University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Virginia Tech | CHAPTER 1:
GE(cid:6)ERAL I(cid:6)TRODUCTIO(cid:6)
1.1 Preamble
Coal is a major energy resource in both the United States and world wide. The extraction
of this resource from the earth requires methods that often mix impurities that do not contain
heating value in with the valuable coal. Removal of this excess material requires that the coal be
washed in a preparation plant to separate the clean coal from the refuse material. Most known
methods for accomplishing this for the fine particle size range (less than 1 mm) are wet
processes. Water, when mixed with coal, also reduces its BTU (British Thermal Unit) value, thus
creating a need to remove it from the final product.
Traditionally, coal producers have handled fine waste slurries in two different ways. The
first is to pump the slurry to an impoundment pond where the valuable coal is wasted and creates
an environmental hazard. This is done because it becomes increasingly difficult and costly to
dewater coal as particle size decreases. The other option is to install thermal dryers. Thermal
dryers, however, have high capital and operating costs. It is also difficult to get a permit to install
thermal dryers due to air pollution problems.
Thus, there is a need for a third option. An attractive alternative would involve the use of
a filtration unit that greatly reduces the cost and improves the effectiveness of fine coal
dewatering. This cost is reduced because the energy input needed to evaporate water is high
compared to the cost of mechanically removing water. The use of displacement filters combined
with surface tension reducing reagents can create a high volume filtration unit that would allow
for the economical recovery of these fine coal slurries. Unfortunately, a piece of equipment that
takes advantage of all of these factors in a cost effective way does not currently exist.
1 |
Virginia Tech | 1.2 Literature Review
1.2.1 Definition of Filtration
Filtration is the separation of two forms of matter by passing one of them through a
porous medium. This study is limited to the principles relating to the separation between solid
particles and liquids. In addition, the primary liquid that is being separated in this case is water,
allowing the distinction to be made that this is a dewatering process.
1.2.2 Filtration Methods
Dewatering is primarily concerned with removing water that is bound in the capillaries
between solid particles. This can be achieved using three basic known methods. The first is to
compact the slurry by adding stress to the perimeter of the slurry with a roller, pistons, screw
presses, or other mechanical means. The second method is to displace the water using a gas that
is applied through either vacuum or pressure. The third method can be achieved by adding an
electrical field to a slurry of charged particles. Each of the methods has its advantages and
limitations.
Dewatering that uses mechanical compression is limited to the compressibility of the
filter cake. Once the cake has been compressed to the point of virtual incompressibility, no
further dewatering is possible since the water has been fully compacted into the capillaries and
pores.
In dewatering by gas displacement, the important factors are the pressure difference
across the filter cake and the kinetic dewatering characteristics. This process results in an
exchange between water volume and gas volume in the filter cake.
3 |
Virginia Tech | Dewatering using electrical fields creates charged ions to generate the pressure required
to force the water out of the capillaries. This process is called electro-osmosis. The primary
difference in this method of dewatering is that rather than the driving force being generated from
outside the capillaries, it is formed from the electrical forces within the capillaries.
1.2.3 Filtration Theory
Scientists and engineers have been working to develop the fundamental theory behind
filtration for over 150 years. Despite this large time frame and much attention, the theory
regarding filtration is not complete. Most of the work that has been done is in regard to cake
formation rates. The ability to predict the moisture content of a cake is beyond the current
knowledge of filtration theory. Most work that has been done on this topic is empirical in nature.
The filtration theory that is known is an extension of fluid mechanics. The rate of
filtration is directly proportional to the pressure drop across the cake and inversely proportional
to the resistance. It is known that the pore sizes in the filter cake are small and the liquid’s
velocity through the cake is also slow, the filtrate flow maybe considered laminar. Since these
assumptions can be made Poiseuille’s law is applicable. From this it can be seen that the variable
rate of filtration may be expressed as:
dV
u = (1)
Adt
where V = volume of filtrate
A = area of filtration
t = time
u = filtration rate
4 |
Virginia Tech | By making the laminar flow assumption and by applying Poiseuille’s Law the basic filtration
equation, it is possible to show that:
dV ∆P
=
( ) (2)
Adt µawV +r
A
where ∆P = pressure drop across the filter (including the cloth and drainage system)
µ = liquid viscosity
a = specific cake resistance
w = weight of the dry cake solids per unit volume of filtrate
r = resistance of the filter cloth and drainage system
Rearranging Equation (2) yields,
dV dV
µawV +r = ∆P dt (3)
A2 A
It can then be assumed for a thick cake that r ≈ 0 since the resistance of the cake is much greater
than that of the filter cloth and drainage system. This expression can be integrated between the
limits of 0 and t (where t is the cake formation time per cycle) and 0 and V where V is volume
f f f f
of filtrate in time t. From these limits, it is possible to show that:
f
V VdV t
µaw∫ f = ∆P∫ f d t (4)
0 A2 0
The result of the integration gives:
5 |
Virginia Tech | µaw
V2 =∆Pt (5)
2A2 f f
When both sides are divided by t2, rearranged, and square root taken, it can be shown that:
f
V 2∆P
f = (6)
At µawt
f f
The final step is to get the equation in the form Z . Z is measured in weight of dry solids per unit
s s
area per unit time of cake formation. If t is the time involved in the cake formation, this variation
f
of the equation is realized by multiplying Equation (6) by w, i.e.:
wV 2w∆P
Z = f = (7)
s At µat
f f
This expression shows that the cake production rate is proportional to the square root of the
pressure and inversely proportional to the square root of time.
1.2.4 Cake Filtration Using Displacement
The most critical aspect of any filtration process is the source of the driving force. The
importance of this force is shown in Equation (2), the fundamental relationship for filtration. In
that equation, ∆P is a critical term and this pressure difference is often referred to as the driving
force. In all of the equipment that was studied and analyzed, the driving force is provided by air
pressure applied to the filter cake. The filtration theory states that the higher the ∆P the greater
will be the cake formation rate. Other empirical evidence also shows that the higher the ∆P the
lower will be the final cake retained moisture (Cheremisinoff, 1995).
6 |
Virginia Tech | 1.2.5 Filtration Equipment
There are several types of equipment that are commercially available for fine particle
dewatering. The most prolific of these are disc filters, drum filters, and belt filters. All of these
units are currently used in a vacuum mode and some are used in a pressure mode. Each of these
units has some advantages over the others. There are other novel pieces of dewatering equipment
that are made in various places, but these three are the most prevalent.
1.2.5.1 Disc Filters
Disc filters are used in both a vacuum mode and a pressure mode. Pressurized disc filters
work on the same principal only the disks are contained within a pressure vessel. Disc filters use
a multitude of rotary discs that are vertically mounted on a horizontal shaft, which are suspended
in a slurry reservoir. This has the advantage of having greater floor space utilization than other
filters. Each disc is divided into separate sectors often ranging from 12 to 30 per disc. The disc
must be mostly submerged because cake formation occurs while the sector is still submerged in
the slurry.
Disc filters do have some major disadvantages due to the vertical nature of the disks this
limits the thickness of the cake since too thick of a cake will not stick to the filter. This limits the
capacity for the filters. Also, it is difficult to wash the filter media thus plugging can be a major
problem. Since the cake is submerged through much of the rotating cycle of the filter it can result
in too high moisture content due to too short of dewatering time.
1.2.5.2 Drum Filters
Drum filters are the most commonly used air displacement filter. This is a device that has
a large drum that rotates slowly about the horizontal axis. The drum is similar to the disc filter in
that part of the drum is submerged in the slurry. The drum has multiple sections, each containing
7 |
Virginia Tech | a separate vacuum chamber. This filter like the disc filters can be enclosed in a pressure vessel
and operated under pressure. The disadvantage of the drum filter is the relatively poor use of
space. The entire volume of the drum contributes little to the dewatering process and yet
consumes much space. Otherwise, the operating principal of the drum filter is similar to the disc
filter.
1.2.5.3 Belt Filters
Horizontal belt filters are the third prevalent type of displacement filtration equipment.
The operation of the horizontal belt filter is quite different from that of the disc filter or drum
filter. In this unit, a filter belt is continuously rolled across two rollers much like a conveyor belt.
The slurry is feed onto one end and then slowly turned across the length of the belt. While
traversing the length of the belt, a vacuum is applied to the underside of the belt. This type of
unit has an advantage of lower capitol costs than the other filters and it has a high area available
for dewatering. Also, the speed and feed rate of the belt can be easily controlled. The primary
disadvantage is that the unit has a poor dewatering rate per square footage of floor space. These
units are not made in a pressurized version at this time.
