate the shear strength of the muck in different rings, a hand vane shear
device was used. Three different tests were done on muck samples throughout each ring and
normalized into a single value of shear strength (see figure 3.3). Having the vane size as well as
the torque values, the vane shear strength is then derived using the ASTM standard (ref). Vane
shear testing was performed under atmospheric conditions. A density test was conducted to
estimate the muck density per ring using ASTM standard (ref).
Figure 3.3: Hand vane shear test
For a rectangular vane of height to diameter ratio (H/D) = 2, the vane shear strength can
be determined as follows where Tmax is the maximum measured torque:
(3-1)
6.𝑇𝑚𝑚𝑚
( 𝑆𝑢) = 7𝜋𝐷3
3.3. Ground Condition and Muck Characteristics
Figure 3.4 illustrates the tunnel alignment and geology based on the GBR. Approximately
60% of the tunnel is in granular soil and 30% in full face of cohesive clay and silt. This study
focuses on the initial 10% of the alignment that is in a combination of different layers. This
section is expanded in Figure 3.4 and the soil units introduced at the beginning of this chapter are
illustrated with their associated colors. The tunnel profile from the crown to the invert is shown
for rings 55-150. The grain size distribution (GSD) results from muck testing are superimposed
on this Figure. Gravel, sand, and fines content of the conditioned soil are illustrated in red, green
and blue, respectively. GSD test results from ring 75 to 145 are shown in figure 3.5 and 3.6.
31
|
Colorado School of Mines
|
Figure 3.4 also depicts that the range of gravel, sand and fines content varies among
selected rings. However, Figure 3.6 shows that GSD test results are very similar in some
consecutive rings. Consistency of GSD during one ring and close test results from consecutive
rings suggest that cutterhead tools consistently cut and mix different soils at the face. Muck in
the chamber is a combination of all soil layers, which means each layer in the tunnel profile
contributes to the muck final characteristics. With interpretation of geotechnical investigation
boreholes from the GDR, the tunnel profile in different rings are compared due to the ring
chainage (see figure 3.7).
Figure 3.7: Tunnel profile interpretation based on GDR borehole result
Table 3.3: Grain size distribution based on muck grain size distribution
Ring %Gravel % Coarse Sand %Fine Sand %Fines
75 11.17 18.31 26.18 44.34
77 12.2 17.5 24.5 45.8
82 18.72 12.69 16.75 51.84
83 20.77 12.75 16.99 49.49
87 10.55 8.21 15.66 65.58
90 7.76 9.62 14.77 67.85
91 9.20 15.73 19.40 55.67
103 10.98 19.52 17.83 51.67
112 5.77 21.47 24.23 48.11
113 6.25 18.87 26.02 48.86
114 6.58 19.24 27.84 46.35
131 10.67 22.54 30.35 36.44
132 7.93 22.83 31.47 37.77
145 4.12 26.69 31.88 37.32
34
|
Colorado School of Mines
|
The GSD, vane shear strength, slump, density, LL device taps and moisture content
results from muck testing are presented in Figure 3.8. GSD was determined for 14 rings (see
Table 3.3). GSD test results are also shown in Figure 3.8a. The moving average per 2 data of
GSD is also shown to provide an overall trend. An increase in muck fines content and a decrease
in muck sand content is evident from ring 75 to 90. Figure 3.8a indicates that from ring 90 to
145, muck fines content decreased significantly and sand content is increased remarkably.
Besides an exceptional increase at ring 82 and 83, muck gravel content is lower than 10%. Figure
3.8f shows a decrease in density from ring 55 to 90 and an increase from ring 90 to 145. With
comparing figures 3.8a and 3.8f, a clear relationship is seen between density and GSD test
results.
Figure 3.8b is provided the vane shear test result under atmospheric pressure. In this
Figure, the average value of the test results on each ring is shown. Figure 3.8c is presented the
number of Casagrande device taps on the muck samples passing the sieve #40. Recall that the
Casagrande device is used to determine the liquid limit (LL) of cohesive soil (the LL
corresponds to the moisture content when 25 taps closes the gap over a 1 cm length of soil).
From rings 75-87, the number of taps varied from 30-45 indicating that the muck consistency
was above the LL. This is consistent with low shear strength behavior observed in the vane shear
tests. The LL of the original cohesive soil in this area ranged between 58-82% and liquid limit
moisture content is between 23-35%. With focusing on the results, it is noticeable that exact
same Su value results in exact same number of taps. For instance, ring 75 and 131 with Su value
of 2.6kPa both results in 30 Casagrande device tapes. Similarly, consecutive rings 82 and 83, 89
and 90, as well as 111 and 112 provide comparable results.
The reduction in Casagrande device taps observed at ring 90 is significant and is not
reflected by an increase in Su that one would expect. Perhaps the soil type has changed here.
Casagrande tap numbers beyond ring 100 are not meaningful since this is primarily a granular
soil. Figure 3.8d provides average slump test results on each ring. Due to the close relationship
of material shear strength and viscosity, it is possible to establish a relationship comparing slump
and vane shear test results. It is depicted in figures 3.8b and 3.8d, that muck with vane shear test
result ranged between 2-3 kPa presents 0.6-4.6 cm of slump.
35
|
Colorado School of Mines
|
Figure 3.9 presents soil conditioning changes throughout 100 rings of the tunnel. Figure
3.9a shows injection ratios and their standard deviations. An increase in standard deviation
means that foam flow rates or advance rate of the EPBM is fluctuating. FIR’s standard deviation
from ring 115 forward is very high, due to a malfunction in foam injection (ref). Figure 3.9b
presents the volume of injected foam and bentonite (slurry) and their percentage to the ring
volume. It is evident that bentonite slurry injection increased from ring 60 and reached to a peak
value of 18% at ring 90. After ring 90 bentonite slurry injection decreased to zero at ring 112.
Polymer is directly injected between rings 67-71 and 116 -149. Total ring foam volume from
rings 58-113 fluctuates around 200 m3 (FIR = 40%) and decreases to around 100 m3 (FIR =
25%) from rings 113-146 (with fluctuations). Foam volume per ring volume fluctuates between
25-35% and is decreased remarkably from ring 112 forward in range of 5-20% with minimum
values at ring 132 and 145. FER gradually increased from 5 to 10 from ring 59-158 with
exceptional increases at ring 59, 74, 104, 125, and 131.
Figure 3.9: Soil conditioning representation
37
|
Colorado School of Mines
|
CHAPTER 4 – RING MINING OPERATION ANALYSIS
4.1. Introduction to Operation Data
Ring 75 excavation is used here to introduce machine operation. Figure 4.1 presents the
key EPBM operational parameters vs. time during mining of ring 75 (t < 1:09 where h:mm is the
time convention used throughout). The mining of this ring is continuous without any stops. The
tunnel profile is 40% cohesive clay and silt, 40% till deposits, 10% cohesive sand and gravel,
and 10% recent granular deposit based on the geotechnical data report.
As mentioned in chapter three, the on-site GSD of the muck material in this ring shows
11.2% gravel, 18% coarse sand, 27% fine sand, and 44.3% fines. Material flows on the machine
conveyor belt quite easily and the slump test (the standard test to rate the flowability of the
muck) result is 0.8 cm. The material on the belt was found to be visually consistent, i.e., no
visible chunks of unconditioneded material. The measured vane shear strength was 3.1 kPa. The
average muck moisture content is 24% andLL device taps was 30, which suggests that the muck
consistency was above the LL.
EPBM operation is predicated on forward displacement and cutterhead rotation speed
control. The EPBM uses feedback control to apply the appropriate torques and thrust to maintain
the desired displacement and rotation rates. During mining, the operator manually controls the
cutterhead rotation speed ( ), the center agitator rotation speed ( ), and the thrust cylinder
pressure together with the
𝜃
𝐶fi𝐶rst screw conveyor rotation speed ( 𝜃𝐶𝐶) and the second screw
conveyor rotation speed ( ). In this ring the is kept constant
𝜃
𝑆e𝐶x1cept from a small change
at time 00:08). 𝜃𝑆𝐶2 𝜃𝑆𝐶1
The advance rate of the machine is is influenced by several factors such as thrust
cylinders pressure (thrust force), geology of the ring, shield friction, and trailing gear. The
advance rate also depends on the soil conditioning and material flow from the cutterhead, which
will be discussed in chapter five.
38
|
Colorado School of Mines
|
fluctuations. varied from 27,000-35,000 kN-m during mining. A comparison of figures 4.1a
and 4.1b rev 𝑇e𝐶a𝐶ls the often complex relationships between machine operating parameters. In
general, F, , and the AR are correlated. They increase and decrease with each other as
evidenced i 𝑇n𝐶 t𝐶he
𝜃
n𝐶u𝐶merous notch reductions throughout mining (e.g., t =23, 28, 52 min). The
period t = 35-50 min reveals the more subtle relationship between these four variables. Here,
and AR increase steadily while and F remain constant. The increase in AR under const 𝑇a𝐶n𝐶t
produces a greater penetratio 𝜃n𝐶 𝐶rate (advance distance per revolution or AR/ ). This larger
𝜃‘b𝐶i𝐶te’ results in a greater . 𝜃𝐶𝐶
𝑇𝐶𝐶
The soil conditioning inputs are presented in figure 4.1c, including the foam injection
ratio (FIR), the polymer injection ratio (PIR), the bentonite injection ratio (BIR) and the water
injection ratio (WIR). Recall from Chapter two that these reflect volumetric ratios of injected
foam, polymer, etc. to encountered or formation soil (including ingested pore water). Each of
these ratios reflects the sum of respective injections from all ports through the CH, in the
excavation chamber, and in the SC1 and SC2. FIR varies significantly among 20 pumps and the
average FIR is between 45-70% with an increase in the beginning and end of mining. The BIR is
kept constant around 20% and the WIR around 10%. No polymer injection was used throughout
this ring. The overall FER was 5.0 during this ring. The magnitudes of FIR, BIR and WIR are
considerable. Collectively, FIR+BIR+WIR ≈ 80%, indicating that 8 parts of foam, bentonite and
water by volume is added to every 10 parts of encountered ground.
Figure 4.1d shows the opening of the screw conveyor gates. G1 is very crucial during
mining to control the out flow of the material. This gate controls the discharge of material from
the SC1 to the SC2. In ring 75, G1 is all the way open to incorporate the maximum discharge
from SC1. G2 is used for safety purposes; in case of water invasion or unusual pressure increase
resulting in blow outs (i.e., high pressure air). G2 is partly open to control the discharge of
material on the belt. The operator controls these gates by looking at the camera on the discharge
chute as well as the feed back from the SC1 and the SC2 pressure readings.
Figure 4.2 illustrates additional data from ring 75. In addition to the operating data in
figures 4.2a - 4.2c, the excavation chamber earth pressure sensors (EC-EPS) (Figure 4.2d) and
the screw conveyor earth pressure sensors (SC-EPS) in Figure 4.2e are also presented. These
40
|
Colorado School of Mines
|
gage pressure sensor readings reflect muck pressure above 1 atm. The EC-EPS are oriented on
the stationary vertical bulkhead, and therefore measure total lateral pressure. Similar colors have
been used for sensors at the same elevation. As introduced in chapter two, sensors R1-R6 are
located on the right side of the bulkhead from top to bottom and the L1-L6 sensors are located on
the left side of the bulkhead from top to bottom.
Six gage EPSs were installed in each screw conveyor. The SC pressure sensors are
positioned at mid-elevation of the SC casing and capture the lateral total pressure. Figures 4.2d
and 3.35e convey EC and SC pressures during mining (t < 1:09) and standstill / ring build (t >
1:09). The vertical difference in chamber pressure due to the self-weight of the muck is
significant. The chamber pressure near the crown (top) of the EPBM is approximately 1.5 bar
during mining while the chamber pressure at the invert (bottom) of the EPBM is 4 bar. The
vertical pressure gradient will be explored in detail in chapter six.
A horizontal difference in chamber pressure is clearly evident at all elevations within the
chamber. Rightside pressures are higher than leftside pressures. This coincides with
counterclockwise CH rotation (see figure 4.2a). Note that in other rings where the CH rotation is
clockwise, leftside chamber pressures are higher than rightside. This phenomena will be
analyzed later. The EC-EPSs record a decrease in chamber pressure upon transition to standstill
that plateaued to a constant value with time. This decrease is due to the absence of rotational
mechanical influence, a stop in volume flow (into and out of EC) as well as the material coming
to equilibrium with the ground pressure.
The chamber muck experiences pressure fluctuations during mining and standstill that are
quite informative. All pressure readings increase considerably during the three minute ramp up
(initial values reflect the equilibrium position from ring 74 standstill). They all then gradually
decrease until reaching a fairly steady state behavior from t = 15-60 min. This steady pressure
suggests that the operator had good control over the muck flow in and out of the chamber.
Chamber pressures increase again significantly during the 3 min ramp down, and gradually
decrease during standstill. The nature of chamber pressures will be discussed in greater detail
ater in chapter five.
41
|
Colorado School of Mines
|
The cut and scraped formation soil flows into the chamber through the CH openings
during machine advance. The changes in AR are associated with , , F, and together
with the muck characteristics. and along with th 𝜃e 𝑆𝐶g1at 𝜃e 𝑆𝐶o2pening a 𝜃n𝐶d𝐶 material
characteristics control the outflow
𝜃
𝑆r𝐶a1tes. The
𝜃
𝑆s𝐶c2rew conveyor takes advantage of the pressure
difference between the EC pressure and the atmospheric pressure on the machine conveyor belt.
Advance rate and SC facilitate and control the flow.
Figure 4.2e depicts the pressure sensor readings of both SCs. The SC-EPS 1 and 4 are not
working due to damage during operation. Figure 4.2e shows the dissipation of pressure from
sensor 2 to 10. Sensor SC1-2 is very close to the chamber and presents values near 3.0 bar,
which indicates initiation of the pressure dissipation (compared to excavation chamber’s R5, R6
and L5, L6). The sensors SC1-10 is the last sensor in the screw conveyor right before the belt
and shows a pressure of approximately 0.25 bar.
The muck pressure (Figure 4.2e) decreases to atmospheric pressure as the material travels
through SC1 and SC2 and onto the belt (to the right of SC2). Screw conveyor pressures exhibit
rapid and gradual fluctuations similar to those observed in the excavation chamber. A
comparison of SC and EC pressure reveals they are related. For example, the increase in SC
pressure at the end of mining (beginning at t = 1:07 and due in part to the reduction in SC
rotation) causes the increase in excavation chamber pressure.
Most of the subtle changes observed in the SC pressure are coincident with similar EC
pressure changes, though their polarity (increase or decrease) is not always the same. For
example, during ramp-up, EC pressure increases while SC pressure decreases. During this period
SC rotation speed is increased and therefore relieves pressure in the SC. This will be discussed
again in Chapter five.
The screw conveyor torque (here ) is a valuable parameter to evaluate the muck
characteristics in each ring. remains q TuSitCe1 constant at 75 kN.m throughout excavation with
an increase to 90 kN.m durin TgS Cth1e first 5 min of excavation which is coincident with few changes
SC1 and SC2 (i.e. time 00:08 and 00:11).
43
|
Colorado School of Mines
|
4.2. Ring 67 to 78 Operation analysis
As an example of more variable machine behavior, ring 67 operation data is depicted in
the series of plots in figure 4.3. The mining of this ring was continued for 1:46 with four stops.
Per the GBR, the tunnel profile is about 50% cohesive clay and silt, 30% till deposits, 10%
cohesive sand and gravel, and 10% recent granular deposit. GSD testing was not conducted for
this ring.
The mining of ring 67 included multiple stop-start sequences that provide insight into
relationships between chamber and SC pressures and the EPBM operating parameters. The first
stop lasted 22 minutes, the second stop five minutes and third stop lasted three minutes. The
fourth stop was brief to change cutterhead and center agitator rotation direction at time 1:41.
After this change mining continues for five minutes.
The AR fluctuates frequently, and and are also changed by the operator
throughout mining. is kept higher than 𝜃𝑆𝐶1 to pr 𝜃e𝑆v𝐶e2nt cloggage in between. In this ring
operator adjust the sc 𝜃r𝑆e𝐶w2 conveyor gates durin 𝜃g𝑆 o𝐶1petaion. Mining operations initiate with G1 and
G2 wide open (1500mm) and G3 partially open (500mm). After 10 minutes of mining and 20
minutes stop, the operator adjusted the gates at mining restart to control the material on the
discharge chute. The reason behing operating the gates at this point of mining can be due to
watery muck splash out of the chute.
and AR follow the same trend. During mining, F remains constant around 125,000
kN. How T𝐶e𝐶ver, minor fluctuation of F is not visible due to the scale of the plot. Figure 4.3 shows
the magnitude of oscilations. Comparing the changes in F with AR one can find a relationship.
This relationship will be discussed in chapter five.
Soil conditioning inputs are presented in Figure 4.3c. The FIR varies significantly among
20 pumps and the average FIR is between 25-45%. The BIR is kept constant around 12% and the
WIR around 8 %. No polymer direct injection used throughout this ring. The magnitudes of FIR,
BIR and WIR are considerable. Collectively, FIR+BIR+WIR ≈ 60%, indicating that 6 parts of
foam, bentonite and water by volume is added to every 10 parts of encountered ground. FER is
averaged 4.0 in 20 pumps.
44
|
Colorado School of Mines
|
Ring 71 operation is shown in a series of plots in Figure 4.6. Mining of this ring is
continued for 1:06 and stopped once due to a change of cutterhead rotation direction. The mining
of this ring started without rotating the CA. Just before changing the CH rotation direction the
operator starts CA in a clockwise direction, and after changing the CH rotation to clockwise, the
CA rotation is changed to counterclockwise. EC-EPS readings are fairly constant during mining
except for a rapid decrease at time 0:45 for 10 minutes. This rapid decrease is inversely related to
the AR. The is constant and the AR is increased. At time 0:54 the operator decreases the AR
back to 30mm 𝜃𝑆/m𝐶 in and shuts down the . At time 56:00 the operator restarts the back to 5
rpm and mines for 10 minutes with cons 𝜃ta𝑆𝐶nt EC pressure. 𝜃𝑆𝐶
Soil conditioning inputs are presented in Figure 4.6c. The FIR varies significantly among
20 pumps and the average FIR is between 25-45%. The BIR is kept constant at around 12% and
WIR around 8%. The magnitudes of FIR, BIR and WIR are considerable. Collectively,
FIR+BIR+PIR+WIR ≈ 90%, indicating that 9 parts of foam, bentonite, polymer and water by
volume is added to every 10 parts of encountered ground. The FER is averaged 5.0 in 20 pumps.
Unlike rings 67 and 75, polymer was injected in this ring to reduce water in the muck. PIR was
constantly injected throughout the ring at around 22%.
The and the AR are strongly correlated in this ring giving enough evidence to
support that Ta𝐶n𝐶 increase in AR is accompanied by an increase in the . The is associated
with the and the material characteristics. In this ring they are no Tt𝐶 𝐶correlat Te𝑆d𝐶 to each other
when com 𝜃𝑆p𝐶a1red to ring 67. The is around 35 kN-m, which is well below the in ring 67
that ranged between 70-90 kN- Tm𝑆𝐶. This lower torque and absensce of a strong
T
𝑆𝐶relationship
suggests that the shear strength of the material in this ring is lower than in ring 67.
Ring 77 operational data is shown in Figure 4.5. The geology in the tunnel profile was
similar to ring 75. The mining of this ring is continued for 01:17. The observed slump was 0.7
cm and the vane shear strength was 3.1 kPa. The mining of ring 77 included two stop-start
sequences that provide insight into relationships between chamber and the SC pressures and the
EPBM operating parameters. The first stop lasted 10 minutes; the second stop was brief to
change cutterhead and center agitator rotation direction at time 0:54. The CH rotaion direction is
changed at time 0:54 and the right and the left EPS reading illustrate a very clear change. For
47
|
Colorado School of Mines
|
Ring 78 operational data is shown in Figure 4.9. The mining of this ring is continued for
01:09, stopping once to change cutterhead rotation direction. The geological profile of this ring is
very close to ring 77. AR, and fluctuate considerably and they follow the same trends.
The is constant at 1.2
θ
SrpC1m.
θ
SaCn2d AR are clearly associated. Overall F is constant and
aroun θdC H58,000 kN and is 30- T4C0H kN-m, both are much lower than ring 77. The ranges
between 20,000-32,000 TkSNC-m, similar as in ring 77. TCH
In this ring, like ring 71, using one half of the F, the same range of AR is generated,
comparable to rings 67, 75, and 77. Rings 78 and 71 both exhibit lower and F. Unlike rings
67, 75, and 77, the does not show a strong correlation with AR. Esse TnStCially, the AR and the
F are correlated, whi TcShC suggests that the is also corelated to the F. Comparing EC-EPS at the
crown from ring 67-78 shows an increase
T
dSuCe to the machine mining deeper underground.
The influence of AR on the EC-EPS data is not signicant when comparing to ring 71 and
77. For instance, in time frame 0:37 - 0:45, except from a small fluctuation in the SC data, the
AR is decreasing; however, the EPS readings are constant. The same phenomenon happened in
ring 67 right after a stop in mining. The absence of a strong correlation can be due to dynamics
of flow in the chamber, described in Chapter 5.
Similar to ring 77, the SC-EPS and the EC-EPS show the same trends. Having similar
changes (i.e. time 0:10) in both screw conveyor and excavation chamber can suggest that the
scraped and cut material presents a continuous structure to develop flow and dissipate the
pressure from the EC eo atmospheric pressure on the machine conveyor belt. Interestingly, SC-
EPS2 and SC-EPS3 in the SC1 and SC-EPS5-10 in SC2 depict the change in chamber pressure
gradually. The SC-EPS2 and 3 show this change at the same time of the EC-EPS change.
However, SC-EPS5-10 show the change with a delay of three minutes maximum (looking at
SC-EPS 8 and 9, figure 4.11). It is evident that the is higher than the at time frame
0:02-0:07, which suggests that the SC2 is partiallly fil θleSdC 2with material. Figure
θ
4S.C110 shows a drop
in chamber pressure after mining (during standstill) from time 1:09 to 1:40. After time 1:40 EC-
EPS is plateaued.
51
|
Colorado School of Mines
|
4.3. Ring 81-101 Operation Analysis
Ring 81 operational data is shown in Figure 4.12. GSD test results shows that the muck
contained 18.6% gravel, 13.4% coarse sand, 17% fine sand (29.4% sand) and 52% fines. This
translates to an 8% increase in gravel, 15% decrease in sand and 8% increase in fines compared
to ring 75 and 77. The soil exhibited 0.2 cm slump and the muck condition on the belt was
visually observed to be more sticky than rings 77 and 75. Chunks of unconditioned material were
observed. The vane shear strength was 3.8 kPa under atmospheric pressure. Average muck
moisture content is 26% and recalling from chapter three, the LL number of taps was 41, which
demonstrates that muck consistency was below the LL.
The mining of this ring was continued for 01:07. Mining stopped once for five minutes at
time 0:21. All machine operation parameters in this ring are very consistent with minimum
fluctuation compared to previous rings. F is around 115,000 kN and is 30,000 kN-m in this
ring. and show similar changes. and AR also followin TgC Hthe same trends. In this
ring, t ThSeC gate G θ2S Co1pening was about half op TenC H(700 mm).
Chamber pressure is relatively constant throughout the ring and the level 1 EPS reading is
higher than ring 78 at around 1.5 bar. The CH is rotated counterclockwise and the CA is rotated
clockwise. EC pressure on the right side is higher than the left side, and is consistent with the
counterclockwise rotation of the CH.
The EC-EPS experiences a drop after mining and during standstill on the right hand side.
Left side pressure mainly plateaus after mining. The SC-EPS shows a consistent pressure drop
during standstill. This can suggest that material on the right hand side is counteracting the ground
pressure. This will be discussed in chapter six. The EC-EPS R1 and L1 are recording similar
values, while R2 and L2 recording a horizontal pressure difference. This horizontal pressure
difference is consistent in all levels except from level five and six. Horizontal pressure difference
will be discussed in greater detail in chapter six.
54
|
Colorado School of Mines
|
The tunnel profile for ring 82 is located in the same geology as ring 81. The muck GSD
test results revealed 20% gravel, 13% coarse sand, 17% fine sand (29% sand) and 51% fines.
Figure 4.13 illustrates the slump test result is 0.2cm (same as ring 81). The muck condition on
the belt is more sticky than ring 77 and 75. Muck consistency is still fairly good on the belt, and
no chunks of unconditioneded material are observed. Vane shear strength equaled 3.7 kPa,
around the same value of ring 81. The average muck moisture content is 22% and the number of
LL device taps is 40. Ring 82 mining was halted after 5 minutes due to undesirable muck
conditions and low chamber pressure (see figure 4.15). Note G3 is only slightly open and then
closed at 5 min. The long standstill after mining of ring 81 (30 hours) may have contributed to
this. Mining was re-initiated at t = 54 min. By the end of mining at time 01:47, EC pressure
decreased significantly. The CH and the CA were rotated clockwise to elevate the pressure
(Figure 4.15).
Figure 4.13: Slump test from the material on the belt - Ring 82
Figure 4.14: Slump test failed- Ring 83
56
|
Colorado School of Mines
|
Ring 87 operational data is shown in Figure 4.19. The geological profile within the face
per the GBR included 50% cohesive clay and silt (high clogging potential), 40% till deposits, 5%
cohesive sand and gravel, and 5% recent granular deposits. The muck GSD test results showed
10% gravel, 8% coarse sand, 16% fine sand (24% sand) and 65% fines. This demonstrates a 20%
decrease in sand and 20% increase in fines portion compared to rings 75 and 77. The slump
equaled zero. The muck visibly appeared much more sticky that observed during rings 77 and
75. The muck appeared inconsistent on the belt, with many chunks of unconditioned material
visible. Muck vane shear strength was 3.3 kPa. The average muck moisture content was 35% and
the number of LL device taps was 46. Ring 87 mining was completed in 54 minutes. EC and SC
pressures are fairly constant throughout mining. EC and SC pressures increased significantly
after mining was stopped.
The geological profile of ring 90 was similar to ring 87. The muck GSD test results
revealed 8% gravel, 10% coarse sand, 15% fine sand (25% sand) and 68% fines. The observed
slump was zero and the materials was sticky (see figure 4.18). Muck on the belt is observed a lot
more sticky than ring 77 and 75. The consistency was very chunky. The vane shear strength was
4.4 kPa, slightly higher than previous rings. Average muck moisture content was 38% and the
number of LL device taps was 19 ( which is inconsistent with the LL of the formation soil = 58-
82%). Ring 90 operational data is shown in Figure 4.20. The mining of this ring continued for 55
minutes with no stops.
Figure 4.18: Slump test failed - Ring 90
60
|
Colorado School of Mines
|
4.4 Ring 103-132 Analysis
The operational data for rings 103 and 105 are presented in figures 4.26 and 4.27. The
geological profile for these rings is 60% clay and cohesive silt and 35% till deposits. The slump
determined from a ring 103 sample was 13.3cm (highest value up to this point) and the muck on
the conveyor belt was very watery (muck splashed out a few times from the muck chute). The
vane shear strength on the muck samples was 1.2 kPa. This value is also represnts the lowest
Vane shear strenght among all test results.
During ring 103 mining, screw conveyor gate G3 was controlled at 400 mm constantly.
In ring 105, the operator controlled the opening from the beginning of mining and gradually
increased the oepning from 300mm to 700mm. Gate G2 is also showing the same operation from
the start of ring 105. This can be the reason behind the availability of watery muck in the
chamber that is observed on the machine conveyor belt.
EC pressures decreased gradually throughout mining despite constant machine operating
parameters in ring 103. This will be discussed further in Chapter 5. SC pressures also decreased
during mining. was very low - 20,000 kN-m and is decreasing from time 00:30 forward to
the end of minin TgC.H The observed reduction in torque whithin two rings from 101 (30,000 kN-m)
to 103 was significant. EC-EPS shows a lower increase by the end of operation when compared
with previous rings.
EC and SC pressures remained constant during ring 105 mining. The differences in
operations included much lower F (80% lower than ring 103), lower BIR (10% vs. 20% in ring
103) and increased G3 opening (500 mm). Torque remained low at 20,000 kN-m and AR was
10% greater than ring 103.
As illustrated in figures 4.26a and 4.26b, in ring 105 is strongly associating with the
. Comparing , , and during time frame 0:1 T4S C- 0:23, it is evident that at time
θ0:S1C81, the chan θgSeCd1 fr θoSmC 233kN-m TS Cto 36kN-m with an increase in from 6.5 to 8 rpm and
from T 8S Cto 9.5rpm. At time 0:18, reduced back to 33kN-m. θSC1
θSC2 TSC
67
|
Colorado School of Mines
|
Figure 4.28: Ring 105 Operation a) , and , b) Tsc data
𝜽𝑺𝑪𝑺 𝜽𝑺𝑪𝑺
Ring 112 operational data is shown in Figure 4.29. The mining of this ring is continued
for 01:21 with a 20 minutes stop. The geological profile is about 45% in till deposits, 35%
cohesive clay and silt, and 20% in cohesionless sand and gravel based on geotechnical data
report. The muck GSD results revealed 6% gravel, 22% coarse sand, 24% fine sand (46% sand)
and 48% fines. This constitutes a 4% reduction in gravel, 11% increase in sand, and 4% decrease
in fines comparedg to ring 103. The slump was 4.6 cm and the conveyor muck was visibly
flowable but with less visible water than rings 103 and 105. Vane shear strength equaled 2.0-2.4
kPa, average muck moisture content was 26% (10% lower than ring 103), and the LL device taps
was 8. Ring 112 mining included one stop and restart. Excavation chamber and screw conveyor
pressures vary considerably during mining of this ring. Gate openings varied widely as did the
key operating parameters. This can also shows that muck operator is using SC gates to control
muck flow on the discharge chute. As shown in Figure 4.30, does not closely follow , AR
and F. TSC θSC
70
|
Colorado School of Mines
|
Figure 4.30: (a) AR, , , (b) F and
𝜽𝑺𝑪𝑺 𝜽𝑺𝑪𝑺 𝑻𝑺𝑪
Ring 113 is in identical geology and tunnel profile to 112. The operational data is shown
in figure 4.31 which is also very similar to ring 112 . Collectively, FIR+BIR+WIR ≈ 75%,
indicating that 7.5 parts of foam, polymer and water by volume is added to every 10 parts of
encountered ground.
Similar to ring 112, EC pressures vary considerably throughout mining. They drop
steadily during stoppage from 0:20 to 0:50, pick up considerably during restart at 0:50 and then
exhibit rapid by small fluctuations throughout the remainder of mining. Upon standstill,
pressures decrease steadily. SC pressures are abnormal (significant drop from SC3 to SC5 due to
gate G1 being tightly controlled throughout mining. The SC2 pressure does not dissipate linearly
with some significant differences between SC-EPS3 and SC-EPS5. Gate G1 is located between
these two sensors which can indicate that flow of much is intrupted by the smaller gate G1
opening.
72
|
Colorado School of Mines
|
Ring 119 operational is depicted in figure 4.34, with soil conditioning inputs shown in
figures 4.35 and 4.36. The mining of this ring is continued for 57 minutes with no stop. The
tunnel profile is about 40% sand and gravel, 20% cohesive clay and Silt, 30% in till deposits, and
10% in cohesionless silt and fine sand based on geotechnical data report. The Muck GSD test
was not obtained in this ring. Operating parameters varied considerably during ring 119 mining.
Significant PIR was employed at the outset of mining. The EC and SC pressures remained very
constant throughout mining. Fluctuations in pressure from 0:05 to 0:20 can be connected to
changes in volumetric flow rate (AR and SC rotation) that are described in more detail in
Chapter 5.
The operating data for rings 120, 122 and 124 are shown in figures 4.37, 4.38 and 4.39.
As the TBM progresses through these 3 rings, the cohesionless soil percentage grows, e.g., 40%
sand in ring 120 to 60% sand in ring 124 according to the GBR profile. GSD testing was not
performed to verify. As shown in the figures, G1 and G3 were controlled carefully throughout
mining of these three rings – typically around 300 mm opening and G3 around 100 mm during
ring 124 mining. EC pressures remained reasonably constant throughout mining with observed
‘chatter’ that corresponds to similar fluctuations in the main parameters, e.g., AR, F, . F was
considerably higher during ring 122 mining than in rings 120 and 124. TCH
Rings 131 and 132 consisted of 40% sand and gravel, 20% clay and cohesive silt, 20%
till and 20% cohesionless silt and fine sand per the GBR. mining is continued for 01:17:00
(hh:mm:ss) with no stop during mining (see figure 4.40). Muck sampling in each ring revealed 8-
11% gravel, 53-54% sand, and 36-38% fines. Slump values were 5.1 cm and 5.8 cm,
respectively. The muck was visibly watery with material splashing out of the chute occasionally.
Vane shear strengths were found to be 2.6 kPa and 2.8 kPa. LL device taps were less than 10 but
are not a good indicator in cohesionless soils.
Rings 131 and 132 were mined with different operating parameters (see figures 4.40 and
4.41). Magnitudes of F were twice as high in 131 than used in 132. AR were generally similar as
were levels. Both AR and were very choppy throughout. Gates G1 and G3 was
contro TllCeHd tightly, with G1values TrCaHnging from 200-300 mm and G3 values100-200 mm. EC
76
|
Colorado School of Mines
|
CHAPTER 5 – CHAMBER PRESSURE
5.1 Introduction:
In this chapter the influence of specific EPB operational parameters are investigated.
Each of these parameters can influence the muck flow and material characteristics. Operational
parameters are individually discussed and their influence on chamber volume and pressure is
explained. Of particular interest to this analysis is relating operational parameters to observed
changes in excavation chamber pressure. In the last part of this chapter, an analytical estimation
of the volume changes and related pressure changes is provided to estimate material
compressibility.
5.2. Influence of Cutterhead on Chamber Pressure
Although CH design is very critical to cut and flow material through the openings and
into the excavation chamber, it is found that CH itself influence the chamber pressure and this
may alternate due to muck characteristics. Shear strength of the material and the adhesion
between the soil and steel surfaces (face and spokes) are found to be influential.
Normally, during standstill, the operator maintains F to keep the ground stability (see
figure 5.1a). Mining operation is started by rotating the CH (see figure 5.1b). When operator set
the (pre-determined value), it takes around 30seconds for CH motors to reach this
targ θetC Hva =lu 1e .0f rro pm
m
0.0rpm. In this time frame, the AR is zero and the SCs are not rotating.
Varying among different rings, after 30 and 60 seconds, the operator starts SC1 and then SC2. In
this time period, the change in chamber pressure is due to CH rotation (see figure 5.1c, dashed
box).
Figure 5.2 shows EC level 4 pressure changes in the first 30 seconds of mining for rings
70-150. The increase in pressure is ranged between 0.2 to 0.4bar. Interestingly, the change in
pressure is higher before ring 100, and the maximum pressure change is between rings 82-96 at
rings 89 and 91. The on-site lab test on the rings 82-91 shows a higher LL number of taps, failed
84
|
Colorado School of Mines
|
shows left side pressure changes (R4 EC-EPS is only used for simplification) in red and right
side pressure change (L4 EC-EPS) in black color. For instance, in ring 77, pressure in the right
side is higher than the pressure in the left side of the chamber. Although the pressure increase
due to ramping up of CH is the same, during counterclockwise rotation of CH, pressure on the
right side is found to be higher than the left side. This will be discussed more in chapter six.
Figure 5.2: CH influence on chamber pressure
5.3. Influence of Thrust Force on Chamber Pressure
Force equilibrium of the EPBM shield can be defined in Equation 5-1 thrust force F
equals the summation of the chamber pressure force ( ), the cutterhead force ( , the shield
friction force ( , and the trailing gear force FTG. FAPssuming FTG and Ff r FeCmHa )in constant
locally (within Fmf)inutes), the change in F is reasonably approximated by equation 5-2. These
equations are valid during standstill and during excavation where inertia forces are negligible.
(5-1)
𝐹 = 𝐹𝑃 + 𝐹𝐶𝐶 +𝐹𝑓 +𝐹 𝑇 𝑇 (5-2)
𝑑𝐹 = 𝑑𝐹𝑃 +𝑑𝐹𝐶𝐶
86
|
Colorado School of Mines
|
Figure 5.3: EPB mode schematics and force equilibrum
According to Hooke’s law, , where k is the stiffness of a spring under force F. If
the F is applied to two springs, each of them will absorb portion of that force based on the
F = −kX
stiffness. By applying this law to the simplified EPB mode equilibrium 3, it changes as below:
(5-3)
𝐾𝐹−𝐶𝐶
𝐹𝐶𝐶 = 𝐾𝐹−𝐶𝐶+𝐾𝑃
(5-4)
𝐾𝑃
𝐹 𝑃 = 𝐾𝐹−𝐶𝐶+𝐾𝑃
Soil conditioning influence the compressibility of chamber material and stiffness of
muck is inverse of compressibity. With increaseing F, stiffer material takes higher forces when
compared to muck with lower stiffness.
Mining starts with increasing F and which flows the cut and scraped foamy soil
toward the openings of the CH and into the ch θaCmHber. Without the , muck would not flow out
of the chamber. By increasing , AR is also changing. This θsSuCggest that the is also
influence the AR and this is di θsScCussed before when change in the gate G1 ope θrSaCtion was
inevstigated in ring 112-131 (Chapter four).
87
|
Colorado School of Mines
|
Much in the chamber is treated with a specific soil conditioning during minig. When the
CH scraped and cut the in-situ soil formation, it cannot flow them into the chamber when SC
outflow rate is low. By this phenomenon, conditioned material is compressed and when they
reached their maximum compressibility, Although F is constant, AR will drop significantly. This
manifestation can provide cloggage in front of the CH.
Figure 5.4 presents , machine operation data, EC-EPS, and SC-EPS for ring 67. The
EC-EPS level 3, 4 in the ch FaCmHber and the SC-EPS 2, 3, and 5 are only shown for simplification.
In this ring, operator controls chamber pressure through , and F. In each decrease or
increase of , the chamber muck is mechanically influ θenSCc1ed.
θ
TSCh2e discussion here focuses on
the time fram θSeC 01:05-01:36.
Figures 5.4a and 4.79b show an increase in from 0.0 to 6.0 rpm and from 0.0 to
8.0 rpm from 1:05 to 1:06. AR is also increasing du θrSiCn1g this period. Following th θiSsC,2 F is initially
set to 135,000 kN, remains constant for 5 minutes and then it is decreases to 123,000 kN (at
1:12). The AR is generally constant at 30mm/min and the EC-EPS decreases steadily by
approximately 0.4 bar. With a significantly greater , more material is discharged from the
chamber than is entering through the cutterhead. Wh θeSnC the outflow rate is increased, chamber
pressure decreases and consequently, material in the chamber expands (decompresses) due to the
presence of foam under pressure (in case of a proper soil conditioning). This muck expansion
decreases the chamber pressure and decreases the chamber material stiffness. A decrease in
chamber material stiffness would result in the chamber soil absorbing a lower share of F.
Consequently, the CH contributes more and increases as shown in Figure 5.4b.