8 |
Virginia Tech | 1.3 Objectives
A variety of mechanical processes are available for dewatering fine particles in the coal
and mineral processing industries. Most of these units have major problems associated with
them. All of the units currently available suffer from one or more of a number of difficulties,
such as poor dewatering performance, low throughput capacity, high capital costs, and high
operating costs. Primarily, the vacuum units suffer from the poor performance and low capacity,
while the pressurized units suffer from high capital and operating costs.
The first aspect of this research was to develop a new type of dewatering process that
combines the operational flexibility of a continuous horizontal belt filter with the dewatering
efficiency of a batch pressure filter. As such, the project involved the design, construction,
testing, and evaluation of a pilot-scale prototype unit. The prototype hyperbaric belt filter can be
used as a model to assist in the design of a full-size commercial unit. The project demonstrated
that the unit is physically feasible and provided critical operating information such as throughput
capacity, moisture capabilities, power requirements, etc.
9 |
Virginia Tech | CHAPTER 2:
DEVELOPME(cid:6)T OF A HYPERBARIC BELT FILTER
2.1 Introduction
The final stage of a coal processing operation involves the separation of liquid water from
the useful coal particles. For very fine material, this is commonly done using a vacuum filtration
method, a practice that is inherently limiting. One of the primary factors in any filtration process
is the pressure drop across the filter cake. Vacuum methods are limited to a theoretical maximum
of one atmosphere of pressure. By utilizing pressure filtration, the amount of pressure can be
increased much higher.
Research was done to determine how effective pressure filtration would be on fine
particle dewatering of coal. The ultimate goal is to develop a process that, inexpensively, is
effective at reducing cake moisture content to levels that are currently unattainable in coal
processing equipment.
The mining industry is infamous for handling massive amounts of material. To
accommodate this inherent requirement in the industry it is critical that any filtration device have
a large capacity to be of any practical use. This large capacity is one of the major shortcomings
of vacuum filters. Although it is possible to create a high capacity vacuum filter, this is
impractical due to the massive size that would be required. This is due to the low pressure
differential that causes cake formation times to be long and extended dry cycle times. To
understand the relationship between cake formation time and various operating parameters is
imperative to the design of a filter unit. The higher the cake formation rate is, the higher the
capacity of the full size unit. This overall dewatering capacity is of extreme importance to the
success of the project.
10 |
Virginia Tech | 2.2 Hyperbaric Belt Filter Design
The first step in the design process was to determine what type of filter was going to be
developed to dewater coal. It was determined, that currently, a pressurized belt filter does not
exist. This unit has unique advantages over other types of pressure filters. The primary advantage
is that it has the potential to have high capacity since the length of the belt can be increased and
yet run at a relatively high velocity.
2.2.1 Concept of the Design
Once the decision was made to design a hyperbaric filter, the specific details of the
design needed to be worked out. One of the objectives was to make a low cost unit. To this end,
the basic components were considered. The lab tests showed that for the unit to physically do
what was necessary, the top of the belt would need to be exposed to positive pressure while the
bottom of the belt would need be open to atmospheric pressure. This pressure difference across
the belt is the key item to a pressure filter. All of the parts of the unit are designed to allow this to
occur. Figure 2.1 shows what a pressure difference across the belt looks like.
The most feasible approach to achieve a large pressure differential across a moving belt
was one that had the entire filter device enclosed in a pressure vessel. That would create the large
positive pressure on the top of the belt. To support the pressure differential across the belt, a rigid
pressure plate was devised that was capable of handling a large force load.
The other major requirement that this project set out to overcome is that the pressure
filter had to be a continuously operating unit. This is something that virtually all of the existing
pressure filters are not. Most of them are batch processes that simulate continuous operation by
having multiple units operating side by side. This creates numerous difficulties by presenting a
11 |
Virginia Tech | Figure 2.1. Pressure differential across a horizontal belt filter.
situation where material has to both enter and leave a pressurized chamber while maintaining a
pressurized environment.
2.2.2 Design of the Filter Device
From the general concept of having a belt filter contained in a pressure vessel, it is clear
that the project should be separated into two major designs, the filter device and the pressure
vessel. Each of the parts of the filter device is discussed in detail in the following sections.
2.2.2.1 Frame
The purpose of the frame is to support all of the other components that are required to
make the unit operate. This was designed to be as inexpensive as possible while still minimizing
weight and still having the strength characteristics that would be required. The frame was
constructed out of carbon steel tubing. Figure 2.2 provides a scale technical drawing of the filter
device frame. Additionally, the frame had to be conducive to material handling for maintenance
purposes. For this purpose, hooks were added to allow the unit to be handled using an overhead
12 |
Virginia Tech | Figure 2.2. Design of the filter device frame.
crane. The unit also has wheels that can roll along a track to allow easy access to the unit inside
the pressure filter. Figure 2.3 shows the track wheel mounted to the filter device frame.
To assist in the handling of the unit an a-frame attachment was made to fit into the end of
the support structure to assist in removing the filter from the pressure vessel. Figure 2.4 is a
picture of the A-frame structure inserted into the end of the filter device to support the structure
when removing it from the pressure vessel.
2.2.2.2 Filter Cloth
The filter cloth is the medium that
allows the water to pass while holding most of
the solids on the top of the belt. The filter
cloth that is being used on the unit has
openings of 80 microns. Thus, any material
that is larger than 80 microns is retained on the
filter belt as is a large percentage of finer
Figure 2.3. Wheel on the filter device.
material. The pressure differential across this
13 |
Virginia Tech | Figure 2.4, A-frame attachment for the filter device.
medium provides the driving force that allows for the filtration to occur.
One of the major concerns of the filter cloth is that over time the cloth would become
blinded (plugged) with near size particles. This would cause a situation where water would not
be allowed to pass through the cloth and no more filtration would be possible. To prevent this, a
high pressure water spray was installed to force water through the back side of the filter cloth to
push out any trapped particles. The water spray is located just before the fresh feed is put on the
belt to ensure that each run of slurry has a clean cloth to filter through.
Another design concern that had to be addressed in regard to the filter cloth is the
tensioning of the filter cloth. This tensioning had to be done separate of the rubber belt so that
the water spray would be able to spray the back side of the belt to effectively remove any trapped
14 |
Virginia Tech | Figure 2.5. Tensioning roller and water spray on the filtering device.
particles. This tensioning is done just behind the drive roller. The roller in the center of Figure
2.5 is the tensioning roller.
The tension is provided by a pair of springs that pull an idler roller back to create the
tension on the filter cloth. The filter cloth is in contact with the rubber belt during its entire travel
around the non-drive primary roller and the length of the dewatering stage as well as the trip to
the drive roller. This means that to tension the filter cloth and allow washing, the amount of
distance the filter cloth is apart from the rubber belt is minimized.
To keep the slurry from running off the feed end of the filter, a trough was built to
provide a reservoir of slurry to ensure continuous feed to the dewatering section of the belt. This
was accomplished using two Teflon guiders for the filter cloth and a roller. A picture of this
apparatus is shown in Figure 2.6.
15 |
Virginia Tech | Figure 2.6. Trough apparatus used to hold slurry on the filter cloth.
2.2.2.3 Pressure Plate
The pressure plate is the part of the filter device that creates the seal necessary to separate
the high pressure region from the low (atmospheric) pressure region. The pressure plate consists
of an aluminum plate, which has had some machine work done to it, along with a steel box that
collects the water. Figure 2.7 shows an isometric view of the pressure plate.
The plate has a slot machined down the center that allows the notch in the rubber belt to
fit inside of it. Also, there are holes drilled in a slightly lower slot to allow the water and air to
pass through both the pressure plate and the rubber belt. At the ends of the pressure plate, two ¾
inch rubber strips are mounted. The rubber strips also have a slot machined in them to help seal
the ends of the belt and pressure plate as the belt travels across the plate. Also, mounted to the
pressure plate are two V-shaped plastic strips that help the length of the belt to seal. Commercial
16 |
Virginia Tech | Figure 2.7. Design used for the pressure plate.
weather stripping attached with silicone was found to be ideally suited for sealing the underside
of the belt (see Figure 2.8).
On the bottom of the pressure plate is the discharge box that collects the air and water in
a central location and sends it to a drain pipe to leave the pressure vessel. The pressure plate is
attached to the frame by four one-inch diameter bolts that have a ten-inch adjustment on them.
This allows the pressure plate to be pushed against the belt with varying degrees of force. Figure
2.9 shows how the bolts are screwed through four steel blocks that are welded to the frame.
2.2.2.4 Belt and Drive System
One of the key aspects of the filtration device is the rubber belt that carries the filter cloth
over the pressure plate. The rubber belt provides a material with the strength properties needed to
take the filter cloth across the pressure plate. The rubber belt supplies the drive to overcome the
force that the pressure differential creates to hold the filter cloth in place. The belt has a notch in
17 |
Virginia Tech | Figure 2.8. Plastic strips used to seal the Figure 2.9. Bolts attaching frame to the
underside of the belt. pressure plate.
the center that contains holes for the air and water to pass through. Also, the belt has slots across
it that allow for the pressure to be applied across the entire width of the belt. These attributes are
shown in Figures 2.10 and 2.11. The belt has two raised angular rubber pieces on each side to
prevent the slurry and cake from falling off the sides of the belt.
The rubber belt is mounted on two nylon rollers. The drive roller is on the same side as
the feed is put on the belt. This roller is chain driven by a DC electric motor that has been speed
reduced to increase the torque needed to pull the belt across the pressure plate. Both of the nylon
rollers are mounted on the frame by two pillow block ball bearings. A diagram of one of the
pillow block bearings used to support the two nylon rollers is shown in Figure 2.12.
2.2.2.5 Filter Cloth Scraper
After studying a lab batch unit that was developed at Virginia Tech, it was determined
that to remove the filter cake from the filter cloth after all of the dewatering has been completed
18 |
Virginia Tech | The other issue that had to be
overcome was how to position the scraper in
an orientation that the cake would not
excessively build up on the scraper. This
was accomplished by scraping at an angle
with two separate scrapers to cover the entire
width of the belt. Figure 2.13 shows the
scrapers used to remove fine material Figure 2.12. Pillow block bearing for the
primary drive.
coming off of the filter device. Each of the
scrapers is adjustable to change the amount
of force with which each of the scrapers
presses against the belt. The angle of scrape
can be adjusted also.
2.2.3 Pressure Vessel Design
The pressure vessel was designed
with two primary concerns in mind. The first
was to construct a structure that could safely
hold 60 PSI of air pressure for a long period
of time. The second concern that had to be
addressed was it must be sufficiently large to
contain the filtration apparatus. Both of these
conditions had to be addressed before the
Figure 2.13. Scrapers used to remove fines
from the filter cloth.
smaller technical issues could be resolved.
20 |
Virginia Tech | To form the body of the pressure vessel, a cylindrical steel pipe made the most sense.
First, it would be easy to find a pipe that was rated to hold more than 60 PSI of pressure. The
second reason for using a steel pipe would be that it is easy to attach accessories to it since it can
be welded. The final motivation for using steel is that it is relatively inexpensive. The only major
drawbacks to using steel for the pressure vessel is that it is heavy and that there is the potential
for corrosion due to the wet environment that exists on the inside of the unit. The other type of
material that was seriously considered to form the pressure filter was polyvinyl chloride (PVC).
This type of pipe is less accessible but is still widely manufactured. PVC is lighter than steel and
has better corrosion resistance. It is more expensive and it would be difficult to cap off the ends.
Also, adding a structure to the inside of the pipe to hold the filtration apparatus would present a
challenge. After much consideration, the decision was made to use steel pipe. The steel pipe that
was used is a schedule 40 pipe that is rated to hold 150 PSI. Figure 2.14 shows the steel pipe in
the condition it was in when it was received.
Figure 2.14. Pipe used to construct the pressure chamber (as received).
21 |
Virginia Tech | The second major concern was to get a vessel that would be large enough to hold the
filtration apparatus. From the design specifications it was determined that the smallest circle that
could practically hold the entire unit would have to be 30 inches in diameter. From this plan the
pipe that was acquired is 30 inches in diameter and 10 feet long.
2.2.3.1 Supporting Frame
Before the pressure vessel could be assembled, a frame had to be built to hold the vessel
and filtration apparatus. This frame was built out of tubular steel. Figure 2.15 shows the
dimensions on the tubular steel that was used to construct the frame shown in Figure 2.16. To
support the weight of the pressure vessel and all of its components, it was decided that a simple
box frame would be sufficient to hold the pressure vessel. This frame was welded together and
Figure 2.15. Steel used to construct the Figure 2.16. Schematic of the primary
main frame. support frame.
22 |
Virginia Tech | 2.2.3.3 Feed System
The pressure vessel had to allow for material to be continuously fed to the filter device.
Since the input stream to the filter is slurry the only option was to pipe in a line and to pump the
slurry onto the filter belt. The main challenge that this presented was that the filter device and
pressure vessel where designed to allow operation at various pressures. This means that a
normal centrifugal pump could be optimized to create flow at a given pressure but it would be
hard to operate the same pump under different conditions. Also, most centrifugal pumps could
not produce the pressure required to overcome the back pressure caused by the pressurized tank
at a low feed rate. These deficiencies lead to the search for some form of positive displacement
pump that could suit the requirements. Since a positive displacement pump could function and
get various feed rates with little reaction to the changing back pressure caused by the air. A
progressive cavity pump seemed to be the best option for the combination of factors that this
situation required. It could generate a low flow rate pulse less flow against high pressure and
could handle abrasive slurry. Figure 2.19 shows a picture of the progressive cavity pump that
was used with this hyperbaric belt filter unit.
2.2.3.4 Discharge System
The discharge system is required to allow a large amount of solids to exit the pressure
vessel and to not lose air pressure. For this system, two basic concepts were considered. The first
was to use a rotary star valve. This is a valve that has several compartments in it that rotate
around an axis. While some of the compartments are exposed to the inside of the pressure vessel,
others are facing the outside allowing the material to pass but not depressurizing the vessel. This
system has the advantage of only requiring one valve. The disadvantage is that these valves are
25 |
Virginia Tech | 2.3.2 Pressure Vessel Construction
The pressure vessel portion of the unit took longer to construct. The construction began
with making the frame to hold the pressure vessel. Then, the pipe was placed on the frame. Once
on the frame the rail system was added to the inside of the pressure vessel. Then, there was a
long break in construction while bids were taken for some work to be done by outside
contractors. This work included welding on the pipe fittings for all of the input and output
system on the vessel. Also, welded on were the flanges for the end caps and knife gate valves.
Once this was done, the end caps were put on and the unit was hydrostatically tested at 100 PSI
for one hour. The unit was then returned to the plantation road facilities and painted with water
resistant epoxy paint. Figure 2.24 shows a picture of the fully assembled pressure vessel.
Figure 2.24. Completely assembled pressure vessel.
30 |
Virginia Tech | 2.4 Shake-Down Testing
When construction of the unit was completed it was necessary to test the hyperbaric belt
filter for its mechanical operation. These tests lead to two primary discoveries.
2.4.1 Air Requirement
The first operational discovery was that for the unit to operate effectively it would need
more volume of air than could be provided by the air compressor at plantation road. The lack of
an appropriate air compressor, lead to renting a 185 CFM unit from RSC Rental in
Christiansburg, Virginia. Once the air requirement for the unit was met the unit began to perform
as expected.
2.4.2 Start-Up Procedure
The second important discovery during shake-down testing was that the unit like most
complex machinery has a start up procedure that must be followed to ensure that the unit will
perform as expected. The steps to properly start up the unit are listed below.
1. Check to see that all of the valves on the unit are closed.
2. Start the water spray.
3. Start the belt at the desired speed.
4. Pressurize the vessel to the desired pressure (not to exceed 60 PSI).
5. Check the pressure vessel to ensure that there are no leaks.
6. Start the pump to allow slurry to flow on the belt.
7. Open the valve to allow the dewatering to begin.
8. Open the water discharge valve to ensure that water level in the waste tank is brought to
zero.
31 |
Virginia Tech | CHAPTER 3:
HYPERBARIC BELT FILTER OPTIMIZATIO(cid:6)
3.1 Introduction
The experimental testing of the hyperbaric belt filter was initiated after the design and
construction activities were completed. This work is the final phase of the process for developing
the hyperbaric belt filter. During this phase, the performance of the prototype unit was tested,
analyzed and calculated. The performance was measured on several different aspects. Some of
these measures focused on the functionality of the unit, while others focused on the dewatering
capabilities.
The test program included both batch laboratory tests as well and continuous tests
conducted with the new prototype. Because of the importance of particle size in determining
filter performance, the tests were conducted using coal feeds that were coarse, fine and
intermediate (mixed coarse and fine). Several sets of tests were also performed with dewatering
chemicals in order to further enhance the dewatering performance of the unit. Once these tests
were completed, the optimal operating conditions for the new hyperbaric belt filter were
established. This data was used to develop empirical correlations that could be used to predict the
performance of the filter under other conditions not tested in the current study.
33 |
Virginia Tech | 3.2 Experimental
Described in this section are the precise procedures that were necessary to complete the
test work for the filtration device. This is recorded to ensure that proper scientific procedures
were followed and to allow these experiments to be replicated elsewhere if desired.