FCH
Figure 5.5 (page 89) presents EPBM data from ring 78 where EC and SC pressures vary
throughout mining. The analysis here focuses on two time periods: (1) t = 0:10 to 0:16 where EC
and SC pressures steadily but significantly decrease, and (2) t = 0:16 to 0:23 where EC and SC
pressures steadily increase. During the first time period ramps up signficantly and quickly
(4-7 rpm from 0:10 to 0:11) and remains at 7 rpm until 0: θ16SC. Chamber pressure decreases during
this time frame due to increasing discharge of muck (and chamber decompression). The decrease
in muck stiffness that results in turn causes Fch to increase as observed. At 0:16 when are
decreased quickly down to 4 rpm, the behavior reverses. The chamber pressure increases a θnSdC Fch
88
|
Colorado School of Mines
|
5.4 Chamber Pressure Fluctuation Induced By Chamber Muck Volume Alteration
The volumetric flow rate of soil into the chamber through the cutterhead (inflow rate) and
discharge volumetric flow rate through the screw conveyor (outflow rate) can be estimated and
analyzed in comparison to pressure.
The inflow material volume through CH openings is calculated using AR as follows:
(5-5)
2
π∙(DCH)
V in = � 4 × (ARt2−ARt1)× α�× ∆t+ VB
3
Vin = Theoritical volume of material into t he chamber(m )
ARt1 = Advance rate at time 1 (mm/min)
ARt2 = Advance rate at time 2 (m m/min)
DCH = Cutter head diameter (m)
∆t = t2 −t1
VB = Bentonite Volume
α = CH opening ratio (%)
The change in material outflow volume through SC-1 is calculated using SC1 as follow:
(5-6)
2
π∙(DSC1)
V out = � 4 ×((θSC1)t2 −(θSC1)t1)× (aSC− p)�×∆t
3
Vout = Theoritical Volume of material leaving the screw conv eyor(m )
(θSC1)t1 = Screw conveyer1 rotation speed at time 1 (rpm)
(θSC1)t2 = Screw conveyer1 rotation sp eed at time 2 (rpm)
DSC1 = Screw Conveyer 1 diameter (m)
∆t = t2 −t1
aSC = Flight Length
P = Blade Thicknes
91
|
Colorado School of Mines
|
the chamber muck is compressing (to accommodate the increase). The compression of muck
increases the excavation chamber pressure. The behavior is quiet similar in reverse in that an
increase in net outflow volume rate leads to muck decompression and chamber pressure
decrease.
The chamber pressure relationship with flow rate is further analyzed for ring 77
(primarily cohesive clay and silt, till deposits) in the following plots. In Figures 5.7 - 5.84, ten
zones of chamber pressure change are identified. Only L4 chamber pressure is reported for
clarity; however, all EPS readings exhibited similar behavior.
During each of the ten highlighted excavation chamber pressure changes, the volumetric
flow rate is changed as shown in Figures 5.7e-5.7e. Recall that Qin-Qout is calculated from the
AR and responses (see figures 5.7b-5.9b) via the equations above. Figures 5.7e-5.9e reveal
varying θvSoClumetric flow rate behavior where Qin-Qout is positive for some stretches and
negative during other stretches. The magnitudes of Qin-Qout as well as it’s time rate of change
are also variable.
It is evident from these Figures that a positive Qin-Qout yields an increase in chamber
pressure and a negative Qin-Qout yields a decrease in chamber pressure. The Figures also show
that chamber pressure continues to increase/decrease after the volumetric flow rate change
plateaus. This is evident in highlighted zones 1, 3 and 5 in Figure 5.7 where sharp monotonic
decreases in Qin-Qout followed by constant negative Qin-Qout are synchronous with steady,
monotonic chamber pressure increases.
It is logical that the magnitude and rate of the volumetric flow rate change would be
proportional to the resulting magnitude in chamber pressure change. More specifically, a
magnitude of Qin-Qout over a certain time interval yields a magnitude of ∆Vin-∆Vout. Figure
5.10 presents the magnitudes of these changes ∆Vin-∆Vout vs. ∆P for the 10 highlighted zones
in ring 77.
93
|
Colorado School of Mines
|
CHAPTER 6 - VERTICAL AND HORIZONTAL CHAMBER PRESSURE
DISTRIBUTION
6.1. Introduction
EC pressure in large diameter machines varies vertically and horizontally. These
variations can be due to operation and material characteristics. In this chapter, the vertical
excavation chamber pressure distribution and related pressure gradient, as well as horizontal
pressure distribution are discussed during mining, after mining, and in few rings after long
standstills.
Muck density testing was performed under atmospheric pressure (hereafter as . The
vertical chamber pressure gradient (hereafter as ) is also calculated during stan γdatsmti )ll and
plotted versus atmospheric density test results. Reca ∇llPing from chapter three, although geology is
consist of several different layers of material, GSD test provide an insight into change in muck
grain size from ring 75 to 145.
6.2 Chamber Pressure Distribution and Gradient
To investigate vertical and horizontal pressure distributions, a set of plots are developed
as follows. In Figure 6.1, earth pressure sensor (EPS) data at six elevations on the left side of the
excavation chamber (looking forward) are plotted. The top pressure sensor is 2.5m from the
crown and the bottom sensor is 2.0m from the invert. These values are gage pressure readings,
i.e., pressure above atmospheric (1 bar). These data reflect total lateral earth pressure that is
theoretically equal to pore fluid pressure plus effective lateral earth pressure (Equation 6-1), and
related to vertical effective stress through the coefficient of lateral earth pressure K (Equation 6-
2). The soil is assumed to be partially saturated due to the presence of foam-induced air bubbles,
and therefore, pore fluid pressure u reflects some combination of pore air and pore water
pressures.
105
|
Colorado School of Mines
|
dσ dσ′ du
x =K z + =Kγ′+γ
f
dz dz dz (6-3)
γ′=γ−γ
f (6-4)
dσ x =K( γ−γ ) +γ =Kγ+γ ( 1−K)
f f f
dz (6-5)
Estimating K of the chamber soil is very difficult. The scraping and ingesting process
destroys any soil fabric and therefore K0 conditions no longer prevail. There is no literature on
the topic of K resulting from mixed soils. If the chamber soil is well-conditioned such that the
shear strength is negligible and the effective stress is relatively low, the coefficient of lateral
earth pressure can be assumed = 1.0 (isotropic, fluid-like). In such a case, these measurements
also reflect the vertical total earth pressure and the gradient reflects the total unit weight of the
conditioned chamber soil γ.
In Figure 6.1, the inverse of the slope is the gradient of the total lateral earth pressure in
the chamber, which in this case is 0.19 bar/m (19 kPa/m). The on-site density test under
atmospheric pressure is 1885 kg/m3 for this ring (18.85 kPa/m). The density under pressure is
higher than the density under atmospheric as expected. Figure 6.1 illustrates the significant
difference in the lateral (and presumably the vertical earth pressure) from the crown to the invert
in the chamber. Projecting the gradient of the data, the crown chamber pressure of 0.8 bar is 3.4
bar less than the invert chamber pressure of 4.2 bar. The data also indicate that the chamber is
either full of muck or there is an air gap at the crown. In both cases, excavation chamber is fully
pressurized (no atmospheric conditions at the crown).
This approach of estimating the vertical pressure gradient was used along 50-150 ring
alignment (see figure 6.2). Figure 6.2 shows that gradient is correlated with fines and granular
grain size distribution data. With increasing fines content ∇aPnd decreasing granular content of the
muck, and decreased (ring 75-90). With increasing granular content and decreasing fines
content γ, atm and ∇ P increase (rings 90-145).
γatm ∇P
107
|
Colorado School of Mines
|
The chamber right side pressure gradient ( ) and the left side pressure gradient ( )
follow the same trend as expected but not the sam ∇eR Pvalue. Before ring 101, is higher ∇thLaPn
(e.g, ring 77 and 78) and sometime is lower than (i.e. ring 67 and
∇
L8P1) or equal
∇(eR.gP., ring 75, 85 and 99). However, afte ∇r LrPing 101, is
∇
aRlPways higher than and in so ∇mRPe
cases the difference is significant ( ring 122, 124, 131 ∇, LaPnd 133). ∇RP
(a)
(b)
Figure 6.2: (a) Geology and GSD; (b) Muck atmospheric density vs right and left chamber pressure gradient
One finding of chamber pressure analysis is that left and right side pressures and their
gradients are different, similar to the findings of Bezuijen A., Joustra J.F.W., Talmon A.M.,
Grote B., (2005) for a project in the Netherlands. One example during ring 77 mining is
presented in Figure 6.5. Here, the right side pressures are higher than the left side pressures at all
EPS levels except level 1 that is 2.5 m from the crown. The horizontal pressure difference at the
EPS level 6 is also negligible. It is worth noting that the horizontal distance between EPS levels
108
|
Colorado School of Mines
|
varies (see figure 6.3). The level 6 EPSs are very close (7.6 m apart) and are strongly influenced
by the screw conveyor.
The increased right side pressures in Figure 6.3 are coincident with counterclockwise
cutterhead rotation and clockwise agitator rotation. While the aggregate gradients are similar, the
left and right side pressure gradients vary locally in the chamber. Recalling from chapter two,
EC-EPS locations are dimensioned in Figure 6.3. The scale and position of the cutterhead,
agitator and bulkhead EPSs is important when evaluating these data (see figures 6.4).
(a) (b)
Figure 6.3: (a) EPS horizontal distances, (b) EPS vertical distances
The causes of the horizontal pressure difference are multiple and the process of chamber
mixing is complex. The horizontal pressure difference is influenced by the cutterhead and center
agitator rotation rates, shear strength of material, and adhesion between the muck and the steel
surfaces. The cutterhead pushes the muck in the direction of rotation and above all, material flow
in the chamber is dominated by the cutterhead rotation direction. As a result, if the chamber is
not full of soil, material in one side of the chamber would be higher than the other side causing
an increase in observed pressure on that side. A center agitator is used in larger EPB machines in
order to help mix the material in the chamber The thrust cylinders and articulation jacks can also
affect the pressure in tight curves. Material shear strength, slump (flowablity quality), stickiness
109
|
Colorado School of Mines
|
and adhesion can also influence the effectiveness or machine mechanical forces ( i.e. cutterhead
and center agitator).
(a) (b)
Figure 6.4: Excavation chamber; (a) Cutterhead assembling process; (b) Mixing chamber with center
agitator in red, static bars and cutterhead legs all attached to bulkhead
This horizontal chamber pressure difference changes during mining and also during
standstill. To investigate this difference the horizontal pressure for different rings is explored
hereafter. Figure 6.5 shows that sensors L1 and R1 show similar readings. Sensors L2 and R2
exhibit 0.65 bar difference, demonstrating the material is either lower than sensor elevation 1 or
slightly above it (L1 and R1). Figure 6.5 illustrates the right and the left side pressure distribution
of ring 77 during mining. The pressure gradient on the left side is 0.20 bar/m and on the right is
1.96 bar/m.
By continuing the pressure gradient slope trendline up to the chamber crown it shows 0.7
bar pressure. This suggests that either the chamber is full of material or there is pressurized air on
top of the material.
110
|
Colorado School of Mines
|
Figure 6.5: pressure distribution on ring 77 during mining on right and left side
6.3 Vertical and Horizontal Pressure Distribution of Ring 77
The pressure gradient during mining at time 0:25 of ring 77 is shown in Figure 6.6a.
is counterclockwise looking forward and is clockwise. Chamber pressure is higher on
t θhCeH r ight side than on the left side of the chamb θerC.A T he pressure distributions on the right and left
side of the chamber during standstill are the same in this ring, indicating that without the
mechanical influence of the machine, the material is level at the top.
Figure 6.6d shows the pressure gradient 1:10 after mining ended. The gradient remained
unchanged during this time. Figure 6.6 indicates that chamber pressure has dropped 0.6 bar
similarly on the right and left sides. Figure 6.8 shows a rapid drop from time 1:16 to 1:20 and a
gradual long lasting drop in pressure from time 1:20 to 2:30. At time 1:16, CH and CA are
stopped by the operator (Figure 6.7). Right after that time, the first drop (rapid decrease) in
pressure occurs and is due to the absence of mechanical influence of CH. This change in pressure
111
|
Colorado School of Mines
|
As shown in chapter five, the chamber pressure fluctuates during mining due to muck
volumetric flow rates and compression/decompression of the material. In either case, the height
of muck in the chamber may change (if an air pocket exists at the chamber top). To investigate
this, pressure changes between level one and two (local gradient) on the right and left sides
(referred to as ) were divided by the overall gradients and (see figure 101).
∆P ∇RP ∇LP
Figure 6.9: Height of right and left side chamber muck level
can be different on the right and left sides of the chamber due to local muck
conditio ∆nP. ⁄F ∇ivPe different scenarios can happen between EPS level one and two. The first scenario
is that muck level in the chamber is located between level one and two (
). The second senario is that calculated is in range
11.8m ≤
∆P⁄∇, Pbu <t
a 1c 4t .u 5a mlly material is locally compressed and exh ∆iPb ⁄it ∇s Pa local gradie 1n 1t .8h migh ≤er ∆thPa ⁄n
∇
Pch <am 1b 4e .5r
m
. In
the third scenario, muck fills the space between the two sensors ( ). The f ∇oPurth
scenario is when is in range , but material is locally
H = 14.5m
decompressed and ha ∆sP a
⁄
∇loPcal gradient lowe 1r
1
t .h 8a mn ≤the
∆
cPh ⁄a ∇mPb <er 14.5. T mhe last scenario is when the
local gradient of material between EPS level one and two is hig ∇hPer than chamber (
). The first and third scenarios are the most probable when material has
∇
aP c ∆oPn ⁄si ∇stPen >t
gradient throughout the chamber (no local compression or decompression). Hereafter, material of
14.5m
different rings is assessed to investigate whether muck is locally influenced or not.
As discussed in chapter two, port 28 is the only open port in the chamber that shares foam
pump P8 and foam generator FG8 with port F8 and F32. F8 is at the CH and is open during
mining. F32 is located in the SC1 and is closed during mining. Recall from chapter two that
foam flow rate is a combination of air and foam liquid flow rate ( ). The flow rate
QF = QA+QL
114
|
Colorado School of Mines
|
As discussed in chapter five, during each ring, is the first parameter that is increased,
SC1 and SC2 follow, then AR is ramped up along wi θthC HF and other operational parameters. To
illustrate, Figure 6.11 presents the initial five minutes of mining for ring 77. Figure 6.11a shows
that F is increased from 55,000 to 65,000kN during time frame 0 to 40 sec. In this time frame,
the AR is zero meaning that the EPBM is not displacing. CH is started at t = 0 and is ramped
up to -1.2 rpm (counterclockwise) by t=40 sec. Figure 6.11a shows that
θ
CiHs decreases
significantly during the ramp up. TCH
For the initial 90 seconds, CH is rotating and AR, SC1, and SC2 are zero. As it is
illustrated in figure 6.11c, in this time frame chamber pressure increased 0.5 bar on the right side
and 0.35 bar on the left side. Figures 6.11e, 6.11f, and 6.11g provide some insight into how muck
behaves during this time frame. Figure 6.11f indicates that during ramp up of the CH rotation (in
counterclockwise direction), increases from 0.17 to 0.20 bar/m. also increases from 0.15
to 0.17 bar/m. After 20 sec,
∇
wLPhile is still increasing (muck c ∇oRmPpression), begins to
decrease and the muck level on the l ∇efLt Pside remains constant. The left side chamb ∇eRr Pmuck level
is likely lowered due to compression as the cutterhead rotated, and it is likely that material on the
right side decompressed and the right side muck height increases.
Figure 6.12 presents the vertical pressure gradients before (t = 15 sec) and after (t = 40
sec) the center agitator is operated. At t = 15 sec, it is evident that material density on the right
side of the chamber (red dashed line) is locally compressed and decompressed. As explained
earlier, cutterhead rotation compresses material between levels one and two, and decompresses
the material between levels two and three on the right side of the chamber. This results in locally
compressed and decompressed muck.
With CA operation, a more consistent pressure gradient is provided (comparing red and
black dashed lines). The muck on the left side of the chamber remained under compression with
no inconsistency except from level five (R5 and L5). While there is no SC operation during this
time, this compression can mainly be caused by back pressure of the SC muck (muck self-
weight).
116
|
Colorado School of Mines
|
Figure 6.12: Verical pressure gradient ring 77 time 00:00:15-00:00:40
SC1 and SC2 rotation are initiated at 90 sec and 110 sec, respectively (Figure 3.11c). At
this point, the volume of material in the chamber is dramatically influenced, as is illustrated in
Figure 3.11e with a negative chamber muck volumetric flow rate. This implies that more
material is leaving the chamber than entering. This chamber material reduction continued for 75
sec, and the chamber pressure decreased 0.4 bar.
Chapter five explored the parameters that influence AR. At t = 3 min, there is an increase
in thrust force F from 65,000 to 120,000 kN at t = 5 min. Due to this increase and the rotation of
CH, SC1 and SC2, AR increases dramatically from 0 to 36mm/min over the course of 15 sec.
As discussed in Chapter 4, cutterhead torque is influenced by penetration rate (AR/ ).
Immediately after AR increases, begins to
T
CstHeadily increase. With an increased AR, m θCuHck
flow rate into the chamber incre TaCsHes, as shown in Figure 3.11e. Interestingly, this positive
volumetric flow rate does not yield compression-induced chamber pressure increases within the
5 minutes of data presented. For the first 90 sec of mining, there is a steady increase in right and
left pressure gradient. The gradient flattens as the SC begins. Foam injection initiates at t = 2 min
with an average foam flow rate of 200 L/min varied among 20 pumps (FIR does not report until
AR increases by definition). It is plausible that foam injection is influencing the gradient,
particularly the decrease observed on the left side.
118
|
Colorado School of Mines
|
Ring 77 mining is stopped at 1:17 (Figure 6.14). increases from 0.17 to 0.20 bar/m.
During standstill and both decreases gradually f ∇oLr Pabout 0.01bar/m. The data suggests
that the muck heig ∇hRt Plevels ∇oLfPf at around 14 m on the left side and 13.8 m on the right side.
Figure 6.15: Ring 77 operation, time (00:11:00 - 00:23:00)
The horizontal EC pressure differences are presented for ring 77 in Figure 6.16, together
with pressure gradients and estimated muck heights. The horizontal pressure differences (HPD)
are calculated based on right side reading – left side reading at the 6 sensor levels. For the first
53 minutes of mining when CH rotation is counterclockwise, all HPD levels are positive. When
the CH rotation is clockwise (t > 0:53), HPD readings are negative. The HPD magnitudes during
counterclockwise CH rotation are noticeably greater than these during clockwise CH rotation.
The reason for this is unknown.
Level one HPD ( ) fluctuates between 0.0-0.2 bar and does not change polarity
with CH rotation direction.
H
T Ph Dis1 supports the existence of an air pocket and suggests that level 1
EPSs are influenced equally by the air pocket, independent of CH rotation direction.
HPD2
121
|
Colorado School of Mines
|
exhibits the greatest values, averaging 0.5 bar during counterclockwise CH rotation and -0.2 bar
during clockwise CH rotation. HPD in these two levels are important to judge the muck
condition near to crown level and possible air pressure. and provide the pressure
difference at the springline of the chamber. Figure 6.16 Ha Psh Do3ws tha Ht
P
d Du4ring mining,
and they follow the same trend. Condidering a constant vertical Hpr Pe Dss2ur >e
g Hr Pad Di3en >t o Hn
P
t Dh4e >lef Ht h Pa Dn6d side and a locally changing gradient on the right side, it is evident that
HPD will be locally changing. One can claim that muck at level 1 and 2 can compress and
decompress more easily than levels 3, 4, 5, and 6 due to lower pressures and an air gap.
fluctuates around zero and is mainly influenced by SC1. When the EPBM transitions Hf Pro Dm5
mining to standstill, HPD levels change but do not all reach zero. trends towards zero
during standstill yet actually increases slightly during standstill. H PD2
HPD1
Figure 6.16: Ring 77 (a) HPD, (b) Right and left chaber pressure gradient, (c)possible muck level height in the
chamber
122
|
Colorado School of Mines
|
gradient is varying locally in the chamber. The HPD near SC1 is influenced by the operation
significantly. At times 0:41 and 0:45, the gate G3 opening is increased (Figure 6.17b).
Consequently, the SC-EPS 5-10 readings decrease significantly. This change in operation
influences the volumetric flow rate and EC pressure. increases at these mentioned points
in time. This demonstrates that changes in gate openin Hg
P
l Dea5ds to muck compression on the right
side at level 6. Moreover, ranges between -0.4 and -0.9 bar, indicating that muck on the
right side of chamber invert H(a Pr Dou6nd the SC1 opening) is compressed significantly during mining
(Figures 6.22a and 6.22c). Figure 6.22b shows the EC pressure distribution during the short
standstill on the second stop. It is evident that due to SC stoppage, muck compression is
alleviated.
In the last part of ring 67 mining, CH rotation is counterclockwise. By comparing
gradients during clockwise CH rotation (see figure 6.22c) and counterclockwise CH rotation (see
figure 6.22d), one can see a significant change. changes from 0.21 to 0.17 bar/m and
changes in the reverse fashion, from 0.17 to 0.2 ∇1R bPar/m. During standstill and the absence
∇
LoPf
rotational influence, chamber muck is counterbalancing the earth pressure. Interestingly, after
mining is completed the increases (see figure 6.22c). As illustrated in Figure 6.22a, EC-EPS
reaches 0.2bar/m immedia ∇tPe ly. After 4 hours of standstill, decreases 0.03 bar/m, from 0.20 to
0.17 bar/m; however, remains unchanged. Interesting ∇lLyP, right side EPS level 5 and 6 data
suggest significant loca ∇l RcPompression.
Figure 6.21: Pressure drop during standstill of ring 67
126
|
Colorado School of Mines
|
level. Figure 6.24c shows a significant drop in pressure on the right side during the stop in
mining and also after mining is completed (standstill). Figure 6.24d shows that this drop is due to
muck levelling off at the top. During standstill, HPD gradually decreased 0.4 bar at levels 2 and
3. During the decrease in HPD level 3 and 4, and reach constant values from 1:25
onward. This suggests that and between l ∇eRvPel 2 and
∇
L4P are locally changing.
∇RP ∇LP
Muck weight in the SC is negligible compared to the muck weight in the chamber. It is
hypothesized that the self-weight of muck in the chamber compresses the material in the SC.
However, Figure 6.24d indicates that SC-EPS2-10 are decreasing (if SC muck were compressed,
these pressures should have increased). Exceptionally, increased after mining. It is
possible that the muck is compressed near the SC1 inlet ( Hb Pet Dw5een levels 5 and 6). Moreover,
counterclockwise CH rotation increases the muck stiffness (lower compressibility) on the left
side of the chamber. Although L5, L6, R6 EC-EPS decrease very slightly (Figure 6.24c) during
standstill, R5 EC-EPS decreases significantly. One can conclude that muck with lower stiffness
(higher compressibility) is absorbing higher portion of SC back pressure. In this case, muck
between level 4 and 5 on the right side has a lower gradient (higher compressibility). This
phenomenon is depicted in figures 6.23a-6.23c.
Figure 6.23: Ring 81 drop in chamber pressure during standstill
128
|
Colorado School of Mines
|
Figure 6.25: Ring 81 chamber pressure assessment
Figure 6.26 depicts the operational data during mining stoppage (left) and during the
subsequent restart (right). Figure 6.26L(b) shows that AR and SC both stop at the same time.
Counterclockwise CH rotation creates higher pressure on the left. With the absence of AR and
SC operation (19 minutes, 30 seconds), left side and right side pressures increase. This increase
shows the influence of CH as the counterclockwise rotation creates higher compression on the
left side. During the CH rotation ramp down (beginning at t = 20 min), the gradients on both
sides decrease. After CH and CA stoppage, and tend toward the similar value of 0.17
bar/m. Interestingly and unlike ring 67 and 77 ∇, RmPuck in
∇
tLhPe chamber experiences a slight increase
in gradient during stoppage.
To rebuild the lost pressure throughout five hours of standstill, the operator turns the
cutterhead and advances the machine approximately 24 mm over one minute with = 0.6 rpm
(clockwise). SC and CA are not initiated. This introduces material into the chamb θeCr Ha nd results
130
|
Colorado School of Mines
|
Standstill was continued for 15 hrs during ring 81. Figure 6.28 presents chamber pressure
distributions during this standstill, immediately after the t = 5:30 machine operation and eight
hours later. The superposition of these two distributions shows a 0.6 bar maximum chamber
pressure drop. Although left and right side EC-EPS readings are different, their drops are very
consistent on both sides.
The gradients suggest local decompression between levels 3 and 4. One hypothesis is that
foam has disintegrated after 5 hours such that the muck density increases and pressure in the
chamber is lowered. Another hypothesis is that muck self-weight causes compression and
chamber pressure is decreased. Either of these hypothesis should decrease . In this ring, in
increased on the left side of the chamber and decreased very slightly on
∇
Pthe right side. ∇APs a
result, neither disintegration of foam nor muck compression from self-weight is likely. A
reasonable explanation for this phenomenon is that muck exhibits local gradients due to lack of
flowability.
A time dependent compression and de-compression is evident by investigating this
influence throughout time of standstill. Muck with lower flowability (lower viscosity - it is not
measured among on-site tests) like this ring provided more visible inconsistency in pressure
distribution both vertically and horizontally.
Figure 6.28: Pressure drop during standstill
133
|
Colorado School of Mines
|
6.6 Vertical and Horizontal Pressure Distribution of ring 87, 90, 91, and 101
Figure 6.30 presents a snapshot of pressure distribution at time 0:30 during excavation of
rings 87, 90, 91, and 101. It is evident in all four plots that EC-EPS R5 is inconsistent. This
phenomenon was discussed previously in this chapter. Unlike previous rings, in these rings the
R5 reading is lower than the R4 reading above it (L5 and L4 does not exhibit this inconsistency).
Test results from rings 87, 90, and 91 indicate high vane shear strength and zero slump. The
geological profile between rings 87 and 101 is mostly clay (CL and CH layers) and the tunnel
face for these rings is consistent (Figure 121). 80% of the tunnel face in this location is till
deposits and cohesive clay and silt soil units. The GDR indicates these soil units have high
clogging potential.
Figure 6.29: Tunnel profile geology based on BDR
Figure 6.30 shows that the chamber material exhibits local compression and
decompression. It is likely that, due to high shear strength and stickiness (high adhesion), the
material would not flow appropriately. A lack of flowability can provide gaps in the chamber or
highly decompress material. The gradient between levels 2 and 4 is very low on the right side
compared with the left side. A possible explanation is that CH pedestals and adhesion between
CH steel surfaces and muck are greatly decompressing the muck on the right side.
134
|
Colorado School of Mines
|
Figure 6.33: Ring 103 chamber pressure assessment
The change in pressure distribution and local gradient are shown in Figure 6.34 during
mining. The left side gradient exhibits a concave shape with depth while the right side gradient is
more constant with depth. Three scenarios are possible to explain this curvature. In the first
scenario, due to availability of water in this ring and very high slump comparing to previous
rings, one can claim that counterclockwise CH rotation is conveying and compressing muck on
the right side of the chamber. Assuming such a phenomenon, Figure 6.34 shows that loose soil-
bentonite-water-foam mixture on the left side between levels 2 and 3 is available and it is
providing a lower gradient. The second possible scenario is that water is available on top of
material on the left side. Figure 6.35 illustrates a decrease in local gradient between levels 2 and
3. In the third scenario, it is possible that heavier particles are sedimented during operation due to
the soil-bentonite-water-foam mixture with lower vane shear strength and high slump providing
138
|
Colorado School of Mines
|
6.8 Vertical and Horizontal Pressure Distribution of ring 105:
As discussed in chapter four, ring 105 is located in the same geology as ring 103, and
material was very flowable. The CH is rotating clockwise and the CA is rotating
counterclockwise (see figure 6.37a). Due to the presence of water, gate G3 was initially closed
and opened gradually from 0 to 500 mm by the operator. Gates G1 and G2 were for the most part
fully open. This control over gate G3 produced some fluctuations on SC-EPS5-10 readings as
discussed in chapter four. The combination of operational parameters such as , , and gate
G3 opening along with AR, changes the flow rates of material through the c θhsacm1b θesrc (2see figure
6.38b).
Changes in EC pressures with muck volumetric flow rate changes are not clearly evident
in this ring. The pressure gradient is clearly showing these changes (see figure 6.38c, i.e. time
0:14, 0:18, and 0:22). It is evident
∇
thPat with a slight increase in and (see figure 6.37a),
decreases on both sides of the chamber (time 0:14). Similarly θ,s ca1t time θs s0c:218 and 0:22, with a
d ∇ePcrease and an increase in and , increases and decreases consecutively ( shows
a more clear change). θsc1 θsc2 ∇P ∇RP
Due to clockwise cutterhead rotation, the horizontal pressure differences are negative.
Interestingly, HPD decreases from top to bottom, from 0.6 bar difference at level 2 to 0.1 bar
difference at level 6. The decrease in HPD is very consistent from top to bottom of the
excavation chamber with 0.1 bar per level ( ). At all
levels except 1 and 6, HPD is increasing the s Ham Pe
D
2am >ou Hn Pt
D
fr3o >m Hti Pm De 40 >:03
H
t Po
D
05:1 >4. HA Pf Dte6r mining,
the material height rapidly levels off with HPD = 0.3 bar at all levels. Figure 6.38d illustrates that
muck height after mining is just below level 1. There is an exceptional increase in right after
the end of mining, and is most likely due to the absence of CH rotation. This in ∇crLePase is very
abnormal and is about 0.03 bar/m which is remarkable when compared with other changesin
gradient. Figure 6.38 shows that the rapid increase in is due to CH rotation stoppage that
allows the chamber material that had been locally influe ∇nLcPed by the CH rotation reach its actual
gradient.
141
|
Colorado School of Mines
|
Figure 6.38: Ring 105 chamber pressure assessment
Figure 6.39a illustrates the influence of and on EC pressure while AR and gate
G3 are constant. is not changing and the pr θesscs1ure de θcsrce2ases 0.2 bar. On the left side, is
decreasing. Intere ∇sRtiPngly, this decrease dissipates gradually from level 3 up toward the t ∇oLpP of
material. As discussed earlier, clockwise CH rotation compresses muck on the left side of the
chamber (Figure 6.39a). With increasing , the flow regime and shear rate is changing, and as a
result, material with lower shear strength θasncd less stiffness (lower compressibility) is expected to
initiate the flow or increase the flow from previous regime. Due to CH rotation, material on the
left and right side of the chamber may exhibit different viscosity. Figure 6.40b shows time
intervals of change from light orange to dark orange. It is evident that pressure on the right side
is consistently dropping without gradient change overall or even locally. However, on the left
side this drop in pressure is not consistent with gradient that locally changing between level 1
143
|
Colorado School of Mines
|
6.9 Vertical and Horizontal Pressure Distribution of ring 114 and 119:
Ring 114 and 119 operations were very similar to ring 103, however, the muck
characteristics of ring 114 are very different. Ring 103 is located in a transition zone from two
completely different geological profiles (Figure 6.41). Rings 114 and 119 are located in a
geology zone consisting of silt and sand. Table 6.1 provides information regarding soil
conditioning of rings 103, 114, and 119. Bentonite was injected in ring 103 but not rings 114 and
119. Polymer was injected during ring 119 only. Foam injection was similar for rings 103 and
114 and dropped considerably for ring 119.
Figure 6.41: Tunnel profile interpretation based on GDR
Table 6.1: Soil COnditioning in ring 103, 114, and 119
Ring C FER FIR BIR PIR
f
103 5 7 45 18 0
114 5 8 47 0 0
119 5 10 30 0 10
Figures 6.42 and 6.43 show chamber pressure data for rings 114 and 119. From these
Figures it is evident that same machine control results in a same overall behavior, including EC
145
|
Colorado School of Mines
|
Figure 6.44 presents the pressure gradients 40 minutes into mining for all three rings. The
curvature of the pressure distribution on the left side is less in rings 114 and 119 . in ring 103
is lower than rings 114 and 119, and for ring 114 is lower than ring 119 (see
∇
Pfigure 6.44).
Atmospheric muck density from rings 1 ∇0P3, 114, and 119 was found to be 1950, 1955, and 1998
kg/m3. As discussed in ring 103, muck height decreased during mining to the same level as rings
114 and 119. In ring 103 between levels 2 and 3, the muck exhibits a very low local gradient.
However, this is not the case in rings 114 and 119 and they are presenting higher gradients
between levels 2 and 3. During all three rings the influence of SC is very minimal; however, this
influence in ring 103 more noticeable than ring 114 and 119.
Figure 6.44: Ring 103, 114, and 119 pressure distribution during mining
6.10 Vertical and Horizontal Pressure Distribution of ring 131
As discussed in chapter four, the operator used the gates to control the discharge of
material, particularly when the material was flowable. Gate operation clearly influenced the AR
and consequently affected EC-EPS and SC-EPS. Operations data from ring 131 mining is
presented in figure 6.45. CH rotation is counterclockwise and increased pressure on the right side
to values higher than the left side of the chamber. FIR was reduced to 20% and FER was
increased significantly to 17 during ring 131 excavation.
148
|
Colorado School of Mines
|
CHAPTER 7 – CONCLUSIONS AND RECOMMENDATIONS
CONCLUSIONS:
Many factors are involved in excavation chamber pressure distribution during mining and
standstill. Operational parameters such as , , , , F and also gate G1,G2, and G3
openings play individual roles in realized θpSrCe1ssu θrSeCs2 a θnCdH pr θeCssAure changes in the chamber. The
following conclusions can be drawn from this study:
Testing of conveyor belt muck from rings 67 through 145 provided valuable information
about the conditioned soil behavior. Vane shear strengths ranged from 1-4.5 kPa. The more
cohesive soils in rings 67-91 exhibited higher vane shear strength (2.5-4.5 kPa) while the
granular soil from rings 111-145 exhibited lower vane shear strength (2.0-3.0). Slump levels in
the granular soils ranged from 3.6-14 cm.
Chamber pressures varied significantly from the crown to the invert of the EPBM. Up to
3 bar differences between crown and invert pressures were observed. Chamber pressures varied
from left to right at similar elevations. The side exhibiting the higher pressure depended on
cutterhead rotation direction. When CH rotation was clockwise, left side pressures were higher.
These pressure differences often receded during standstill but not always.
A clear relationship was observed between changes in muck volumetric flow rate
into and out of the chamber and chamber pressure. When was
posit (iv ∆e Q, inch )amber pressu (r ∆e Qinoucrt)e ased, and when was negative,
∆
c Qhianm −be ∆r
Q
poruetssure
decreased. The magnitude of the pressure cha ∆ng Qeisn −wa ∆s
Q
ofoutund linearly proportional to the
incremental change in volume change. These results convey the compressibility of conditioned
soil.
Increases in chamber pressure caused by flow rate changes, in turn influenced the
cutterhead force. When the chamber pressure increased as a result of more compression, the
cutterhead force decreased, as is consistent with a relative stiffness analogy.
153
|
Colorado School of Mines
|
ABSTRACT
Mining accidents related to ventilation problems that often occur deep within a mine are
costlyeventsintermsofproductionloss, equipmentloss, publicrelations, andultimately, the
cost of human lives. Although the mining process has become safer over the years through
mechanical extraction, the operation still requires human control, monitoring, and repair
at or near the mining face. Larger operations use the longwall mining process to extract
coal by cutting a 3 to 4 foot wide swath out of a long, continuous block of coal. During
the operation, the extraction face is ventilated where miners control the cutting process,
while the roof is temporarily held up by hydraulic supports called shields. As the shields
advance following the cut-out face, the roof behind collapses into a rubblized region called
the gob. Air flowing across the face may permeate into the gob, and mine gases (primarily
methane) liberated from the overlaying, fractured strata or the mine floor enter the mine
ventilationsysteminsidethegob. Continuousmonitoringandadjustmentofmineventilation
is required to prevent hazardous conditions such as explosive or irrespirable atmospheres.
Theinaccessibilityofthegobpreventsanyeffectivemonitoringordirectmeasurementsofgas
concentration, pressure, velocity or flow characteristics. Research has shown that fresh air
fromthemineventilationsystemcanenterthegobwhereitmaycreateexplosivemethane-air
mixtures. Also, with certain coals, ingress of oxygen into the gob can promote spontaneous
combustion of remnant coal.
The interaction of mine gas liberation and oxygen ingress into the gob is examined with
a Computational Fluid Dynamics (CFD) modeling tool developed during this project to
predict hazardous operating conditions and to help design ventilation systems to avoid these
hazards. This tool will be used to model the ventilation of a bleeder-ventilated, underground
longwall coal panel to assess the development of the explosive methane-air mixtures, in
contrast to the current regulatory view that this condition is not recognized as a hazard in
iii
|
Colorado School of Mines
|
a properly working bleeder-ventilated gob system.
ThisdissertationresearchemployedtheCFDmodelingsoftwarepackageANSYSfi Fluentfi
along with the output of a previous model developed using the geomechanical software pack-
age FLAC3D, to determine the permeability of the gob. Earlier studies have examined the
explosive mixture location and total explosive volume found in the gob and near the face,
gob caving characteristics, as well as the ingress of oxygen (Gilmore et al., 2013; Marts et al.,
2013; Wachel, 2012; Worrall, 2012). This research addresses the challenges of applying CFD
to model multiple mine ventilation layouts with the development of a modular meshing ap-
proach. This modeling approach incorporates the variations in mine lithology through the
development of multiple gob flow characteristics that can be applied across a wide range
of panel dimensions, and the challenges of modeling large scale mine networks and panel
lengths through the application of the modular meshing approach.
The modular meshing approach developed in this research project is used to build the
CFD models of a longwall gob ventilation system flexible enough to model a variety of differ-
ent mine layouts, mine ventilation schemes, and gob flow parameters. This is accomplished
by the creation of a library of meshed geometry modules that are interfaced together to
build the ventilation network surrounding the gob. The CFD mesh is tested by creating sev-
eral ventilation schemes and various mining conditions. Common operating conditions that
meet the mine ventilation regulatory statutes are used as the model boundary conditions.
The gob flow characteristics are then validated against a tracer gas study. A mesh module
repository is developed to help the mining industry create CFD models that match their
mine geometry, allowing access to models with fast compute times, and therefore, removing
what once hindered wide spread use of CFD simulations in underground ventilation. This
dissertation also presents a methodology used to determine the porosity and permeability
parameters, scalable in panel length and width, from the output of a FLAC3D model using
a combination of polynomial and exponential functions to fit the data.
iv
|
Colorado School of Mines
|
ACKNOWLEDGMENTS
I would like to thank my advisor, Dr. Bogin for his willingness to take on a new project
and the long discussions that ensued. I would also like to thank my co-advisor and Principle
Investigator, Dr. Brune for sharing his knowledge of the mining industry and his contin-
ual support of my efforts to improve this research, write papers and present my work at
conferences.