3.2.1 Feed Material
To test the operation of the filter, feed coal material is needed. It was decided to perform
the tests under three primary conditions. The first is to have the filter tested under good filtration
conditions. For these tests, the size distribution of the particles in the filter was 1 mm x 0. This
material was obtained as screen-bowl feed for the Alpha Natural Resources Tom’s Creek Mine
in southwestern Virginia. From filtration theory (Gregory, 1984), it is known that larger particles
are easier to dewater than smaller ones. This higher moisture is primarily due to the increased
surface area associated with fine and ultrafine particles.
The second set of tests was also run using fine material that was obtained from Tom’s
Creek Mine. This material is column flotation overflow material; it is much finer and has low
percent solids (less than 10%). This material was nominally minus 0.15 mm (minus 100 mesh).
The third series of tests was done with a mixture of the two feeds. The mixture is 50%
coarse and 50% fine material by volume. This results in a feed that is 1 mm x 0, but has a much
more even size distribution due to the greatly increased amount of fine particles compared to the
coarse tests.
Due to the dramatic difference that has been reported in filter performance due to particle
size, the unit was tested with both extremes of materials to see how the unit would perform under
34 |
Virginia Tech | difficult conditions (Dickenson, 1997). From these tests it should be possible to predict the
performance of the filter under a wide range of operating conditions.
3.2.2 Procedure
All of the procedures used in testing the unit were considered and used for the purposes
of functionality and repeatability. In the following sections, each step in the testing process is
described in detail to show how the testing was done.
3.2.2.1 Maintaining the Sample
The samples were obtained in 55 gallon drums and hauled on a trailer to the Plantation
Road processing facilities at Virginia Tech. When it was ready to be used, it was dumped from
the drum into a sump. While in the sump, the sample was mixed and aerated by a two
horsepower mixer/aerator. Also, it was recirculated from the bottom of the sump back to the top
by a pneumatic powered double diaphragm pump to minimize settling. After each had pumped
and mixed for size hours, it was not used again to avoid fines generate via attrition. The pumping
and mixing was done to help ensure that the sample was as uniform as possible during testing.
The sample is meant to be representative of what conditions exist in the processing plant.
3.2.2.2 Percent Solids
The measurement of percent solids was done by pumping the entire sample through a
sample splitter numerous times until it was reduced to approximately 100 ml of sample. Then,
the 500 ml beaker was weighed while dry. The 100 ml of sample slurry was placed into it and
weighed wet. It was then placed in an oven at 80oC overnight and weighed dry. From these
numbers, it is a simple calculation to determine the percent solids in the feed.
35 |
Virginia Tech | 3.2.2.3 Air Flow
The air flow input into the pressure filter was determined using a spring loaded flow
meter. This meter was located after the pressure regulator in the air input line. This allowed it to
measure all of the air that went into the unit so that the air requirements at each operating
condition would be known.
3.2.2.4 Moisture Content
The moisture content of the discharge material is of prime importance. Each of the tests
was conducted over a five minute period. After five minutes, all of the material that the filter
produced was collected by discharging it through the knife gate valve system. Since some of the
tests produce a large amount of filter cake it was necessary to take a sample of the discharge.
This was done by cutting the sample using a Jones Riffler to reduce to sample to a manageable
amount that could be weighed, dried, and reweighed. The sample was reduced in size to less than
50 grams. It was then weighed and placed in an oven at 80oC for five hours. This was to ensure
that all of the moisture had been removed. Once dried, it was weighed again to determine how
much of the original weight could be attributed to moisture.
3.2.2.5 Effluent Characterization
The effluent from the filtration device was also analyzed. This was to determine how
much material was being lost through the filter cloth. This material was collected at the bottom
of a stainless steel tank. Once the test was completed, the tank was drained into a container and
the contents weighed. This material was also dried to see the weight of the solid material that
was being lost during the separation process. From this process, it is not possible to determine
how much water is being released since much of it is carried with the air to the atmosphere.
36 |
Virginia Tech | 3.3 Experimental Results
The results that are described below were arrived at using the methods explained in the
experimental section of this chapter. The results are primarily divided into three sections coarse
feed, fine feed, and mixed feed. This segregation is due to the different behavior caused by the
different feed material.
3.3.1 Lab Batch Testing
The batch testing was done for two primary reasons. The first was to set a baseline for
what types of moisture percentages are possible if conditions in the batch tests could be
duplicated in a continuous filter. The second reason for the batch testing was to determine the
best possible reagents and dosages for the coal that was to be used for filtration. These optimal
reagents would then be tested on the prototype unit.
From the filtration theory described by Equation (2), it is known that one of the
controlling factors in dewatering is the driving force created by a ∆P across the filter cake. For
this reason, all of the tests that conducted were done at 60 PSI. Over fifty tests were done to
create each of the charts in this section. The data obtained from the batch tests are plotted in
Figures 3.2 and 3.3. The difference between these two charts is the amount of material that was
used for each. The later chart shows data from tests run with more slurry, thus increasing the
dewatering time and increasing the cake thickness. The tests were conducted using proprietary
dewatering reagents provided by Nalco Chemical. In this work, these proprietary reagents have
been designated as Reagent RU and RV.
From these charts, it is clear that for this coal sample, which was cut from the larger
sample being used to test the continuous unit, that the best reagent combination is RV at an
38 |
Virginia Tech | 3.3.2 Coarse Feed
On the continuous filter there were four primary controllable parameters that were
considered to affect the moisture content of the product. The first has already been mentioned
and that is the size of the feed particles. The second is the pressure across the filtrate. Third is the
speed that the belt moves and the fourth is the speed at which the material is added to the filter or
the mass flow rate. Using these four parameters, it is possible to affect the air flow rate, cake
thickness, production rate, and the moisture content of the product. Each of these responses will
be addressed separately as they each have their own unique impact on the overall performance of
the filter device. The analysis has also been broken down by size class.
This section concerns feed material which is screen-bowl feed from the Tom’s Creek
Mine. The size of the material is 1 mm x 0. This material has some spiral material mixed in with
it. About 15% of this material is minus 325 mesh. Thus, it contains a wide range of sizes of
particles.
3.3.2.1 Moisture Content
There are several factors that affect the moisture content of the product. First is the size
of the material as mentioned before, second the cake thickness, third the ∆P across the cake,
fourth is the dewatering time, and fifth is the water’s affinity to remain attached to the water.
With this filter it is possible to affect all of these primary factors but not all of them directly.
Since some of these factors cannot be controlled directly it makes sense to focus on
inputs that can be controlled. These are feed rate, belt speed, pressure, and chemical reagents
which affect surface tension.
The first thing that was examined was how consistent the filter’s product is when all of
the external factors are held constant. The results of these tests are shown in Figure 3.6. The
42 |
Virginia Tech | average value of these tests is 11.4%. These tests were conducted at an operating pressure of 60
PSIG. The percent solids for all of the coarse material tests were approximately 24% solids by
weight.
The standard deviation of this data is 0.0074. This is showing that while the range of the
data is 1.1% by moisture percentage the actual difference between the numbers is small. There
are some likely causes for this variation. One unlikely possibility is that the feed material is
changing. As described in the procedure section the feed it highly mixed and is also recirculated
in the sump. The feed pump is a progressive cavity pump that even with high solids contents and
large particles delivers a steady pulseless flow.
Some likely causes of this variation have to do with the moisture in the compressed air
itself. The compressor while compressing the air heats it to a temperature of approximately 60oC.
The air is not conditioned in any way - it is straight from the atmosphere in Blacksburg, Virginia.
This means that there is a significant amount of moisture in the air. This hot air rises to the top of
18.0%
16.0%
14.0%
12.0% 11.9% 11.6% 1 2.1% 11.4% 11.5% 12.3%
10.5%
10.0% 10.2%
8.0%
6.0%
4.0%
2.0%
0.0%
0 2 4 6 8 10
Test #
Figure 3.6. Variability of the coarse data (30% feed pump speed).
43
erutsioM
ekaC |
Virginia Tech | Design-Expert® Software
Moisture
Design points above predicted value
Design points below predicted value
0.653 0.66
0.088
0.5125
X1 = A: Pressure
X2 = B: Belt
0.365
Actual Factor
C: Feed = 25.00 0.2175
0.07
90.00
75.00
60.00
45.00
52.50
60.00
45.00
37.50
30.00 30.00
Figure 3.7. Response surface with coarse feed and no chemicals.
The chart shows the pressure on the x-axis and it is measured in PSI. The y-axis is the
belt speed and it is shown as a percentage of the maximum speed of the belt, and the z-axis
displays the moisture content at a given operating condition expressed as a decimal. Table 3.1
shows the conversion factors between the belt speed as a percentage and feet per minute.
Figure 3.7 shows how the filter device performs at every operating point in the given set.