Special thanks to Dr. Grubb for initiating this research, writing the proposal and con-
necting this research to the coal mining industry in the Western United States. This research
project would not have been possible without the financial support of the National Institute
for Occupational Safety and Health (NIOSH) under contract number 200-2009-31409.
Thanks also to Dr. Worrall for completing the first CFD models in this research area,
and to the CSM High Performance Computing group for their administrative support of Mio
and upgrades in disk performance.
This research would not be possible without the work of Jon Marts who completed
the refined FLAC3D model that provided the data necessary to determine the gob flow
characteristics used throughout the CFD simulations. Thanks for completing the early
testing of the meshing approach by developing the progressively sealed gob models and
back return modeling approach, and for the long modeling discussions over the years.
Finally special thanks to my wife, Andine, for her continuous support, enabling me to
finish the work required to complete my degree.
xxxi
|
Colorado School of Mines
|
CHAPTER 1
INTRODUCTION
The coal mining industry remains an essential part of electrical power generation in
the United States. Mine ventilation challenges arise from a system consisting of dozens of
miles of underground pathways and the inability to take direct measurements in abandoned
regions. While mine operators seek to monitor and maintain the mines’ ventilation systems
to eliminate the accumulation of explosive mixtures of methane gas an improper ventilation
design, incorrect ventilation control or any change in ventilation pattern can cause severe
consequences including fires and explosions, the loss of the mine, equipment, and ultimately
the loss of miners’ lives.
Generally the approach to prevent methane explosions has been to dilute the methane
to concentrations below 1%, which creates a margin of safety to the lower explosive limit of
4.5% methane in air. Hazards also exist from fresh air or oxygen passing through reactive
coal that may cause spontaneous combustion in some mines.
Ventilation hazards are often the result of operators’ uncertainty about the ventilation
system response to system changes. The Computational Fluid Dynamic (CFD) modeling
tool can provide insight into the effectiveness of the ventilation system; these insights can
result in safer working conditions and more productive mines.
1.1 Motivation for Longwall Mining Ventilation Research
The coal mining industry supplies the United States with 38% of its electricity (U.S.
Energy Information Administration, 2015), through which society bears the significant cost
of miners’ lives lost. A total of 10,031 mining fatalities due to coal mine explosions have
occurred since the beginning of the twentieth century, in the United States, according to
United States Mine Rescue Association, 2013.
1
|
Colorado School of Mines
|
Two significant mining tragedies occurred as recently as April 5, 2010, when a methane
and coal explosion at Upper Big Branch Mine in West Virginia took the lives of 29 miners.
Another methane explosion occurred later that year at the Pike River mine in New Zealand,
killing 29 more miners. These two accidents were preceded by 104 coal mining deaths from
1980 to 2009 due to methane related explosions in the United States United States Mine
Rescue Association, 2013. Most of the accidents from methane related explosions occurred
in mines operating with a bleeder-ventilated gob system.
Currently, mine operators’ often have a reactive response to hazardous conditions. When
detector measurements trend towards hazardous conditions, a planned response goes into
effect in an attempt to avoid possibly catastrophic events. An automatic or manual de-
energizing of the mining machinery is the common response, after which the workers may
need to be evacuated from the mine. However, the response may not be fast enough when ex-
plosive methane mixtures ignite and propagate into a coal dust explosion extending through-
out the mine. Furthermore, a response to high carbon monoxide levels, which is a product
of self-heating coal, requires the immediate reduction of oxygen in the vicinity. This slow re-
sponse process often takes hours to implement. This response can cause severe consequences,
even possibly resulting in the total loss of the mine. This research seeks provide a predictive
tool capable of optimizing a response strategy and a tool capable of designing the initial
ventilation system to avoid the onset of problems well in advance.
Ventilationmodelingtoolscurrentlyavailabletomineoperatorsandventilationengineers,
such as VnetPC and Ventsim, are limited to one-dimensional ventilation network analysis of
the mine and are used to predict changes in new scenarios. However, limiting the domain to
linear network flows using a laminar flow assumption does not provide any insight to flow
patterns in the three-dimensional, inaccessible regions of the mine.
This research provides a viable alternative to modeling large-scale mining operations
involving flow patterns deep within the mine, as well as giving mine operators’ the ability
to evaluate hazardous gas compositions related to spontaneous combustion and explosive
2
|
Colorado School of Mines
|
methane-air mixtures. The tools created in this project support increased use of CFD
modeling to study ventilation effects. The modular meshing approach and scalable gob
flow characteristics developed for this project can be adapted by future modelers to evaluate
proposed changes to the ventilation system, promoting miners’ safety. This project seeks
an objective baseline to evaluate and implement safer ventilation operating conditions and
designs for mines in the United States. A review of current literature in the area of mine
ventilation research is presented in Chapter 2 to properly place this work in context.
1.2 Introduction to Underground Longwall Coal Mining
In the United States, 49% of the underground coal production is by longwall mining.
The longwall mining method produces 95% of the coal mined underground in the Western
United States (U.S. Energy Information Administration, 2015). This process yields higher
extraction rates compared to room-and-pillar mining. The longwall mining process begins
by driving development gateroads in a “U” shape around a solid block of coal, called a panel,
as shown in Figure 1.1(a). The primary haulage gateroad, called the headgate, is also the
ventilation intake from the main development section, called the mains, and the secondary
gateroad, called the tailgate, is either a return airway or a secondary intake depending on the
ventilation layout. The development gateroads are mined by continuous mining machines
and become the headgate and tailgate of the outlined panel. The headgate and tailgate are
connected to the mains, where fresh air is distributed and exhaust air leaves the mine.
(a) Mining Plan View (b) Section View XX
Figure 1.1: Longwall Mining Plan View Layout (The Davies Family Website, 2013)
3
|
Colorado School of Mines
|
The longwall panel consists of a solid block of coal, typically ranging from 3,000 m to
6,000 m (10,000 ft to 20,000 ft) in length and from 240 m to 460 m (800 ft to 1,500 ft) in
width. The height of a typical longwall mined coal seam ranges from 2.7 m to 3.4 m (9 ft to
11 ft) in the Western United States, and 1.5 m to 2.4 m (5 ft to 8 ft) in the Eastern United
States. Mining of the longwall panel starts at the end opposite from the mains, where the
longwall mining equipment is installed in a special startup entry along the panel width. The
panel will be extracted in 0.9 m to 1.2 m (3 ft to 4 ft) increments by moving a shearer, which
cuts across the panel width as the longwall face retreats towards the mains. A cross-section
of the coal face is presented in Figure 1.1(b): Section View XX, showing the major elements
of the longwall mining equipment: the shearer, the armored face conveyor, and a series of
hydraulic supports or longwall shields. In the right of the figure, as coal is extracted, the
roof rock collapses and forms the gob. The ventilation air provided to workers at the coal
face, flows through this “U” shaped ventilation system and traverses the longwall through
the cross-section shown in Figure 1.1(b). A discussion of ventilation system details are in
Section 1.4.
The longwall mining equipment, as shown in Figure 1.2: A, and the expanded view
Figure 1.2: C, spans the width of the coal panel using about 150 to 250 shields with each
shield approximately 1.5 m to 2 m (5 ft to 6.6 ft) wide. The shearer cuts across the coal face
as it traverses its width, while the shields advance into the newly created space to maintain
roof support. Also shown in Figure 1.2: A are the development entries (gateroads) on both
the headgate and tailgate sides in the upper and lower part of the Figure. These gateroads
are held open by the blocks of coal forming pillar supports. The area of collapsed roof behind
the shields, called the gob, is shown in the center of the Figure and is better seen in the
section view of Figure 1.2: B.
The gob consists of various sizes of rock blocks from the failing overlying roof material,
filling the void of the mined out coal seam. The crushing weight of the overburden compacts
the gob further leaving behind a rubblized zone filled with rocks ranging from large boulders
4
|
Colorado School of Mines
|
to finely crushed gravel. The harsh environment and the inaccessibility of the gob zone make
direct measurements of any kind nearly impossible.
Fresh air is supplied to the face where mine personnel operate the shearer and retreat the
shields. The cross-section view in Figure 1.2: B shows the seam height and overlying rock
strata (overburden) supported by the shields. The immediate mine roof caves into the space
behind the shields as they retreat, and the remaining rock above fractures as it subsides,
creating cracks and fissures that may allow methane from the upper strata to migrate into
the gob.
Hydraulic roof support shields, shown in Figure 1.3(a), are matched to the height of the
coal seam and overburden load. The shield design height can vary from 1.2 m to 5.5 m (4 ft
to 18 ft) depending on the geology surrounding the coalbed and the extraction height. Each
shield is designed to support the roof with two hydraulic jacks. To retreat a shield, it is
lowered and then pulls itself backwards by a hydraulic ram, called the relay bar, attached to
the armored face conveyor. This connection can be seen on the left of Figure 1.3(b), with
the shearer on the right cutting the coal face.
1.3 Description of Mines – Mines C, E, and W
The modeling data comes from two Western United States underground longwall coal
mines and from observations made by the research team during mine visits. The geome-
chanical model developed by Wachel, 2012, refined and validated by Marts et al., 2014b is
used to determine the gob flow characteristics of Mine C and E. The gob flow characteristics
of Mine W are from Wachel’s geomechanical model based on Mine C data and adapted by
Worrall, 2012 for use in the first iteration of CFD modeling. A review of background work
in geomechanical model is in Section 2.2, and the scalable implementation into Fluent is
discussed in Section 4.3
The coal seam at Mine C is 3.35 m (11 ft) with a depth of cover of 138 m (453 ft) and
an immediate roof consisting of weak mudstone and shale layers forming the gob material.
The rock diameter in the gob directly behind the shields and in the fringe of the gob is
6
|
Colorado School of Mines
|
(a) Roof Support Longwall Shields for Different (b) Longwall Shearer and Armored Face Conveyor
Seam Heights (Caterpillar Inc., 2011) (United States Mine Rescue Association, 2013)
Figure 1.3: Longwall Shields and Shearer
approximately 1 m (3 ft) in size and smaller. The final subsidence, expressed as a percentage
of extraction height, was 77%, with a panel width of 304 m (1,000 ft).
The coal seam at Mine E is assumed to be a uniform 2.9 m (9.5 ft) with a depth of
cover of 123 m (404 ft). However, the coal seam joins a rider seam increasing the thickness
occurring under a mountain where the depth of cover increases to 240 m (800 ft). This
large shift in parameters is not included in the geomechanical models. The immediate roof
consists of strong, massive sandstone, which occasionally produces large boulders in the gob
thereby producing a large rock size distribution. The final subsidence is 58%, with a panel
width of 380 m (1,250 ft).
1.4 MineVentilation–Bleeder-ventilatedGobSystemsandProgressivelySealed
Panels
There are two general types of ventilation systems used in underground longwall coal
mining operations: bleeder-ventilated gobs and progressively sealed gobs, sometimes called
bleederless gobs. The first method, typically used only in the United States, involves the
7
|
Colorado School of Mines
|
intake of fresh air into the development entries surrounding the gob in order to ventilate
the gob area in a controlled manner. The process aims to dilute any methane inside the
gob to levels below the explosive range. The second ventilation method, uses a progressive
sealing process around the gob and is often combined with inertizing the gob atmosphere
with nitrogen or other inert gases. Progressively sealed gobs are commonly used throughout
the rest of the world, but only with special permits in the United States.
Figure 1.4 shows an example of a bleeder-ventilated gob system. Fresh air, shown in blue,
is supplied from the mains at the bottom of the figure to the active longwall face. The air
then splits in three directions: down the active face, returning out with the conveyor belt
(called neutral air shown in green), and inby the face to the bleeder fan (direction heading
deeper into the mine). Once the air travels the length of the face to the tailgate, there are
several options for design possible. Shown in the Figure 1.4 is a “back return” at the tailgate,
where the air is directed one-crosscut inby and returns further inby the bleeder to the bleeder
fan with addition fresh air added at the tailgate. This is the more common design option for
the tailgate side to supply fresh air and force all the return air to exhaust out the bleeder
fan. Further details of the modeling environment used are in Chapter 4.
The progressively sealed system in Figure 1.5 shows how the ventilation is isolated from
the gob area with concrete seals in the crosscuts along the headgate side, inby the longwall
face. Additionally, nitrogen may be injected through these seals at various locations to
inertize the gob atmosphere. For seals that separate the main ventilation system for the gob
the completed longwall panels have seals that encase a balance chamber filled with nitrogen.
This is to prevent oxygen-rich mine air from leaking into the gob by maintaining a pressure
differential between the chamber and the mine air (shown on the left of Figure 1.5). Mine C,
studied during this research, uses two headgate nitrogen injection locations, one tailgate
injection point and balancing chambers. This mine also uses Gob Ventilation Boreholes
(GVBs) drilled from the surface into the fractured zone to extract methane from the gob.
The GVBs are equipped with vacuum pump systems to assist the mine ventilation system
8
|
Colorado School of Mines
|
in withdrawing methane from the mine. The additional cost of this ventilation system over
using a bleeder-ventilated gob system is estimated by Grubb et al., 2015 to be 25 cents per
ton of coal produced.
1.4.1 Design of Ventilation Systems for Fire and Explosion Prevention
The choice of ventilation design for an underground longwall mining operation depends
on suspected ventilation hazards of methane gas and the particular coal’s propensity for
spontaneous combustion. For example, coal mines in the Eastern United States are typically
gassy mines, which have coal’s not generally prone to spontaneous combustion and therefore
use the bleeder-ventilated gob systems to dilute the methane gas, as required by law under
the Code of Federal Regulation (CFR) Title 30, Part 75.334 (United States Department of
Labor, Mine Safety & Health Administration, 2012). The goal of the bleeder system is to
dilute the methane inside the gob into the non-explosive range before it reaches an area with
potential ignition sources. This is done by providing as much air as necessary in all entries
surrounding the gob.
Bleeder systems present problems in underground coal mines in the Western United
States. The coal found in the West is often more prone to spontaneous combustion, and
excessoxygeninthegobmayreactwiththecoalleftbehind. Therefore,someWesternUnited
States mining operations use a progressively sealed ventilation design to avoid spontaneous
combustion hazards.
The progressively sealed gob ventilation system has the additional benefit of helping limit
oxygen penetration into the gob with the injection of nitrogen, thereby further reducing the
possibility of spontaneous combustion events. Details about spontaneous combustion are
reviewed in Section 2.4.
1.5 Computational Fluid Dynamics – Fluent Software
Computer simulation of fluid flows using equations of flow physics in a discretized volume
of fluid is termed Computational Fluid Dynamics or CFD. There are a number of available
9
|
Colorado School of Mines
|
The three-dimensional mesh creation and manipulation available with ANSYS products
offers great flexibility to control the geometry solved with Fluent. Many ventilation scenarios
can be created by assembling mesh parts or suppressing existing parts. For example, a large
model mesh encompassing a full longwall panel can be created from individual high quality,
low skewness mesh modules to be assembled and solved on a supercomputer. Figure 1.6
shows a mesh made from eight individual mesh modules, creating the desired geometry.
Figure 1.6: Example of Panel Mesh
Through UDFs in Fluent, custom output variables are generated based on properties
solved in the model. For example, a graphical output scheme is created that translates
methane and oxygen concentrations into a color coded explosive range plot, shown in Fig-
ure 1.7, called a gob gas analysis plot. This quickly identifies the Explosve Gas Zones (EGZs)
through the application of Coward’s triangle (Coward & Jones, 1952). This figure is recre-
ated by applying the same boundary conditions from Worrall, 2012 and nitrogen injection,
but a new mesh is used with a greater panel length. Details of the gob gas analysis and the
color scheme used, which changed blue to dark-green, are found in Sections 5.5–5.6.
1.6 Research Objective and Hypothesis
The objective of this research is to provide the mining industry with a CFD predictive
modeling tool for in-house evaluation of the effects of proposed, planned and unexpected
changes to the mine ventilation system, and mitigation of hazardous gob gas compositions.
The CFD modeling results can then be used for planning and risk assessment to build a
customizable risk management plan for the mine.
12
|
Colorado School of Mines
|
Figure 1.7: Gob Gas Analysis Explosive Results
The modeling approach developed in this research is flexible to a variety of geometry
options through the creation of an easily modifiable mesh geometry library to enable sim-
ulations of actual mine ventilation layouts. This modeling approach is fully adaptable to
underground longwall gob ventilation when combined with a scalable equation for gob com-
paction to determine the gob flow characteristic parameters of porosity and permeability,
including variables to change the host rock properties involved in the conversion.
Using a trend-based analysis of potential hazardous ventilation changes that may not
represent actual conditions, but rather provides insight into the overall response of the sys-
tem, this research examines the explosive gas mixture development in longwall gobs and the
surrounding ventilation networks. It is known that methane must follow the dilution path
from fuel-rich through an explosive zone to fuel-lean (see Section 5.5), therefore this research
asks the question: Do EGZ exist in bleeder-ventilated gobs, and if so, where are they located
and how large is the explosive gas volume?
13
|
Colorado School of Mines
|
This research considers the following hypothesis:
(cid:136) EGZs exist in a bleeder-ventilated gob, then EGZs will be found in CFD modeling
results when using acceptable regulatory operating conditions in bleeder-ventilated
gobs.
1.7 Specific Aims and Research Tasks
The specific aims of this research project are to mitigate hazardous ventilation conditions
in underground coal mines, helping the mining industry achieve their goal to reduce fire
and explosion accidents to zero. This project will benefit mine workers, companies, and
stakeholders in the industry by decreasing risks related to ventilation accidents, which has
the added benefit of improving public perception of underground mining operations. The
following sections present the related tasks to reach these aims: development of a modular
meshing approach, creation of a scalable equation fit for gob permeability, and building the
CFD model to guide mine ventilation systems design.
1.7.1 Universal Mesh Assembly for Multiple Mine Geometries
One of the most challenging parts of CFD simulations is creating a mesh that can be
solved under a variety of conditions. In this project, a library of mesh files of basic geometric
parts is prepared. The creation of high quality, low skewness meshes facilitates the reduction
in simulation time, increase stability, and reduces overall resources needed to converge to
a solution. The meshes can be assembled into a layout mimicking the ventilation network
surrounding any longwall gob area.
The mesh library consists of mesh parts or modules (see Figure 1.6) that can be inter-
changed and scaled as needed to match a specific mine layout. A basic layout, for example,
contains the parts for the ventilation airways of the headgate, face and shields, tailgate, and
any crosscuts or connecting entries with the option to include regulators. It also contains
the gob, open fringe surrounding the gob, upper and lower strata, and GVB parts. These
14
|
Colorado School of Mines
|
modules can be assembled into an interfacing network of mesh connections for the CFD
solver.
The mesh module creation follows the guidelines provided by ANSYS for quality and
skewness. The ventilation airways that often contain turbulent flows are modeled with a
combination of inflation layers and tetrahedral mesh elements, while the porous regions that
dominate the simulation contain hexahedral elements, only. The hexahedral element is ideal
for modeling slow moving flow. This combination of mesh types ensures faster simulation
times, which have previously taken many hours to solve on dedicated processors; this makes
it possible to achieve reasonable solution times employing desktop resources. Using this
approach maintains a higher quality mesh ensuring stability under a variety of operating
conditions. Details of the mesh creation process are in Chapter 6.
The modular mesh library is tested using several mining scenarios including progressively
sealed gobs optimizing nitrogen injection (Marts et al., 2015), active face ventilation schemes
(Marts et al., 2014a), modeling a large bleeder-ventilated gob panel (Gilmore et al., 2014),
and multiple panel bleeder-ventilated gobs. More details are in Section 7.2.
1.7.2 Piece-wise Scaled Equation Fit Implementation of FLAC3D Output
A geomechanical model created by Marts et al., 2014b is used to determine the gob
flow characteristics required to solve the flow inside the gob as it responds to changes in
the ventilation system. The FLAC3D (Itasca Consulting Group, 2013) output of volumetric
strain increment (VSI) describes the change in the porosity inside the gob. Detailed in
Section 4.3, a piece-wise equation is built, which is applicable in certain bounds within the
gob. This approach permits scaling of the mine geometry in length and width enabling a
comparison of different mine lithologies; the resulting EGZs within the gob are presented in
Chapter 8.
A previous project completed the FLAC3D modeling and refined and calibrated it to
known conditions Marts et al., 2014b. The approach to calculating the porosity and perme-
ability from the VSI follows the work of previous researchers Esterhuizen et al., 2010; Lolon
15
|
Colorado School of Mines
|
& Calizaya, 2009; Wachel, 2012; Worrall, 2012; Yuan & Smith, 2010. Where the conversions
from VSI take place after the equation fits enabling greater control to the modelers. The
equation fits are in the form of a set of piece-wise equations bounded to a sectional area
of the gob. In addition, the VSI conversion to porosity and permeability yields parameters
that can be adjusted by the modeler for a given mine’s caving characteristics. A review of
porous media modeling is presented in Section 2.2.
1.7.3 Computational Fluid Dynamics Gas Flow Model
AFluentCFDmodelisassembledusingagroupofrepeatingmodularmeshfilestomodel
a bleeder-ventilated gob network surrounding a large gob area in a steady-state case. The
approximation of a steady-state analysis is considered valid under the circumstances of an
advancing longwall face because the rate of advance is small, 9 m to 30 m per day (30 ft to
100 ft), relative to the hundreds of meters of the panel length. This is also applicable during
longwall maintenance periods when the face advance is stopped.
The bleeder-ventilated gob model is calibrated to acceptable operational parameters at
the inlets and outlets of the ventilation as provided by statutory requirements. Section 7.1
presents model comparison of predicted velocities inside the gob to the velocity of a tracer
gas release study for validation.
The CFD simulation results of the formation of EGZs in bleeder-ventilated gobs are
discussed in Chapter 8. Safety concerns and suggestions for future work follow in Chapters
9 and 10, respectively.
16
|
Colorado School of Mines
|
CHAPTER 2
BACKGROUND
A review of ventilation safety technologies commonly found in the mining industry and
the federal regulations that govern the operation of underground coal mines are introduced
in this chapter. A review of research conducted to determine the gob flow characteristics
begins with an examination of the roof caving process. This is followed by a review of
the ventilation research and CFD studies, and a brief overview of the chemical reaction
of spontaneous combustion. This chapter concludes with proposed improvements for mine
ventilation safety.
2.1 Ventilation Safety Technologies
Mine ventilation systems dilute the explosive gases released during the mining process,
which consists of primarily methane. In addition to dilution, explosion hazards may also be
controlled through inertization, e.g., injecting nitrogen. Companies use mine atmospheric
monitoring to detect harmful gases and to analyze changes in gas composition and resulting
trends. There are a variety of monitoring systems available to use throughout the mine,
from handheld systems to continuous monitoring systems. A review completed by Grubb,
2008 outlines the safety and operational risks for the efficacy of many of these technologies in
miningoperations, andrecommendstheuseoftubebundles, progressivesealingandnitrogen
inertization in longwall mining operations.
2.1.1 Ventilation Regulatory Standards
The United States CFR, Title 30 covers all mandatory safety standards for the mining
industry including underground, surface, hardrock, coal, metal, and non-metal mining. Sec-
tions dealing with mine ventilation and air standards in underground coal mining are found
in 30 CFR, Part 75.300 and the following parts. These standards include sections regulating
17
|
Colorado School of Mines
|
breathable air in the mine and measurement locations, gas concentration limits, dust control,
temperature and inertization (Mine Safety & Health Administration, 2012).
For example, where miners work and travel oxygen content must be greater than 19.5%,
with carbon-dioxide less than 0.5% (30 CFR §75.321), and methane less than 1% (30 CFR
§75.321). The maximum allowable methane concentration is 1.5% or 2% in certain return
airways, including bleeder entries (30 CFR §75.323(d) and (e) 2012).
2.1.2 Monitoring and Measurements Technologies
Themonitoringofthemineatmosphereisrequiredunder30CFR§75.360, whichspecifies
locationsintheminewhereairqualitymustbemonitoredpre-shift,andatcertainintervalsof
the shift, day, week, or month. These points are marked in the ventilation plans by the mine
operatorandaremeasuredandrecorded. Themonitoringequipmentusedtomeasurequality
parameters may include handheld gas monitor devices, continuous atmospheric monitoring
systems (AMS), gas bag sampling and the use of tube-bundle monitoring systems.
A handheld gas monitor is typically capable of measuring the concentration of oxygen,
carbonmonoxide, methane, carbondioxide, and, asneeded, hydrogensulfideandothergases.
Air quality data from bleeder-ventilated gob systems are usually gathered with handheld
devices during the weekly inspections conducted by certified mine examiners.
The continuous AMS are most commonly used on the headgate and tailgate drive areas
of the longwall armored face conveyor and on the shearing machine. The AMS measure-
ments can sound alarms and de-energize the mining equipment if hazardous concentrations
of methane are detected.
Gasbagsamplesaretakenmanuallyandaretypicallyanalyzedwithagaschromatograph.
Thesamplesareanalyzedforthemaincomponentsofcommonexplosivegasesfoundinmines
and spontaneous combustion products, typically consisting of the following gases: hydrogen,
methane, oxygen, nitrogen, carbon dioxide, carbon monoxide, ethylene and acetylene.
A tube-bundle system continuously draws gas samples through tubes from locations
within the mine unreachable by other sampling techniques, including from behind seals.
18
|
Colorado School of Mines
|
These samples are drawn to the surface for analysis of methane, oxygen, nitrogen, carbon
dioxide and carbon monoxide concentrations.
2.1.3 Gob Ventilation Boreholes (GVBs)
The objective of GVBs is to remove methane by means other than ventilating it through
the mine. GVBs are similar in fashion to natural gas wells, except that the fracturing of
the rock occurs from undermining the gas reservoir with the longwall operation. Figure 2.1
shows a cross-section of the strata layers of a coal mine. The active mining face at the
bottom is ventilated with fresh air; the methane comes from the reservoir in the strata and
desorption from the coal seams. A bore hole is drilled from the surface to within 12 m to
24 m (40 ft to 80 ft) above the coal seam with approximately 60 m (200 ft) of slotted casing
at the bottom. The optimum size, locations and depth of GVBs for the Pittsburgh Coalbed
are modeled by Karacan et al. (2007a), recommending the addition of GVBs for mines with
a wider longwall panel width.
Figure 2.1: Gob Ventilation Borehole Capture System (Karacan, 2009c)
19
|
Colorado School of Mines
|
2.2 Gob Porosity and Permeability
Much research has been done in the field of defining fluid flow in porous media in the past
decade. The works of Nield & Bejan, 2013, ”Convection in Porous Media”, Bird et al., 2007,
”Transport Phenomena” and De Lemos, 2012, ”Turbulence in Porous Media: Modeling and
Applications” guide the discussion of this topic. This section presents a review of porous
media modeling and the assumptions that define its use. Additionally, the process of gob
formation, compaction and surface subsidence is reviewed; and finally, the geomechanical
numerical modeling results and ranges of permeability are presented.
2.2.1 Flow through Porous Media
A porous medium is defined as a region containing a distribution of solid particles and
interconnecting void spaces. The percentage of connecting voids is termed the porosity,
n (ANSYS, 2014b documentation uses ε and γ, and mathematical derivation is often ϕ).
Therefore, the volume of fluid, V is given in Equation 2.1, and the volume of solid, V
fluid solid
is given in Equation 2.2
V = V ×n (2.1)
fluid
V = V ×(1−n) (2.2)
solid
In Table Table 2.1 some values of porous materials are given. The predicted values based
on Karacan, 2010 experimental work for gob material from two Eastern United States coal
mines are shown. The initial uncompacted porosity ranges from 0.629 to 0.711 and the final
compacted porosity ranges from 0.216 to 0.383 after applying a correction factor from the
laboratory scale.
Figure 2.2 shows a Representative Elementary Volume (REV) in the center with length
scale, D,theporesizelengthscaleontheright, p, andthedomainlengthscale, L,ontheright
(Teruel & Uddin, 2008). At the pore scale, the flow variables will have large fluctuations,
but space-averaged across many pores on the REV scale, the variables become steadier. A
20
|
Colorado School of Mines
|
2.2.2 Porous Media Model
The flow velocity through a volume of the porous medium is termed the superficial
velocity, also referred to as the seepage, filtration or Darcy velocity, based on Darcy (1856)
law of fluid flow through a porous media, given in simplified form in Equation 2.4
K ∂P
u = − (2.4)
µ ∂x
where u is the velocity component in the x-direction across some pressure gradient P, and
µ is the fluid viscosity. This law defines the permeability unit of Darcy’s, K, equal to
0.987×10−12m2. The value of permeability is related to the porosity and particle diameter,
D , by the Carman-Kozeny relationship given in Equation 2.5
p2
(D )2n3
p2
K = (2.5)
180(1−n)2
where D may be represented by a stochastic model of the density function, h(D ), of the
p2 p
form given in Equation 2.6
´
∞ (D )3h(D )dD
D =
´0 p p p
(2.6)
p2 ∞
(D )2h(D )dD
0 p p p
The constant of 180 in Equation 2.5 is based on a tortuosity of 2.4, a slightly higher value
than the value of 2, which would approximate spherically shaped particles. A refinement by
including a second term for inertial damping is suggested for use by ANSYS (2014b) using
the Ergun approach in Section 3.7.
Figure 2.3 presents the work of Ward, 1964 that identifies the two major flow regimes,
Darcy flow and Forchheimer flow, and the smooth transition that occurs in experiments.
This smooth transition, shown as modeled, uses the Ergun approach. Darcy’s law holds
for values of Reynolds numbers less than 0.1, where the Reynolds number scale with the
square root of the permeability and the friction factor remains linear. In the Forchheimer
(1852–1933) region, the friction factor remains constant with increasing Reynolds numbers.
Other length scales for the Reynolds number are suggested by Bird et al., 2007 based on the
pore scale instead of the permeability. Since the transition to the region of non-linear drag
22
|
Colorado School of Mines
|
Figure 2.3: Porous Media Flow Regimes (Ward, 1964)
is smooth, it should be noted that this is not an effect of turbulence, and the flow in the
pores remains laminar (Bird et al., 2007).
The recommendation given by Nield & Bejan (2013) is that “until further experimental
work is carried out, the simple quadratic expression for the form drag be used, on the
understanding that the coefficient is not necessarily given by the Ergun formula.”
2.2.3 Porous Media Boundary Condition
There is considerable debate regarding the matching of the interface between the Navier-
Stokes free fluid flow solution, as discussed in Chapter 3, and the solution obtained in the
porous medium. The problem is illustrated in Figure 2.4 by the lack of the development of
a boundary layer in the velocity profile. The velocity at the impermeable wall at the top is
well defined, as discussed in Section 5.3.3; however, the permeable interface at the bottom,
a relationship between the free fluid velocity, u and the velocity in the porous medium, u
f m
must be defined.
In the case of a liquid medium exposed to air, the boundary condition may be expressed
with the following relationship given by Equation 2.7
23
|
Colorado School of Mines
|
∂v
= 0 at y = 0 (2.7)
∂y
Beavers & Joseph, 1967 proposes that in the case of the same fluid saturating the medium,
the relationship may be expressed by Equation 2.8
∂u α
f BJ
= (u −u ) (2.8)
∂y K1/2 f m
where u is evaluated at some small distance into the free fluid, y = 0+ from the plane at
f
y = 0, and u is evaluated at y = 0− in the porous media. The constant α may range from
m BJ
0.1 to 4 and depends on the geometry of the porous media and problem being considered.
The most recent work by Nabovati & Amon, 2013 uses a lattice-Boltzmann approach that
considers the colliding particles on the surface of the porous medium and predicts results
similar to parameters used in experimentation.
Figure 2.4: Porous Media Boundary Flow (Nield & Bejan, 2013)
2.2.4 Porous Media with Turbulence
De Lemos, 2012 discusses the treatment of turbulence and buoyancy in a porous medium,
as well as an impinging jet onto a porous layer, among many other porous media flow cases.
The author presents the coupling of the turbulence equations with the porous media model
assumption using a double-decomposition model based on volume averages that include both
24
|
Colorado School of Mines
|
spatial deviations and time fluctuations. The algorithms presented in De Lemos, 2012 are
not mentioned in ANSYS, 2014b, release and remain an area of ongoing research.
2.2.5 Gob Formation
Thegasflowcharacteristicsidentifiedinthegobdependontherockstratathatultimately
will form the gob and overburden that compacts it, crushing the rock and reducing the
particle size. Numerical modeling, direct stress measurements, tracer gas and laboratory
block modeling have all been use to study the gas flow characteristics of the material that
forms the gob behind the retreating longwall mining machine. These studies predict the
formation, structure and variability that create the porosity and permeability used in flow
models.
The gob formation occurs as the longwall shields retreat out, allowing the roof to collapse
into the void of the extracted coal seam. Initially, the gob consists of loose rock and rubble
with a high porosity and permeability. The gob formation from the roof caving process
causes failure and delamination in the upper strata rock and creates three distinct regions as
shown in Figure 2.5: a rubble zone of broken pieces of rock or gob, a zone of fractured rock
that is being supported by the gob, and a strata bending and delamination zone extending
to the surface (Peng & Chiang, 1984; Singh & Kendorski, 1981). The surface subsidence,
defined as the downward movement with respect to the original elevation (dotted line), is
also shown in Figure 2.1. Once the coal is removed, the overburden load redistributes across
the gob. As the face advances farther, the overburden load compacts the gob material and
reduces the porosity and permeability.
The height of each zone shown in Figure 2.5 depends on the geological conditions of
the mine site: lithology of the overburden strata, mine depth and mine extraction height.
The height of the caved zone of the gob is dependent on the overburden strength and the
orientation, rotation and stacking of the falling blocks and their bulking factor (Esterhuizen
&Karacan,2007). Thegobheightistypically3to6timestheextractionheight(Esterhuizen
& Karacan, 2005). The fractured zone above extends from 30 to 60 times the extraction
25
|
Colorado School of Mines
|
height and consists of bedding plane separation and vertically aligned fractures. This zone
may be shaped as an arch or saddle depending on overburden properties (Bai et al., 1995).
2.2.6 Gob Compaction
Studies have been conducted to understand the compaction of the gob material for both
porosity and permeability. An indicator for the compaction is surface subsidence that must
be considered for building damage estimates or continuing ground movement (Karmis et al.,
1984; Landsberg, 1936). Gob compaction depends on the following two factors: the bulk-
ing factor describing how a block initially falls into place, and the rock fragment strength
(Karacan et al., 2007a; Yavuz, 2004).
The bulking factor is determined from shape, falling height of the rock fragments, size
of the fragments, and size distribution of rock fragments. The bulking factor initial value is
determined by the fall height created by the removed coal, and decreases to zero as the fall
height decreases within the caving zone (Karacan et al., 2007a; Yavuz, 2004). The bulking
factor is inversely proportional to gob compaction (Esterhuizen & Karacan, 2007), which
will result in a maximum bulking factor directly behind the shields and near the gate roads,
and reduced bulking toward the center of the gob with higher compaction. Pappas & Mark,
1993b determined that the initial void ratio is 30% to 45% in laboratory tests.
The compacting gob receives the majority of the load from the strata a short time after
the face retreats. Once the initial load compacts the gob, the gob material begins to respond
withstrainhardeningandstiffeningbehavior; thiscrushestherockandleadstoanon-elastic,
irreversible response. An estimate of the gob material response to loading is first presented
as a hyperbolic curve given in Equation 2.9 (Salamon, 1990)
a(cid:15)
σ = (2.9)
b−(cid:15)
where a and b are empirical parameters, and (cid:15) is the amount of plastic-volumetric-strain.
The value of a is the stress before the material begins to harden, which occurs when (cid:15) equals
twice the value of b. The value of b is the initial value of the bulking factor when the gob
26
|
Colorado School of Mines
|
W(t) = W (1−e−ct) (2.11)
o
where W(t) is the subsidence at time t, and W is the final subsidence at any given point,
o
and the constant, c, represents the properties of the overburden. The mining extraction rate
relates to the subsidence rate (Cui et al., 2001), confirming a modification to the Knothe
equations results in Equation 2.12
(cid:16) (cid:17)
cr
K
Dyn
= K
Final
1−eVp (2.12)
where K is the final subsidence, K is the dynamic subsidence value, c is a time
Final Dyn
coefficient, r is the radius of influence, and V is the face advance rate (Peng et al., 1992).
p
The immediate roof rock strata properties play an important role in the final form of
the gob and the dynamic gob compaction. Two caving response cases studied by Hill, 1995
discuss, parting-plane controlled and bulking controlled caving response. He notes that when
the roof contains thick, strong rock, the roof tends to hang and fracture at the parting-plane
inlargepiecesasitfalls. Iftheroofiscomposedofweakrock, theblockrotationandstacking
(bulking factor) play a greater role in the formation of the gob. The final formation of the
gob may contain a void between the broken strata layers and the gob, or it may completely
fill. More often though, there are combinations of the two formations, and direct observation
will only determine the size and shape of the void (see Section 4.2.2).
2.2.8 Volumetric Strain Increment (VSI)
Thevolumetricstrainincrement(VSI)isameasureofthechangefromtheinitialporosity
ofagranularmaterial. LaboratorytestsperformedbyJozefowicz,1997, proposedamodifica-
tion to the Pappas & Mark test. Tri-axially loaded samples of sandstone, shale, or gritstone
were compressed uniformly in a Hoek-Cell acting as permeameter shown in Figure 2.7, where
the sample is compressed in all directions.
During the compression process, nitrogen gas is injected at a known pressure, while the
flowrateofgasismonitored. Darcy’slawisusedtodeveloptheintrinsicpermeabilityateach
29
|
Colorado School of Mines
|
Figure 2.7: Tri-axial Compression Apparatus Hoek-Cell (Hoek & Franklin, 1968)
compressive pressure. As a result, a relationship between volumetric strain and permeability
determined the constants in Equation 2.13
(cid:8) (cid:9)
k = −4×10−16(cid:15)3 −6×10−15(cid:15)2 −7×10−14(cid:15) +1×10−11 m2 (2.13)
gob Vol Vol Vol
where (cid:15) is the volumetric strain, and k is the gob permeability.
Vol gob
A coefficient of permeability from direct in-situ rock property measurements is first in-
troduced by Szlazak, 2001, in a study of gob air flow patterns and spontaneous combustion.
Assuming a laminar flow in the gob, the coefficient of permeability is derived from linear
porous media filtration as given in Equation 2.14
L(cid:52)Q
(cid:8) (cid:9)
k = µ m2 (2.14)
S (cid:52)p
where µ is the coefficient of absolute viscosity of air, S is the cross-sectional area, (cid:52)p is the
pressure drop across the distance, L and (cid:52)Q is the volumetric flow in the gob.
The porosity of the gob material is initially assumed to be 40−50% (Pappas & Mark,
1993b), depending on the host rock of the roof. The Pappas & Mark value of initial porosity
30
|
Colorado School of Mines
|
is derived from the void volume in the cave zone, which is the ratio of extracted seam height
over the caving height of the gob, plus the inherent porosity value of the rock material that
makes up the host rock forming the gob, (H /H +n ) equals the initial porosity
coal cave inherent
value. ThisinitialporosityvalueminusthecompactionvalueofVSI,(n −vsi), equalsthe
inital
final porosity of the gob material. The Carman-Kozeny relationship between permeability
and porosity, expressed in Equation 2.15
K (cid:18) n3 (cid:19)
o
K = (2.15)
gob 0.241 (1−n)2
where K is the base permeability of the rock, and n is the porosity. Equation 2.15 is used to
o
relate the output of FLAC3D and the inputs required to complete the flow characterization
of the gob. This relationship is used by Esterhuizen & Karacan, 2007; Lolon & Calizaya,
2009; Wachel, 2012 and is generally accepted among researchers.