For each input of pressure and belt speed, it shows how the filter will respond. In this case, it is
showing that at a low pressure and slow belt speed the moisture content is very high. This is
because of the slurry being too deep and not enough dewatering time is available due to the short
length of the filter belt. With thick slurry, the cake formation time is greatly increased (see cake
45
erutsioM
B: Belt
A: Pressure |
Virginia Tech | 3.3.2.2 Air Requirements
The air requirements for the filter device are important to understanding the cost of
operating a production unit. This is without a doubt the most expensive part of the filtration
process since it is here that the energy to do the dewatering is applied.
There were several factors that were investigated to see if they had a significant effect on
the air requirements of the filter device. These were cake thickness, pressure, feed size, and
chemical additions. The other factor that affects the air required is the amount of losses in the air.
These losses are the result of air leaking from the pressure vessel into to the discharge without
passing through the filtrate.
To approximate how much of the air was being lost, the filter belt was completely sealed
off with plastic and then pressurized at the operating pressures to determine how much of the air
was being lost. The results of this study are shown in Table 3.2. This loss is a result of an
imperfect seal that exists between the rubber belt and the pressure plate. All of the airflow
measurements that are shown below have been corrected based on this table.
Table 3-2. Air losses due to imperfect seal.
Pressure Consumption Losses Percentage
(PSI) (CFM) (CFM) Loss
30 88 4 4.5%
45 103 8 7.7%
60 109 12 11.0%
48 |
Virginia Tech | 120.00
110.00
100.00
90.00
80.00
70.00
60.00
0.000 0.200 0.400 0.600 0.800 1.000 1.200
Cake Thickness (cm)
Figure 3.9. Cake thickness vs. air flow rate for coarse feed.
Based on the speed of the belt with the corresponding pumping speed, it is possible to
construct an estimate of the cake thickness. After comparing the cake thickness vs. CFM
requirements graph, it has been determined that there is only a slight correlation between them.
This chart is shown on Figure 3.9.
The pressure vs. CFM chart is shown in Figure 3.10. This chart shows a strong
correlation between the two factors. There is clearly a direct relationship between these two. It
isn’t clear on the chart, there are many more points than it appears, but they are often
overlapping.
The last factor that was thought to control the air requirements was reagents. The plot of
pressure vs. CFM of both the base tests and the RV tests are shown on Figure 3-11. From this
chart it is clear to see that both are directly related and that the coal conditioned with the reagents
required more air than the corresponding base test conditions.
49
MFC |
Virginia Tech | 3.3.2.3 Mass Flow Rate
The mass flow rate is important to know since it will show what the production capacity
of a full size unit would be. This factor is clearly controlled by two factors. That is the percent
solids in the feed and the pump speed. This is due to the fact that the amount of material put on
the filter directly controls how much comes out as product. The amount of material lost through
the filter cloth is important to know. Figure 3.12 shows the mass flow rate into the filter on the
horizontal axis and the amount of losses on the vertical axis. It has been found that although
losses do increase with more material they increase at a rate slower than the input. This is due to
the formation of a bed at the bottom of the filter. This bed once formed greatly reduces the
amount of material lost. It is formed at both low and high feed rates.
Figure 3.13 is informative with regard to how the filter is truly performing. It shows the
0.0016
0.0014
0.0012
0.0010
0.0008
0.0006
0.0004
0.0002
0.0000
0.000 0.200 0.400 0.600 0.800 1.000 1.200 1.400
Inpu t lbs/min
Figure 3.12. Material input vs. losses for coarse feed.
51
nim/sbl
sessoL |
Virginia Tech | 99.2%
99.0%
98.8%
98.6%
98.4%
98.2%
98.0%
97.8%
97.6%
97.4%
97.2%
0.000 0.200 0.400 0.60 0 0.800 1.000 1.200 1.400
Mass Rate (lbs/min)
Figure 3.13. Mass rate vs. recovery for coarse feed.
mass cake rate vs. recovery of the filter as an average of all of the values at each mass rate for
clarity. The data shows a very slight increase in the recovery of the filter with more material on
the filter. If it were possible to put more material on the belt, it is expected that the recovery
would asymptotically approach 100%.
3.3.3 Fine Feed
All of the tests and analysis that was done for the coarse material was also duplicated for
the fine material. This material is column product (minus 0.15 mm) from Tom’s Creek mine. It
was tested to see how the belt filter would perform under poor conditions for dewatering. The
feed material contains 8% solids by weight. This material contains 61% minus 325 mesh. For
this reason, all of the tests were conducted with flocculant at a concentration of 1 gallon per 70
tons of coal. The flocculant addition was done by a small pump used to produce very small drops
at a slow speed.
52
)%(
yrevoceR |
Virginia Tech | Design-Expert® Software
Moisture
Design points above predicted value
Design points below predicted value
0.902
0.35
0.92
X1 = A: Pressure
X2 = B: Belt
0.795
Actual Factor
C: Feed = 25.00 0.67
0.545
0.42
90.00 60.00
75.00 52.50
60. 00 45.00
45.00 37.50
30.00 30.00
Figure 3.15. Response surface for fine feed and no chemicals.
The most noticeable difference is that the moisture content is much higher than for the fine
material. This is due to the fact that it takes more force over a longer period of time to dewater
this material. The belt filter was not long enough to give this material sufficient time to fully
dewater. From the response surface it is easy to see that the best operating point for moisture was
at 60 PSI of pressure and a belt speed of 60 percent.
Based on the results found in the batch testing of this material it was determined that the
best reagent to use is the RV (1:10) at 3 lbs/ton. This was then added to the sample and tested.
54
erutsioM
B: Belt A: Pressure |
Virginia Tech | The results were input into Design Expert which created the response surface shown in Figure
3.16.
The response surface shows how the filter behaved in this situation. This material has a
lower moisture at every operating point compared to the results without the dewatering
chemicals. Most of this gain was realized because of the greatly improved rate of dewatering that
was realized. The worst point occurred when the pressure was low and the belt ran slow. This
created a situation where the driving force was low and the cake thickness was high. This
situation simply did not produce cake formation rates that were fast enough. The best operating
point was at high pressure and high belt speed. This was because this had the thinnest cake which
realized the greatest moisture reduction. The other corners of the design also exhibit the same
Design-Expert® Software
Moisture
0.793
0.291
0.8
X1 = A: Pressure
X2 = B: Belt
0.6775
Actual Factor
C: Feed = 25.01 0.555
0.4325
0.31
90.00
75.00
60.00 60.00
52.50
45.0 0 45.00
37.50
30.00 30.00
Figure 3.16. Response surface of fine feed with 3 lb/ton reagent RV (1:10).
55
erutsioM
B: Belt
A: Pressure |
Virginia Tech | 160
140
120
100
80
60
40
20
0
0 10 20 30 40 50 60 70
Pressure (PS I)
Figure 3.18. Pressure vs. air flow rate for fine feed (base vs. reagent RV).
seen that pressure is a controlling factor for air requirements and that the use of the reagent also
increases the requirement.
The comparison between the air requirements for the two different feed types is shown on
Figures 3.19 and 3.20. On the chart with no reagent it is difficult to discern any meaningful
difference in air flow from the two types of feed. This is not true on the material with RV. On
this diagram, it shows that the coarse material has a higher airflow than the fine material. This is
most due to the greater size in the cracks in the and due to the higher porosity that is created in
the cake that contains large particles compared to the small ones.
3.3.3.3 Mass Flow Rate
The mass flow rates are essential for developing a scale up model for this pressure filter.
Figure 3.21 shows the mass flow rate vs. losses rate. From this plot, it is shown that the loss rate
57
MFC Base
RV (1:10) |
Virginia Tech | distribution. The size range is still 1 mm x 0 as with the coarse feed but it has more minus 325
mesh material. This material is 38% minus 325 mesh, compared to the coarse feed which was
16% and the fine feed which was 61%. Figure 3.23 show the consistency of the product obtained
using the mixed feed. For this group of tests, the standard deviation is 0.0048.
3.3.4.1 Moisture Content
Once the variability of the data was established, the same series of tests was once again
run for mixed sample. The results were then put into Design Expert and analyzed. From this
analysis, it was possible to generate a response surface. Figure 3.24 shows the response surface
for the mixed data without any chemical additions.
From the surface, the optimal operating point for this material is with a high belt speed
and high pressure. This would result in shorter drying times and thinner cake. Also, for this
particular feed, the thin cake helps prevent a layer of slimes from coating the surface and
18.0%
16.0%
12.9%
14.0% 12.3%
11.5% 11.4%
12.0%
12.1%
10.0% 11.8% 11.7% 12.0%
8.0%
6.0%
4.0%
2.0%
0.0%
0 1 2 3 4 5 6 7 8 9
Test #
Figure 3.23. Variability of the mixed feed data (30% feed pump speed).