A complete range of permeability values used by researchers is given in Table 2.2, and
summarizes the range of minimum and maximum values. The work completed previous by
CSM researchers Wachel, 2012; Worrall, 2012 for Mine C became the first models of Western
United States mines.
Table 2.2: Comparison of Gob Permeability Findings
Source Max Perm [mD] Min Perm [mD]
Brunner, 1985 1.01×1010 1.01×108
Ren & Edwards, 1997 n/a 1.01×105
Szlazak, 2001 1.01×109 5.07×106
Wendt & Balusu, 2002 n/a 1.00×106
Whittles et al., 2006 5.07×108 1.01×107
Esterhuizen & Karacan, 2007 n/a 1.00×106
Lolon & Calizaya, 2009 4.74×108 8.10×106
Karacan, 2009c 3.55×107 1.52×107
Karacan, 2009d 1.27×107 5.07×106
Ren et al., 2011 2.00×109 2.00×106
Worrall, 2012 Mine C 6.99×109 2.03×108
Marts et al., 2014b Mine C 5.17×109 2.03×108
Marts et al., 2014b Mine E 6.99×109 2.03×109
31
|
Colorado School of Mines
|
Recent work by Marts et al., 2014b used FLAC3D to model the mining operation of
two mines following the work of Esterhuizen & Karacan, 2007, to validate the gob material
strength against the subsidence data provided by the mines. The values shown in Table 2.2
for Mine C (Marts et al., 2014b) represent a mine that is highly compacted, while the values
for Mine E represent a more loosely compacted gob. Table 2.2 presents a wide range of
values within an order of magnitude and illustrates the vast change in the gob properties
that must be considered for an effective ventilation design.
2.3 Ventilation Studies using Computational Fluid Dynamics
ThesectionpresentstheuseofCFDcodeinmineventilationresearch, someofthegeneral
ventilation problems that have been studied, and then details regarding the specific study of
gob ventilation. This discussion covers a brief overview of methane emission and the related
work on GVB performance, then the work on progressively sealed gob ventilation systems,
and finally tracer gas studies to determine ventilation network connectivity.
2.3.1 Codes used in General Ventilation Problems
Initial CFD studies in mining began in Japan by Uchino et al., 1980 with a coal mining
ventilation requiring coolers on the face air. The researchers used a one-tenth scale physical
model of the 100 m (330 ft) wide longwall panel to evaluate gob conditions and the effect
of air flow and patterns along the face, and headgate curtain effect on gas ingress simulated
on a two-dimensional representative mesh. As computational processing power increased
in the early 1990s, the focus shifted to face ventilation in mining development sections as
studied by Gong & Bhaskar, 1992 in a three-dimensional flow model of an active continuous
miner machine to predict the effects of exhausting, blowing, brattice, or tubing schemes of
auxiliary fans. Later, Banik et al., 1993 studied general flow models in natural mine airways
for friction flow coefficients estimated in mine entries and compared to standard pipe flow
models using surface roughness and Reynolds number to relate friction flow coefficients on a
Moodychart. Discrepancieswerefoundandattributedtothegreatrandomnessinplacement
32
|
Colorado School of Mines
|
and size of wall roughness values, but fractal analysis was suggested to correlate values for
mine airways.
2.3.2 Models used for Gob Ventilation Problems
Initial modeling of air flow in longwall gobs began with Ren et al., 1997 using a two-
dimensional Fluent model for the control of methane transport from the strata and gob
to the gate road airways. The gob properties were taken from a rock sample test using a
tri-axial stress condition to determine permeability and porosity changes of the overlying
strata to simulate the undermining of the longwall mining operation. Shown in Figure 2.8
are the results of static pressure and velocity contour plots of a vertical cross-section of the
strata and gateroads. The two gateroads can be seen in the lower left and right of the figures.
This begins to suggest the flow patterns that might be found in the gob. Wendt & Balusu,
2002 used CFD modeling to examine the effect of overburden and rock types on the flow
patterns in the gob, advancing the work of Karacan & Okandan, 2000.
Balusu et al., 2005a is the most comprehensive study using gas monitoring, as shown
in Figure 2.9, to analyze the effects of ventilation schemes, mine layouts, and GVB design
with respect to gob gas composition. In addition, tracer gas studies (see Section 2.3.6) from
two different mines, as shown in Figure 2.10, validate computer modeling efforts. A CFD
code and coupled geomechanical gas flow software, COSFLOW (Yeh et al., 1997), optimized
the ventilation schemes for the safety of methane-air mixtures in the ventilation system.
The methane is both collected as ventilation air methane, and coal mine methane through
GVBs as useful gas production. These two complete mine gob ventilation studies and their
simulation efforts comprise the earliest retrieved publications.
The largest CFD model of mine ventilation by Ren et al., 2011 used the mine geometry
shown in Figure 2.11, which shows an active face in the lower left-hand corner with several
sealed gobs. A CFD analysis of the area and the active panel using various ventilation
schemes were modeled. One of the results is shown in Figure 2.12, where a nitrogen injection
pointisplacedat200m(660ft)and80m(260ft)behindtheface. Theoxygenconcentration
33
|
Colorado School of Mines
|
Figure 2.10: Tracer Gas First Arrival Times at Different Locations in the Gob for Two Mines
is reduced through the gob area at 200 m (660 ft), but increases at the face when compared
with the 80 m (260 ft) placement injection point. The oxygen concentration reported in
these models is useful for reducing the potential for spontaneous combustion, however, they
fail to identify the EGZ location and size.
Worrall, 2012 used a gob gas composition algorithm to compute the explosive hazard
based on Coward’s Triangle, as shown previously in the gob gas analysis plot Figure 1.7.
The mine geometry used for the CFD study is taken at the recovery room with additional
ventilation to the face through two special entries, called recovery chutes. The recovery of
the longwall shields is modeled in a five-step removal process where the nitrogen injection is
optimized for each step. Other studies published by Marts et al., 2013 and Gilmore et al.,
2013, showtheeffectoffaceventilationqualityandnitrogeninjectionamountsinthetailgate
or headgate on the explosive volumes, tailgate concentrations of methane, and the oxygen
ingress as related to spontaneous combustion potential.
35
|
Colorado School of Mines
|
2.3.3 Methane Emission Studies
The primary source of methane emission in underground coal mines is from the distur-
bance of a gas reservoir in an overlying or underlying coal seam, and the in-situ methane
contained in the coalbed being mined may also be a significant source. The methane enters
the mine ventilation at the assumed concentration of 100%, diluting into the airways as
shown in Figure 2.13. Previous studies used models of an underlying gas reservoir by Balusu
et al., 2005b for Australian mines, and an overlying gas reservoir, used by Worrall, 2012.
Figure 2.13: Methane Liberation from a Feeder Crack into a Mine Ventilation Airway (not
to scale)(Kissell, 2006)
AccordingtoastudybySaghafiet al.,1997, theworldwidemethaneemissionfromunder-
ground coal mining is believed to be 30% of the human industrial contribution. The recovery
of useable methane is increasing with GVB use in ventilation designs. The production of
coal in highly gassy mines is predicted to produce up to 50 cubic meters of methane per
metric ton of coal mined. Using a comparison of production rates in Australian coal mines
for gassy and non-gassy mines, the authors used an empirical relationship to predict foreign
coal mines salable methane gas as shown in Table 2.3. The percentage of utilized methane
comes from GVBs, and the total is the emissions liberated as ventilation air methane, or
freely venting GVBs. The United States is the third highest emitter of methane, with less
than 10% utilization. As recently reported, Poland has improved recovery to 20%, and the
mines in Poland are some of the deepest in the world, with some as deep as 1,200 m (4,000 ft)
below the surface. Poland has achieved 20% recovery only with great efforts spent in the
removal of methane before and after mining (Uszko et al., 2013).
38
|
Colorado School of Mines
|
2.3.4 Gob Ventilation Borehole Studies
In a United States Bureau of Mines study by Diamond, 1994, methane drainage tech-
niques for control in underground mines were reviewed for pre-mining and post-mining
drainage to reduce the methane that enters the ventilation system. In Australia, Xue &
Balusu, 2002, studied designs for optimal GVB methane production. Flow simulations cou-
pled with geomechanical modeling of horizontal GVBs by Kelsey et al., 2003, suggest the
optimal spacing when drilling pre-mining drainage systems in Tower Colliery, UK.
MinesinthePittsburghcoalbedsarethesubjectofextensivestudiesinmethanedrainage,
geomechanical modeling, and flow simulations. Karacan et al., 2007b, studied the placement
and design of GVBs and suggested 60 m (200 ft) of slotted casing with a drilling depth to
just above the formation of the gob in the fractured zone. These recommendations were
based on flow simulation and geomechanical modeling verified by mine data from drill holes.
Also Karacan et al., 2007a found that the impacts of longwall panel width on emissions in
the GVBs result in further fracturing and production of methane. The strata layers above
a coalbed seam are highly influential in the production and release of methane as reported
by Karacan et al., 2008. Additional experimental data gathered by multi-rate drawdown
well tests are used to study the flow characteristics of coalbed methane reservoirs (Karacan,
2009c). Karacan’s work is applied in flow modeling of methane removal (Karacan, 2009d)
to prevent mine explosions. Coalbed methane reservoir life is simulated using intelligent
computing methods by training an algorithm with a partial data set and comparing the
predicted result with the actual well life span (Karacan, 2009a,b). The algorithm suggested
an optimum operating condition of future wells. Karacan & Luxbacher, 2010 used stochastic
modeling of the strata geological formation modeling the heterogeneous rock properties,
which can affect GVB performance.
The application of CFD modeling in optimizing the positioning and operation of GVBs
to reduce methane in Australian mines is published by Ren, 2009, suggesting the drilling of
GVBs in a cross-measure roof borehole to drain the tailgate corner more effectively. Down-
41
|
Colorado School of Mines
|
hole measurements for methane production by flow meters and cameras are conducted by
Wierzbicki, 2013 in Poland to better understand the rock fractures and methane producing
strata.
The removal of salable methane from GVBs as a production method and as a mitigation
strategy for mining coal has over a decade of profitability (Schraufnagel et al., 1994). The
full implementation of profitable methane sales for many coal mining operations is limited
by distance to the end user, and therefore methane is often freely vented or flared off instead
of being captured for sale.
2.3.5 Studies of Progressively Sealed Ventilation Systems
Ren et al., 2005 used Fluent to simulate the inertization practices in bleederless coal mine
gobs. The mine geometry for an active face, recovery room and the sealing of the panel were
modeled. The methane inlet is defined as a constant velocity inlet at 10 m (32.8 ft) below
the mine floor. The mesh is generally hexahedral with refinements near the face. The model
extended 1,000 m (3280 ft) inby and is 250 m (820 ft) wide using a “U-type” ventilation
system.
The nitrogen injection locations from a gas boiler at a rate of 0.5 m3/s (1060 cfm)
are optimized for the minimum amount of oxygen penetration into the gob. During active
mining,theeffectsoftwodifferentnitrogeninjectionlocationsaremodeled. Thefirstlocation
is just inby the face, and the second is three crosscuts inby the face. Figure 2.15 shows the
resulting oxygen contour plots. The near face injection may reduce the oxygen near the face,
but high levels of oxygen still remain inby to the second crosscut. Using the third crosscut
as the injection point reduces the penetration depth, and reduces the remaining gob oxygen
concentration to levels that will stop the self-heating process.
In the recovery room, the ventilation T-splits at the tailgate and the oxygen surrounding
the shields has breathable air quality, as shown in the base model in the top of Figure 2.16,
with no nitrogen injection. The following oxygen contour plots show the effect of nitrogen
injection at 30 m (98 ft), 110 m (360 ft), and 200 m (660 ft) behind the face. The oxygen
42
|
Colorado School of Mines
|
reduction on the headgate side is most significant with the 200 m (660 ft) nitrogen injection
location. The effects of the permeability are shown in the shape of steep oxygen gradients.
Further research by Ren & Balusu, 2009, on the effects of nitrogen injection in a GVBs
showed that if the tailgate becomes inaccessible, then a GVB may be used to inject nitrogen
to reduce the oxygen directly behind the face, sufficiently, to prevent spontaneous combus-
tion. Trevits et al., 2010 determined that 10 hours are needed to reach an inert atmosphere
below 10% oxygen using a 0.26 m3/s (560 scfm) flow of nitrogen injection after the longwall
shield recovery and sealing of the panel.
2.3.6 Tracer Gas Mine Ventilation Analysis
A tracer gas study examines gas flow communication between a release point and mea-
suring point along the gob or from GVBs. The gas of choice is sulfur hexafluoride (SF ),
6
a nontoxic, colorless, odorless, nonflammable gas, and easily detectable at extremely low
concentrations. In a study of the Pittsburgh coal seam by Diamond et al., 1999, tracer gas
is released from one GVB in-taking, and detected at GVBs further inby, suggesting commu-
nication between these GVBs. Also, tracer gas released into the gob generally remained in
the gob except when a GVB is shut down due to low methane levels. Tracer gas released
into the ventilation followed bleeder paths to the mine exit bleeder fan. Another study in
the Pittsburgh coal seam by Mucho et al., 2000, used tracer gas in an effort to understand
and improve GVB production and to limit the amount of methane liberated through the
bleeder-ventilated gob system.
Xu et al., 2013 used tracer gas release to build a CFD model based on the mine geometry
and predict the status of ventilation controls after a mine explosion. This modeling approach
could be useful when the mine is not accessible in other ways.
2.4 Chemical Reactions of Coal - Spontaneous Combustion
The self-heating of coal or spontaneous combustion can take place in the stockpile and in
the gob, when coal is left behind. This primarily occurs under conditions when the oxygen
44
|
Colorado School of Mines
|
concentration is high enough to sustain the reaction, and the air is stagnant enough that
the generated heat is not removed. The spontaneous combustion of coal has led to several
accidents resulting in the loss of production, equipment, and even the loss of life (Bessinger
et al., 2005; Grubb, 2008). The following list of events from recent years in Table 2.4 shows
the significance of spontaneous combustion events.
Table 2.4: Significant Spontaneous Combustion Events (Grubb, 2008)
Year Mine Consequences
1972 Box Flats (Australia) 18 Fatalities
(Cliff et al., 1996)
1975 Kianga (Australia) 13 Fatalities
(Cliff et al., 1996)
1991 Ulan (Australia) Loss of $60 million
(Cliff et al., 1996)
1994 Moura No. 2 (Australia) 11 Fatalities
(Cliff et al., 1996)
1997 Galatia (U.S.A.) Loss of $38 million
1997-1998 North Goonyella (Australia) Loss of longwall
1999 Sanborn Creek (U.S.A.) Mine idled 9 months
2000 West Elk (U.S.A.) Loss of $50 million
2003 Southland (New Zealand) Mine closed
2006 Dartbrook (Australia) Mine closed
2010-2011 Signal Peak (U.S.A.) Mine idled 9 months
2012 Oxbow (U.S.A.) Mine closed
The propensity for spontaneous combustion varies regional with coal deposits. Coal’s
tendency for self-heating is often determined from the coal’s classification or rank as deter-
mined by the carbon content, and the higher the rank the more prone it is to self-heating.
For example, Western United States coal mines often have a higher ranked coal and are more
prone to self-heating than Eastern United States mines. Based on Table 2.4, many mines
in Australia are highly prone to spontaneous combustion. When a mine has both a high
propensity for self-heating and is high in methane emissions, special attention is required for
the ventilation system design. The oxygen ingress concentration and its penetration distance
into the gob must be carefully controlled to prevent self-heating, and the methane released
46
|
Colorado School of Mines
|
into the face must avoid the explosive range. The risk of spontaneous combustion heightens
during slow mining or when a complete halt is required.
Studies by Banerjee, 1985; Cliff et al., 1996; Feng et al., 1973; Funkemeyer & Kock,
1989; Highton, 1979; Kaymakci & Didari, 2002; Mitchell, 1973 highlight the dangers of
insufficient cooling of spontaneous combustion prone coal via air velocity, described as the
critical velocity. Over a sufficient time frame, a lack of critical air velocity can initiate the
self-heating reaction. The reaction is known to slow at oxygen levels below 6%. However,
some coals can continue to react in as little as 2% oxygen (Highton, 1979).
Wang et al., 2003 studied the spontaneous combustion reaction pathway and the depen-
dence on the transport of oxygen to reaction sites within a coal particle. A variety of factors
have been determined to influence the possible reactivity of the coal such as composition,
history of weathering, particle size, temperature, partial pressure of oxygen and moisture
content. However, the carbon content or coal ranking is the only agreed upon factor for
propensity of self-heating.
The chemical pathways proposed for the reaction sequence of coal oxidation are shown
in Figure 2.17. The chemisorption sequence is known to occur, but falls short of the total
amounts of heat and gas produced. The overall production rates require another mechanism
of burn off, although the full extent is still not known. Modeling efforts are therefore limited
to only a few variables due to the complexity of this reaction sequence.
High levels of carbon monoxide are a known indicator of a heating event. NIOSH has
modeled a simplified one-step reaction using CFD (Yuan & Smith, 2006, 2008a,b,c, 2009a,b,
2010). The models use the following reaction pathway: Coal+O → CO +0.1CO+Heat
2 2
(Smith & Lazzara, 1987). The effects of nitrogen injection for Eastern United States coal
mines are modeled using progressively sealed gobs. This is the most comprehensive work on
spontaneous combustion modeling, but it does not explore the methane-air mixture results
inside the gob.
47
|
Colorado School of Mines
|
Figure 2.17: Illustration of the General Reaction Pathways Occurring in the Coal Oxidation
Wang et al., 2003
The classification of coal for self-heating is studied primarily by an adiabatic oven test,
the R test, with a pre-dried sample of coal. This process fails to take into account the
70
effects of water on the overall reaction, which can either help or hinder heating. Recent
research to introduce a new standard is discussed by Beamish & Beamish, 2011; Beamish
et al.,2013. ThetestingofseveralcoalsinFigure2.18showsthewiderangeoftimesthatcoal
exposed to oxygen takes to heat. The moisture inherent in the coal is shown to influence
heat significantly, as compared to a sample that is dried first. However, the researchers
could not draw a direct relationships to possible catalyst components such as pyrite content
to moisture.
The reactions for self-heating follow a set of stages. The first is the oxygen sorption
mechanism heating the coal to 70oC. This is followed by the production of hydrocarbon
such as benzene in the temperature range of 70oC to 150oC, where the release of combustion
products starts to occur. The accelerated heat generation begins from 150oC to 230oC, and
this transforms into full thermal runaway if the process is not stopped. During thermal
runaway, the coal content may contain sufficient oxygen or may be hot enough to liberate
oxygen from surrounding water sources to continue the spontaneous combustion process.
48
|
Colorado School of Mines
|
Figure 2.18: Typical Measurement Determining the Spontaneous Combustion Propensity
of Different Coals – Showing the Temperature Rise over Time for Several Types of Coal
(Beamish & Beamish, 2011)
Although some studies focus on spontaneous combustion reaction pathways for simple
modeling work, a specific mine ventilation system with the added variable of coal ranking
is not applicable to an industry useful general gas flow model. The concentration of oxygen
in the gob and the reduction of ingress distance are better suited for general CFD modeling
results and the ventilation design comparisons in this research.