60
)%(
tnetnoC
erutsioM |
Virginia Tech | Design-Expert® Software
Moisture
0.737
0.108
0.74
X1 = A: Pressure
X2 = B: Belt 0.58
Actual Factor
C: Feed = 25.01 0.42
0.26
0.1
90.00
75.00
60.00
60.00
52.50
45.00
45.00
37.50
30. 00 30.00
Figure 3.24. Response surface for mixed without chemicals.
blocking the air from penetrating the cake. The very high moisture that is shown at a slow belt
speed and low pressure is the result of the belt flooding due to a cake thickness that is too great
and not a long enough time for it to dewater.
Also done were tests on the unit with chemical additions that had been made. These were
also completed with RV (1:10) at 3lbs/ton. The resulting response surface is shown in Figure
3.25. This chart indicates that the moisture content can be lowered by a significant amount from
those done without the reagents. The moisture reduction is approximately 2-3 percentage points.
The best operating conditions for the material under these conditions is with a high pressure and
a moderate ranging belt speed. This is a normal optimal condition which is expected.
61
erutsioM
B: Belt
A: Pressure |
Virginia Tech | 3.4 Discussion
From these results it is possible to scale up this unit and to predict the behavior and
requirements of a larger full production capacity unit. This process has several assumptions that
must be made and can therefore only provide an estimate as to the performance of such a unit.
3.4.1 Discrepancies
In terms of scale up, most of the discrepancies exist in the fine feed batch testing
compared to that of the continuous filter. The lowest moisture numbers for the batch test are
approximately 14%, while the lowest moisture numbers for the continuous unit are
approximately 29%. The primary reason for this difference is the dewatering time. In the batch
unit, the total time can at times be almost 3.5 minutes. Across the belt the longest time that is has
is only 2.5 minutes. Also, the slower the belt moves the greater the cake thickness thus making
dewatering even more difficult. If the belt were longer, it would be possible to maintain a thin
cake and allow the proper amount of dewatering time. If this occurred, then it is possible that it
would produce similar moistures to the batch unit.
The coarse material did not respond in the same manner on the continuous unit as it did
on the batch unit. It was not successful at significantly reducing the moisture content of the
product. It did, however, increase the rate of dewatering thus allowing greater throughput on the
machine. This rate change can be seen from comparing Figure 3.7 and Figure 3.8, where the
great improvement was made when the pressure was 30 PSI. The value of this addition will be
further explored in the scale up section.
65 |
Virginia Tech | 3.4.2 Scale-Up Analysis
To do a meaningful scale up the results of two different size machines should so a
relationship. The most important factor in this design is the moisture content of the product. For
this reason, the scale up analysis will be done primarily based on the coarse feed material since
its findings can be confirmed more definitely with those of the batch unit. The computer models
also suggest that if the dimensions of the filter belt were altered to increase the dewatering time
that the fine material would more closely correspond to the lab results.
The scale up analysis will look into determining approximately how much energy would
be required to operate a unit with a capacity similar to that of other dewatering equipment.
Finally, there are some major design modifications that are suggested for a full time operating
unit.
3.4.2.1 Horsepower Requirements
To build a full size capacity unit, the optimal operating point will be used as the design
criteria to design for. This is the best that a unit built just like this one could achieve. There are
two processes that occur in this unit that require power. The first is turning the rubber belt. This
requirement is essential but its overall power requirements are dwarfed by the second
requirement. Second is the air compressor that will need to produce large amounts of compressed
air continuously.
The optimal operating point produces a cake with 10.7% moisture required 110 CFM of
air, while producing 1½ pounds of coal per minute. Assuming that the air consumption has a
linear relationship with capacity as it scales up, a unit capable of handling 50 tons per hour
would require approximately 120,000 CFM. To produce this much air would require
66 |
Virginia Tech | 3.5 Conclusions
This research set out with two objectives. The first was to design and build and operating
pilot scale hyperbaric belt filter. The second was to test such a device to determine its operating
parameters. The primary purpose of these tests was to determine how low of a moisture content
could be achieved and what would be the work input to deliver this separation.
The first objective was completed in May 2006 when the unit successfully demonstrated
a significant moisture reduction with a coal slurry feed. Once this was accomplished and shown
that it could be achieved repeatedly, the shake-down tested period came to a close and the unit
was declared operational. When this happened, the first goal of constructing an operating device
was completed.
The second objective of testing the unit to determine its operating parameters was also
completed. The results of these tests indicate several key points in the operation of the hyperbaric
filter. The first is that the unit consumes more air at higher pressure. Secondly, the moisture
content of the product depends on many factors including, belt speed which operates best in the
mid-ranges between 40% and 75%. The fine feed material takes significantly longer to dewater
than does the coarse material. Also, pressure across the filtrate has a large impact particularly in
the fine feed material. The chemical additions had the ability to reduce the moisture content of
the mixed material and fine material in a significant way. The reagents had the ability to increase
the rate of dewatering for the fine material. For the fine material, this increase in dewatering rate
resulted in greatly reduced moisture contents for the product material. The lowest moisture
content that was recorded for the fine feed material was 29.1% by weight, for the mixed material
8.4%, and for the coarse material it was 8.2%. Finally, it was determined that if a production
capacity unit operated with the same ratio of feed to air consumption it would use between
69 |
Virginia Tech | CHAPTER 4:
GE(cid:6)ERAL SUMMARY
In the United States there is a great need for energy. To help supply this need, coal is one
of the primary resources. The coal processing industry has a need for a solid-liquid separation
between coal particles and liquid water. This process is currently expensive and often sacrifices
many tons of fine coal particles. To address this need, this project seeks two goals. The first is to
design and construct an operating pilot scale version of a piece of equipment to improve
dewatering performance. The second goal is to test this new piece of equipment to determine
how effect it is at removing water from coal slurry.
The first stage of achieving the goal of building a unit was to decide what type of unit to
construct. It was decided after much literature research that a hyperbaric horizontal belt filter had
much promise. So the unit was designed and constructed. This process took approximately 18
months to complete. At the conclusion of this phase of the project the belt filter was capable of
consistently operating and causing significant moisture reduction.
The second stage began once the first stage was completed and it involved testing the unit
to determine how it would perform under a series of operating conditions. From these tests, it
was determined that the belt speed that produced the best results was in the mid-range of the
machine’s capability. Also, it was much more effective at dewatering the coarse particle material
than it was the fine particle material. The lowest moisture reduction that was achieved was in the
coarse material where the feed contained 76% moisture and the product 9%. This was even
further reduced by adding dewatering aids which brought the final moisture down to 8%. The
fine feed material realized a moisture reduction from 92% moisture to 35%. This was also
enhanced by the presence of dewatering aids to 29%. Then a mixed material was created and it
71 |
Virginia Tech | Blasting Design Using Fracture Toughness and
Image Analysis of the Bench Face and Muckpile
Kwangmin Kim
Abstract
Few studies of blasting exist because of difficulties in obtaining reliable fragmentation data
or even obtaining consistent blasting results. Many researchers have attempted to predict
blast fragmentation using the Kuz-Ram model, an empirical fragmentation model suggested
by Cunningham.
The purpose of this study is to develop an empirical model to relate specific explosives
energy (E ) to blasting fragmentation reduction ratio (RR) and rock fracture toughness
SE
(K ).
IC
The reduction ratio was obtained by analyzing the bench face block size distribution and
the muck fragment size distribution using image analysis. The fracture toughness was
determined using the Edge Notched Disk Wedge Splitting test.
Blasting data from twelve (12) blasts at four (4) different quarries were analyzed. Based on
this data set, an empirical relationship, E =11.7 RR 1.202 K 4.14 has been developed. Using
SE 80 IC
this relationship, based on the predicted blasting energy input for a desired eighty-percent
passing (P80) muckpile fragment size the burden and spacing may be determined. |
Virginia Tech | Chapter 1 Introduction
1.1 Statement of the Problem
The United States National Materials Advisory Board estimates that improving the energy
efficiency of comminution processes using a practical approach could result in energy
savings of over twenty (20) billion kilowatt-hours per year. Practical savings in the
comminution process, especially blasting, is one of the most important steps (Adel, 2004).
Few studies of blasting exist because of the difficulty in obtaining reliable fragmentation
data, and getting consistent results is difficult. Furthermore, although blasting engineers
differ with regards to desired results, all blasting designs rely on the experience of these
engineers. Excessive fines or oversized fragments are examples of what to avoid. Blasting
has historically been regarded as a stand-alone operation and is usually reported as a single
cost in most analyses.
Blasting engineers widely use the Kuz-Ram model to predict the rock size distribution
arising from blasting. Many researchers have attempted to predict blast fragmentation using
the Kuz-Ram model, which is based on empirical studies of fragmentation.
The model has two primary parameters: the characteristic size, derived from blasting
parameters using the model of Kuznetsov (Kuznetsov, 1973), and a uniformity index, based
on geometric parameters of the drilling and blast design. The size distribution curve is
determined by these two parameters. However, this original Kuz-Ram fragmentation model
is limited in its application and erroneously assumes a fifty percent (50%) passing size as
the average adjusted size in the Rosin-Rammler model (Spathis, 2004 and Chung &
Katsabanis, 2000). In addition, even though the fracturing of rock, as well as of other
materials, is usually due to tensile failure, no current blasting model considers tensile
1 |
Virginia Tech | strength. For example, the Kuz-Ram model considers the Uniaxial Compressive Strength
(UCS) and the Young’s Modulus (E).
Figure 1.1 Blasting in Pittsboro
In-situ block size is an important aspect of any blasting model and design because many
pieces of blasted rock are released from the in-situ block, as shown in Figure 1.1. Currently
in-situ block size is still obtained by manually estimating the bench face because there is no
instrument for measuring a whole bench face to obtain easily and cheaply an in-situ block
size.
1.2 Proposed Solution
The purpose of this study is to develop a new empirical model in order to obtain the proper
burden and spacing for target fragment size, P80 (the desired eighty percent [80%] passing
size) in the muckpile and to predict a size distribution curve from the muckpile after a
bench blasting.
A blasting fragmentation model of the same form as the model proposed by Donovan’s for
energy prediction in the comminution is considered, the proposed model is as follow.
E =a RRb K c
SE IC
2 |
Virginia Tech | Chapter 2 Literature Review
The proposed equation, E =a RRb K c, is used to develop a new empirical blasting
SE IC
fragmentation model. To obtain the reduction ratio (RR) data, the bench face and the
muckpile have been analyzed using an image analysis program. The fracture toughness,
K , represents the rock properties. In addition, blasting fragmentation prediction model,
IC
Kuz-Ram, is used for the application of this new empirical model. Thus, image analysis,
fracture toughness and blasting fragmentation prediction models will be introduced in this
chapter.
2.1 Image analysis
Image analysis on the bench face and muckpile was conducted to get the reduction ratio
(RR).
The Reduction Ratio (RR) in blasting
In a crusher, the concept of the reduction ratio, RR, is the feed size over the product size
ratio, and the reduction ratio in blasting is very similar. The feed size is represented by the
bench block size and the product size is represented by the muckpile fragment size. Figure
2.1 shows the reduction ratio in a crusher.
Feed
Crusher
Product
Figure 2.1 The Reduction Ratio in a crusher
4 |
Virginia Tech | For example, the new concept of RR in blasting is shown in the following equation.
50
F50
RR =
50 P50
50% passing size from the bench face image analysis
=
50% passing size from the muckpile image analysis
The feed size before blasting
=
The product size after blasting
Reduction factors for 20% and 80% passing can also be determined in the same way.
2.1.1 Image analysis programs
Recent fragmentation assessment techniques using image processing program allow rapid
and accurate blast fragmentation size distribution assessments.
There are a number of different image-processing programs, and the following describes
some of the commonly used programs.
The IPACS system
The IPACS has the software functions: grabbing scaling, image enhancing, grey level
image segmentation, shape analysis (merging and splitting) and processing parameters. The
host computer for this image system is an industrial PC, and this system is well suited for
industrial purposes. Processing speed and accuracy are good, and the system is conducted
automatically with a video input picture (Dahlhielm, 1996).
TUCIPS system
TUCIPS (Technical University Clausthal Image Processing System) has been developed to
measure blast fragmentation at Technical University Clausthal (Germany). This system
involves general algorithms of image processing and a specially created algorithm for
5 |
Virginia Tech | muckpile image analysis. There is just five percent (5%) deviation in the practical test with
this program, so this system is suitable for practical use (Havermann and Vogt, 1996).
FRAGSCAN
FRAGSCAN measures the size distribution of blasted rock from the dumper or the
conveyer belt due to camera and mathematic morphology techniques. The equipment is
composed of a camera, an Image acquisition card, a control data card, a computer type PC,
a light. Conversion from surface to volume distribution is possible by using a spherical
model and the operating system is fully automatic. This tool provides reliable, consistent
results for industrial usage because extensive experimentation has provided satisfying
results (Schleifer and Tessier, 1996).
WipFrag system
Using digital image analysis of rock photographs and videotape images with granulometry
system, grain size distribution may be obtained by WipFrag. Photographic images are
digitized by using WipFrag from slides, prints or negatives, using a desktop copy stand. In
order to overcome size limitations inherent with a single image, WipFrag has the function
for zoom-merge analysis. Therefore, combined analysis of images taken at different scales
of observation may be analyzed. In addition, Using Edge Detection Variables (EDV),
fragment boundaries are analyzed efficiently, and manual editing can improve edge
detection (Maerz, Palangio, and Franklin, 1996)
SPLIT System
SPLIT is operated from eight bit grayscale images of rock fragments, and was developed
from the University of Arizona to figure out size distribution of rock fragment. There are
two kinds of SPLIT programs; one is used on the conveyor belt and its automatic and
continuous program, and the other is a manual program which uses the saved images.
However, the same algorithm is used in both programs (Ozdemir, Kahriman, Karadogan,
and Tuncer, 2003).
6 |
Virginia Tech | 2.1.2 The errors associated with image processing systems
It is extremely hard to obtain accurate estimates of rock fragmentation after blasting.
Following are the main reasons for error in using image analysis programs (Liu and Tran,
1996).
(1) Image analysis can only process what can be seen with the eye.
Image analysis programs cannot take into account the internal rock, so the sampling
strategies should be carefully considered.
(2) Analyzed particle size can be over-divided or combined.
That means larger particles can be divided into smaller particles and smaller particles can
be grouped into larger particles. This is a common problem in all image-processing
programs. Therefore, manual editing is required.
(3) Fine particles can be underestimated especially, from a muckpile after blasting.
There is no good answer to avoid these problems. In order to reduce these errors, sampling
strategies should be carefully selected and flexibility of system configurations as well as the
change of materials is important.
When rock size uniformity is high and thickness of layer is low, the image-processing
program is useful and efficient. However, if the uniformity of rock size is low and thickness
of layer is significant, the user should be especially careful accepting the results of image
analysis (Cunningham, 1996). Figure 2.2 shows an example of an analyzed muckpile image
using the SPLIT system.
7 |
Virginia Tech | Figure 2.2 Image of the muckpile and delineated image using the Split software
2.2 Fracture toughness
Fracture toughness of rock is an important index property for comminution. Using this
index, crushing equipment may be properly sized to meet specific needs without over sizing
equipment and increasing capital equipment costs. To determine the index of fracture,
toughness samples are loaded so that stress is concentrated on the tip of a crack.
The stress intensity factor for Mode I, K is a measure of the stress field at a loaded crack
I
tip with mode I type (Tada, Paris, and Irwin, 2000). When this value reaches a point of
catastrophic growth, it is said to have reached K . The value of K means Mode I fracture
IC IC
toughness and refers to an index of dissipated energy that was required to propagate a crack
to a point of catastrophic growth (ISRM, 1998).
The value of K is affected by temperature, loading rate, and the thickness of the member,
IC
so the fracture toughness can be the property of the material.
The following laboratory tests may be conducted to determine the value of K .
IC
• Chevron Bend [(ISRM, 1998) and (Sun and Ouchterlony, 1986)]
• Short Rod [(ISRM, 1998) and (Sun and Ouchterlony, 1986)]
• Cracked Chevron Notched Brazilian Disc (Wang, 1998)
8 |
Virginia Tech | • Single Edge Notched Bend (Fenghui, 2000)
• Compact Tension (Sun and Ouchterlony, 1986)
• Semi Circular Bend (Chong and Kuruppu, 1984)
• Flattened Brazilian Disc Specimen (Wang and Wu, 2004)
The Chevron Notched Short Rod and Chevron Notched Round Bar in Bending was
suggested as a fracture toughness test by the International Society of Rock Mechanics
(ISRM).
These methods are all variants of the same technique. A sample is prepared to a certain
specification and then has a notch cut into the sample. The sample is then loaded in such a
way that a crack is propagated from the notch. From the load applied and the geometry of
the sample K can be calculated. The main disadvantage to these methods of finding K is
IC IC
that they have very intense sample preparation procedures and the loading apparatus is
complex. Another difficulty encountered with these methods is the means of measuring the
dilatation of the notch prior to crack growth. Therefore, another method for fracture
toughness testing of rocks is necessary and a relatively easier test, END test, has been
proposed by Donovan and Karfakis (Donovan and Karfakis, 2004).