2.5 Improvement of Previous Research
Other researchers have primarily focused on oxygen ingress or methane concentration in
the gob, and on limiting the application of the research to spontaneous combustion control
measures and tailgate methane control. This research provides the first look at the explosive
nature of bleeder-ventilated gobs and offers hazard mitigation strategies.
The modular mesh assembly approach developed for this project makes it possible to
study multiple mine ventilation layouts, efficiently and quickly. This approach also provides
fastersolutiontimesonlargerventilationnetworksthanpreviouslymodeledbyWorrall,2012,
49
|
Colorado School of Mines
|
CHAPTER 3
INTRODUCTION TO COMPUTATIONAL FLUID DYNAMICS
This chapter summarizes the equations used in fluid dynamics following the work of
White, 2010 and Tu et al., 2008 for the history and fluid physics derivations, and the FLU-
ENT Theory Guide, ANSYS, 2014b for the final equation form. The notation used through-
out follows that of ANSYS, 2014b when discrepancies between reference sources exist.
Fluid flow modeling began with Archimedes (285–212 B.C.E.) describing the laws of
buoyancy, and continued with the addition of one-dimensional conservation of mass by da
Vinci (1452–1519) defined in steady-state flow. Newton (1642–1727) postulated the laws of
motion and defined the law of viscosity in what is now called Newtonian fluids. Bernoulli
(1700–1782) made further developments in theoretical fluid motion with a principle describ-
inginviscidflowconservationnamedafterhim, Euler(1707–1783)explainedthemoregeneral
integrated form of the differential equations of motion. (White, 2010)
The vast difference between theory and experimental results drove engineers to develop
the science of hydraulics. Weber (1871–1951), Hagen (1797–1884), Poiseuille (1797–1869),
Darcy (1803–1858), Manning (1816–1897), and many others conducted foundational work
in this area. The experimental work by William (1810–1879) and Robert Froude (1846–
1924) codified the laws of scaling or similitude of modeling, and further work by Rayleigh
(1842–1919) developing of dimensional analysis built strong connections between experimen-
tal results and theory. Reynolds (1842–1912) expanded his predecessors work with a pipe
experiment showing that flow regimes can be characterized with his dimensionless Reynolds
number. The theoretical work continued to advance with developments by Navier (1785–
1836) and Stokes (1819–1903) studying viscous relationships in the differential form of the
equations of motion. At the time the Navier-Stokes equations were too complex, so their
solutions could not be easily applied.
51
|
Colorado School of Mines
|
Modern fluid mechanics began with Prandtl’s (1875–1953) division of flow into bulk fluid
motion and flow near a surface in a thin viscous region, called boundary layer theory. The
random motion of fluid in development of turbulent eddies had been observed in earlier
research, but was first modeled by Kolmogorov (1903–1987) who developed length scales in
1941. In 1972, Launder and Spalding first proposed the standard k−ε turbulent model.
Over the last four decades, there have been many refinements in modeling with the
increasinglyavailablecomputationalpowerandlinearsystemsolveralgorithms. Thischapter
presents a summary of the fluid transport equations used in this research.
3.1 Mathematical Notation
Themathematicalnotationofvectorsandindicesarepresentedinthissectiontofacilitate
the writing of fluid dynamics. The substantial derivative with respect to time, t, is defined
in Equation 3.1
DV ∂V
≡ +V·(∇V) (3.1)
Dt ∂t
where V is the velocity vector with three Cartesian coordinate components as given in
Equation 3.2
V =ue +ve +we (3.2)
x y z
where u, v, and w are the velocity scalar quantities in the unit direction vectors of e , e
x y
and e . This may also be written using Einstein index notation as
z
V = u (3.3)
i
where the subscript, i = x, y, z referring to each of the terms in Equation 3.2. This notation
is used to refer to a summation over all the coordinate components of a vector or used to
refer to all species involved in the formulation of a given solution. Therefore, Equation 3.1
expands into the following three Equations 3.4–3.6
Du ∂u ∂u ∂u ∂u
= +u +v +w (3.4)
Dt ∂t ∂x ∂y ∂z
52
|
Colorado School of Mines
|
∂ρ
+∇·(ρV) = 0 (3.13)
∂t
A physical meaning of each of the terms in Equation 3.13 can be identified as the change
in mass or density in the control volume due to temperature or phase fluctuation over time,
and the mass flux in and out of the control volume. Using the chain rule to expand the
second term in Equation 3.13 results in Equation 3.14
∂ρ
+ρ·∇V+V·∇ρ = 0 (3.14)
∂t
Using Equation 3.2 to replace each velocity vector gives the following full vector expansion
in Equation 3.15
∂ρ ∂ρ ∂ρ ∂ρ ∂u ∂v ∂w
+u +v +w +ρ +ρ +ρ = 0 (3.15)
∂t ∂x ∂y ∂z ∂x ∂y ∂z
Now simplifying this using the substantial derivative, Equation 3.1, with density as the
variable gives the condensed Cartesian coordinates mass conservation in Equation 3.16
(cid:18) (cid:19)
Dρ ∂u ∂v ∂w
+ρ + + = 0 (3.16)
Dt ∂x ∂y ∂z
For incompressible flows and steady-state fluid flows this can be simplified into Equation
3.17
∂u ∂v ∂w
+ + = 0 (3.17)
∂x ∂y ∂z
generally referred to as the incompressible continuity equation, or written using Einstein
index notation as Equation 3.18
∂u
i
= 0 (3.18)
∂x
i
3.3 Navier-Stokes Momentum Equation
The definition of the governing equations of motion start with Newton’s second law of
motion applied in a Cartesian coordinate reference frame as given in Equation 3.19
F = ma (3.19)
54
|
Colorado School of Mines
|
turbulentflowsinmostcases. Theadditionofrenormalizationgrouptheory(RNG)turbulent
modeling resolves a wider range of cases.
3.4.1 Description of k −ε RNG Models
The RNG turbulent model improvement over the standard k−ε model includes an ad-
ditional term in the formulation of turbulent energy dissipation that improves simulation
accuracy in rapidly strained flows and includes swirling flow modeling. RNG also provides
analytical formulation of turbulent Prandtl numbers compared to constant user specified
values in the standard k−ε model. Furthermore, the standard k−ε model is valid only in
high Reynolds number flows, while the RNG model provides differential formulation for an
effective viscosity in low Reynolds number flows. The use of RNG model is chosen over the
realizable turbulent model for this reason (Worrall, 2012).
The Prandtl number is define by the ratio of molecular diffusivity of momentum to heat
transfer as given by the following Equation 3.27
Molecular diffusivity of momentum υ µC
p
Pr = = = (3.27)
Molecular diffusivity of heat α k
T
where α is the thermal diffusivity or ratio of heat conductivity to thermal capacity ( kT ), k
ρCp T
is the thermal conductivity and C is the specific heat capacity.
p
The two-equation models use two transport equations for the definition of turbulent
kinetic energy k and turbulent dissipation ε. These two variable then calculate a turbulent
viscosity µ that is added to the laminar viscosity in the momentum Navier-Stokes Equations
t
3.23–3.25. The formulation of k and ε are two additional non-linear equations requiring
solutionseveryiteration. Theturbulencetransportequationsaretheturbulentkineticenergy
given in Equation 3.28
(cid:18) (cid:19) (cid:18) (cid:19)
∂(ρk) ∂(ρku) ∂(ρkv) ∂(ρkw) ∂ ∂k ∂ ∂k
+ + + = α µ + α µ
k eff k eff
∂t ∂x ∂y ∂z ∂x ∂x ∂y ∂y
(cid:18) (cid:19)
∂ ∂k
+ α µ +G +G −ρε−Y +S (3.28)
k eff k b M k
∂z ∂z
57
|
Colorado School of Mines
|
and the turbulent dissipation in Equation 3.29
(cid:18) (cid:19) (cid:18) (cid:19)
∂(ρε) ∂(ρεu) ∂(ρεv) ∂(ρεw) ∂ ∂ε ∂ ∂ε
+ + + = α µ + α µ
ε eff ε eff
∂t ∂x ∂y ∂z ∂x ∂x ∂y ∂y
∂ (cid:18) ∂ε(cid:19) ε ε2
+ α µ +C (G +C G )−C ρ −R +S (3.29)
ε eff 1ε k 3ε b 2ε ε ε
∂z ∂z k k
where G is the relationship between turbulence kinetic energy to the mean velocity gra-
k
dient given as G = µ S(cid:48)2 , S(cid:48) is the modulus of the mean rate of strain tensor given as
k t
√
S(cid:48) = 2S S , G is the generation of turbulent kinetic energy due to buoyancy, Y is the
ij ij b M
dilatation dissipation effect in compressible flows, which is neglected for incompressible tur-
bulent modeling, S and S are user defined source terms, α and α are the inverse effective
k ε k ε
Prandtl numbers and the constants are analytically derived as C = 1.42 and C = 1.68.
1ε 2ε
The effective viscosity given as µ =µ+µ is the result of the scale elimination procedure
eff t
in RNG theory defined by Equation 3.30
(cid:18) ρ2k (cid:19) µ /µ
eff
d √ = 1.72 d(µ /µ) (3.30)
(cid:113) eff
εµ
(µ /µ)3 −1+C
eff v
whereC isapproximately100. ThisallowstheRNGturbulencemodelwiththe“Differential
v
Viscosity Model” enabled to accurately predict low-Reynolds number. In the high-Reynolds
number flow limit the solution to Equation 3.30 gives Equation 3.31
k2
µ = ρC (3.31)
t µ
ε
where C is 0.0845, which is approximately the value use in the standard k−ε model.
µ
The RNG model in three-dimensional flows accounts for swirl in the mean flow with a
modification to the turbulent viscosity as given in Equation 3.32
(cid:18) (cid:19)
k
µ = µ ·f α ,Ω, (3.32)
t t0 s
ε
where µ is the turbulent viscosity without swirl modification, α is a swirl constant de-
t0 s
pendedontheflowbeingdominatedbyswirlandΩisacharacteristicswirlnumbercalculated
within Fluent.
58
|
Colorado School of Mines
|
3.4.3 Turbulence Boundary Conditions
The turbulent flow conditions at the model boundaries (i.e. inlets and outlets) must have
initial values based on the flow conditions entering the CFD domain to be computed or have
backflow conditions set when the flow re-enters the CFD domain. However, for the large
scale modeling of an underground mine, this is shown to be relatively insensitive to these
settings (Worrall, 2012). The k−ε model in Fluent uses one of three methods to define the
turbulent flow characteristics:
1. Specifying k and ε values directly
2. Turbulent Intensity, I, and hydraulic diameter
3. Turbulent Intensity and Length Scale, l
The first method is impossible to know without prior modeling results. The second is for
fully developed internal flows such as the mine entries, and the third can be used for external
flows. Fluent guidelines for estimating turbulent intensity is given by Equation 3.39
u(cid:48)
I ≡ ≈ 0.16(Re )−1/8 (3.39)
u
DH
avg
where u(cid:48) is the root mean squared of the average turbulent velocity fluctuation and u is
avg
the mean flow velocity. The length scale is define by l = 0.07L, where L is the relevant
geometric flow length that for this research is the hydraulic diameter of the entries.
3.5 Species Conservation
Molecular species transport is governed by a conservation law stating that species cannot
becreatedordestroyed. However,chemicalreactionswithinacontrolvolumemaychangethe
concentrationofdifferentspeciesfromonetothenextfollowingelementconservationbetween
species. In this research project, chemical reactions between species are not considered as
there is no combustion occurring in the gob. Therefore, modeling is only concerned with
species transportation. On a mass fraction Y basis the species conservation is given in
i
Equation 3.40
60
|
Colorado School of Mines
|
For turbulent flows the mass diffusion is given by the following Equation 3.43
∇T
J = −D ∇Y −D (3.43)
i t,eff i T,i
T
where in the standard and realizable k−ε models the effective turbulent diffusion coefficient
D is given by ρD + µt where D is the mass diffusion coefficient for species i in the
t,eff i,m Sct i,m
mixture and Sc is the turbulent Schmidt number defined as µt where D is the turbulent
t ρDt t
diffusivity. The Fluent default turbulent Schmidt number is 0.7. Turbulent diffusion is
orders of magnitude greater than laminar diffusion in turbulent flows and the details of full
multicomponent diffusion are often unnecessary to model.
The RNG turbulent model uses a difference formulation for the effective turbulent diffu-
sion coefficient given in Equation 3.44
D = αρµ (3.44)
t.eff eff
where α is calculated using Equation 3.33 with the value of α = 1/Sc where Sc is the
0
molecule Schmidt number.
3.6 Energy Transport
Theenergytransportequationisnotsolvedduetoconstanttemperatureboundarycondi-
tions and no significant temperature gradients appearing in the solutions when it is enabled.
Also, in the Fluent pressure-based solver the viscous heating term τ the deviatoric stress
ij,eff
tensor is not computed by default.
3.7 Treatment of Porous Media
The flow through porous media as modeled in Fluent is an empirically determined re-
sistance or momentum sink that is added to the governing momentum Equations 3.23–3.25.
The physical presence of the material producing the porous media is not represented in Flu-
ent. Therefore, a superficial velocity, which is based on the volumetric flow rate, is used
in the convection and diffusion terms in the momentum equations. This formulation may
be changed by using the physical velocity porous media model, which may produce more
62
|
Colorado School of Mines
|
accurate results when velocity values and gradients are important. The superficial veloc-
ity porous formulation generally provides a sufficiently accurate representation of the bulk
pressure loss needed for flow evaluation. A unique pressure interpolation scheme is always
used in the porous media zone, which calculates the necessary pressures at the faces for the
transport equations (see Section 5.2.2).
The porous media momentum source term contains two terms. The first term, Darcy’s
Law is a viscous loss term linear with velocity, and the second term, varies with the square
of velocity as given in the following Equation 3.45
(cid:32) (cid:33)
3 3
(cid:88) (cid:88) 1
S = − D µv + C ρ|v|v (3.45)
i ij j ij j
2
j=1 j=1
where S is the source term for the i-th direction momentum equation, |v| is the velocity
i
magnitude, D is a user defined prescribed matrix for viscous resistance and C is a user
ij ij
defined prescribed matrix for inertia resistance. The porous media source term contributes
to the pressure gradient in relationship to the fluid velocity.
In the simple homogeneous case Equation 3.45 becomes Equation 3.46
(cid:18) (cid:19)
µ 1
S = − v +C ρ|v|v (3.46)
i i 2 i
K 2
where K is the permeability and C is the inertial resistance factor. Fluent uses the input in
2
cell resistance, which is the inverse of permeability. The value of K is calculated by Equation
2.15 and discussed further in Section 4.3.
3.8 Solvers
Fluent has two numerical solvers available: pressure-based and density-based. The
pressure-based solver was initially developed for low-speed incompressible flows, while the
density-based solver was better suited for high-speed compressible flows. Currently, both
handle a wider range of flow problems.
The final solution of the velocity field using either solver is obtained from the momentum
equations. The density-based solver obtains the density field from the continuity equation
63
|
Colorado School of Mines
|
and then the pressure field from an equation of state such as the ideal gas law. The pressure-
based solver obtains the pressure field from a pressure correction equation, which is derived
fromthecombinationofthecontinuityandmomentumequation. Bothsolvers,thencalculate
the solutions to other scalar equation such as turbulence, species and energy equations.
The Fluent solver uses the following steps in a control volume based approach:
(cid:136) Mesh or grid creation of discrete control volumes dividing the domain into cells (see
Chapter 6)
(cid:136) Governing equations integrated over the cells creating algebraic discrete dependent
variables (velocity, pressure, temperature, species, turbulence k−ε)
(cid:136) Discretization of equations creating a linear equation system solved for the updated
values of the variables (see Section 5.2)
The choice of using the pressure-based solver over the density-based solver is clear as
the flow velocity is expected to be sub-sonic and dominated by flow in the gob, which is
primarily driven by Darcy’s law (pressure differentially driven flow).
3.8.1 Pressure-Based Solver
The pressure-based solver algorithm constrains mass conservation of the velocity field
variable through the use of a pressure correction equation. The pressure correction equation
derivedfromthegoverningequationsofcontinuityandmomentumhasthesolutionsuchthat
the velocity field when corrected by the pressure must satisfy the continuity equation. The
governing equations are nonlinear in nature and coupled, and therefore a solution proceeds
by iteratively solving the governing equations until the largest change per iteration reaches a
required minimum. This change per iteration is termed a residual, and a converged solution
is said to be reached when at least a three-order of magnitude drop is achieved. There are
two options for the pressure-based solver algorithms: segregated and coupled.
The segregated algorithm solves the governing equations in a sequential fashion. Thus,
being memory efficient and relatively fast per iteration, but slower to solution convergence
64
|
Colorado School of Mines
|
due to the decoupling of the governing equations. Figure 3.1 shows the segregated solution
process in the following steps:
1. Update the fluid cell properties from a current solution or initialized values
2. Solve the momentum equations in sequence using updated mass fluxes at the face and
pressure values
3. Solve the pressure correction equation using the velocity field and mass fluxes from
step 2
4. Apply the pressure correction solution to correct the mass flux, pressure and velocity
field
5. Solve other governing equations such as energy, turbulence and species
6. Check if the equation residuals meet the set convergence criteria
The coupled algorithm solves the momentum and the pressure-based continuity equation
in a single coupled step, thus, replacing steps 2 and 3 in the segregated solver process with
a single step as shown in Figure 3.1. The solver then continues to solve the other governing
equations. The coupled solver may converge in fewer iterations with increased time per
iteration due to the coupled solver step. In addition, the memory storage requirement
increases 1.5 to 2 times the segregated solver usage due to solving the velocity and pressure
fields simultaneously. Improvements in iteration times is shown to be beneficial as published
by Gilmore et al., 2015a with the use of GPGPU processing units.
3.9 Domain Meshing
CFD requires discretization of the fluid domain into control volumes or cells. This is
considered by many to be the most challenging part of the modeling process, since as the
creation of the mesh impacts solution stability, iteration time and the resolution of gradients
of the transport equations. Therefore, many mesh quality reporting statistics are available,
65
|
Colorado School of Mines
|
describing the cells in the mesh. These qualities include: cell quality, orthogonal quality,
skewness, aspect ratio, warp angle and smoothness. These may also have a dependence on
the flow field in the resulting solution. The guidelines from ANSYS recommend the use of
the statistics of quality and skewness as reported by ANSYSfi MeshingTM software.
The cell quality is a composite quality metric ranging from 0 to 1, where a value of 0 is
a poorly formed cell and a value of 1 is a perfectly formed cell. For three-dimensional cells,
this mesh statistic is computed using Equation 3.47
volume
Quality = C(cid:113) (3.47)
(cid:80) (edgelength2)3
where the value of C is selected for the cell type found in Table 3.1. The ANSYS meshing
guidelines recommend a minimum cell quality greater than 0.01 for all individual cells, with
an overall average much higher.
Table 3.1: Values of C for Computing Quality
Element Type Value of C
Tetrahedron 124.70765802
Wedge 62.35382905
Pyramid 96
The cell skewness is a mesh metric based on either the optimal cell size or a normalized
angle. The method for calculating the cell skewness is chosen to match the type of cell.
For example, Figure 3.3 shows the ideal skewness for a triangle and quadrilateral cell, and
the corresponding highly skewed cell for each. The range of values for skewness is listed in
Table 3.2 with their cell quality ranking. ANSYS Meshing guidelines recommend a skewness
below 0.95.
For example, for triangles and tetrahedral cells the equilateral volume method is calcu-
lated by using Equation 3.48
OptimalCellSize−CellSize
Skewness = (3.48)
OptimalCellSize
67
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.