The relatively easier test proposed by Donovan uses an edge notched disk (END) sample
loaded on a wedge of set geometry. The sample is loaded uniaxially until failure. The peak
load and friction coefficient are used into following equation to determine the fracture
toughness:
⎛ ⎛α⎞⎞
⎜ 1−µtan⎜ ⎟⎟
Κ =2⋅ D ⋅⎜ Ρ ν * ⎝2⎠⎟ ⋅⎜⎛ a + 1 ⎟⎞ .1 (2.1)
IC 2a ⎜⎜ 2tan⎜⎛α ⎟⎞ 1+µcot⎜⎛α ⎟⎞ ⎟⎟ ⎜ ⎝0.355715(D−a)3/2 0.966528(D−a)1/2⎟ ⎠ t
⎝ ⎝2⎠ ⎝2⎠⎠
9 |
Virginia Tech | Where:
K = The critical stress intensity factor (MPa*m1/2)
IC
P = The applied peak load (N)
v
α = The wedge angle (11o)
µ = Friction coefficient (tanφ)
a = The notch length (m)
D = Specimen diameter (m)
T = Thickness of disc (m)
A new fracture toughness test, the Edge Notched Disk Wedge Splitting test, was developed
and verified to permit rapid and easy assessment of the fracture toughness of a rock.
In addition, Donovan’s work has shown that fracture toughness is related more strongly to
the specific comminution energy than any other material property tested. As a result, a
method for predicting the specific comminution energy, E , required to reduce a rock
c
particle to a given size based on fracture toughness, K , was proposed (Donavan and
IC
Karfakis, 2004).
Single Particle Breakage Testing
Bearman and Donovan (Bearman, 1989 and Donovan, 2004) have shown that a strong
correlation exists between the fracture toughness of a material and the power consumption
of a laboratory crusher used to crush the material, indicating that fracture toughness may
have practical application in the evaluation of blast fragmentation.
Single particle breakage test is to achieve crushing energy and product size distribution data
regarding the rocks, and these data will be compared with the fracture toughness of those
rocks. HECT system, the Allis-Chalmers High Energy Crushing Test, is used for single
particle breakage tests. Using HECT system, crushing force and actuator displacement can
be obtained as well as the net energy for rock crushing. Additionally, HECT can simulate
10 |
Virginia Tech | all crusher operating conditions, a wide range of crusher sets, speeds, and throws (Allis-
Chalmers, 1985 and Donovan, 2003).
Using HECT system, Donovan applied this K value to the prediction of jaw crusher
IC
power consumption in 2003. From this application, there is a strong relationship between
rock fracture toughness, K and specific comminution energy, E , and this correlation was
IC C
used to achieve an empirical model for the jaw crusher power consumption prediction as
the change of reduction ratio. As result, the predicted E and actual E were in agreement.
C C
The following is one of the Donovan’s E model.
C
[ ]
E = −0.511+0.511RR K (1≤ RR<1.5)
c IC
E
=[ 0.215RR0.425]
K (RR ≥1.5)
[ kwh/t]
(2.2)
c IC
Where, E is the specific comminution energy given in terms of kilowatt-hours per metric
C
ton, and reduction ratio (RR) is the particle size divided by the closed side set.
Because the relationship between Ec and K is based on only several rocks and two
IC
reduction ratios in Donovan’s test, Equation 2.2 is limited. However, the results of
Donovan’s experiment indicate a strong and proportional relationship between fracture
toughness and specific comminution energy. In addition, fracture toughness was shown to
be related to specific comminution energy more strongly than any other material property
tested, including tensile strength (Donovan, 2003).
Donovan’s test shows the possibility that fracture toughness can be used for practical
applications to predict blast fragmentation. In addition, K value can replace the other rock
IC
properties; uniaxial compressive strength (UCS) and young’s modulus (E) which are used
in Kuz-Ram, the empirical comminution energy prediction model in blasting.
11 |
Virginia Tech | 2.3 Blasting fragmentation prediction
Assessment of the blast performance is critical for optimal blasting, and the size
distribution of the blasted material is essential to determine the degree of fragmentation.
However, fragmentation is influenced by both controllable and uncontrollable parameters:
rock properties, the geometry of rock, and blasting patterns, and the optimal size
distribution of the blasted material depends on the mining objective. Optimal blasting is a
very complex and difficult issue. Furthermore, there is no method or equation which can
predict the blast fragmentation exactly because of varying desired blasting fragmentation
and numerous controlling parameters involved in the process.
Many researchers have recently developed models and computerized simulations.
Following are some of the widely accepted models (Jimeno and Carcedo, 1995).
2.3.1 Kuz-Ram model
Kuz-Ram is the combination of Kuznetsov and Rosin-Rammler equation, and an empirical
fragmentation model. Since its introduction by Cunningham, the Kuz-Ram model has been
used by many mining engineers to predict rock fragmentation arising from blasting, and
many researchers have attempted to improve the Kuz-Ram fragmentation prediction model
(Cunningham, 1983 and 1987).
The model has two main factors;
The characteristic size (X ): It was derived by the Kuznetsov model (Kuznetsov, 1973).
C
The uniformity index (N): It is based on geometric parameters of the drilling and blast
design.
The size distribution of the muckpile rock after blasting is determined by these two main
factors. However, this original Kuz-Ram fragmentation model has the limitation of
application and a high margin of error (Spathis, 2004).
12 |
Virginia Tech | The mean size of the fragments formed by blasted rock
The distribution function, an analytical representation of the fragment size composition of
blasted rock, has been suggested by Rosin-Rammler model (Lilly, 1986), (Chung and
Katsabanis, 2000), and (Kuznetsov, 1973).
⎡ N⎤
⎛ x ⎞
Φ =1−R =1−exp⎢−⎜ ⎟ ⎥ (2.3)
(X) (x) ⎢ ⎜ ⎝x ⎟ ⎠ ⎥
⎣ 0 ⎦
Where:
Φ is the distribution function (the total relative volume of fraction not longer than x).
(X)
X is the Characteristic Size.
0
N is the Uniformity Index.
R is the fraction of material retained on screen.
(x)
Using the Rosin-Rammler equation, the formula for the mean fragment size was suggested
with given rock volume and needed explosives by Kuznetsov (Kuznetsov, 1973).
4/5
⎛V ⎞
<X>=A ⎜ 0 ⎟ Q1/6 (2.4)
⎝ Q ⎠
Where:
<X> is the mean fragment diameter (cm).
V is the volume of blasted rock per hole (m3).
0
Q is the weight of explosives of TNT equivalent explosives per hole (kg).
A is rock factor:
A=7 The medium hard rocks, f =8~10.
A=10 The hard but highly fissured rocks, f=10~14.
A=13 Very hard and weakly fissured rocks, f=12~16.
f is the Protodyakonov factor.
13 |
Virginia Tech | An equivalent quantity of any explosives, Q related to TNT is calculated by Equation 2.5
e
because TNT is not currently used in blasting as a standard explosive (Clark, 1987).
⎛ E ⎞
Q=Q ⎜ e ⎟ (2.5)
e ⎝1090⎠
Where:
E is the absolute weight strength of the explosives (cal/g).
e
The factor 1090 is the absolute weight strength of TNT.
Cunningham
Cunningham used the Rosin-Rammler model for blasting analysis (Equation 2.3). If the
characteristic size (X ) and the uniformity index (N) are known, then the size distribution
0
will be obtained from Equation 2.3, Cunningham suggested following formula for
determining uniformity exponent (Cunningham, 1987).
⎛ B⎞⎛ S/B⎞0.5 ⎛ W ⎞⎛ L −L ⎞0.1 L
N =⎜2.2−14 ⎟⎜1+ ⎟ ⎜1− ⎟⎜ B C +0.1⎟ ×(1.1 or 1.0) (2.6)
⎜ ⎟
⎝ D⎠⎝ 2 ⎠ ⎝ B ⎠⎝ L +L ⎠ H
B C
Where:
D is the hole diameter (mm)
B is the burden (m)
W is the standard deviation of drilling accuracy (m)
S is the spacing (m)
L in the bottom charge length (m)
B
L is the column charge length (m)
C
L is the total charge length (m)
H is the Bench height (m)
14 |
Virginia Tech | The factor “1.1 or 1.0” means that if a staggered drilling pattern is employed, then ‘N’ will
be increased by 10% (1.1).
Usually ‘N’ varies between 0.8 and 2.2. High values indicate uniform sizing, but low values
indicate non-uniform sizing, high proportion of fines and the oversize. Therefore, ‘N’
higher values and a staggered pattern is preferred for uniform sizing.
If the burden of hole diameter is decreased, drilling accuracy is increased, the charge length
of bench height is increased, and spacing of burden is increased, then the uniformity index
is increased (Cunningham, 1983). This relationship may be derived by Equation 2.6.
In addition, the characteristic size, X was suggested by adjusting Rosin-Rammler
C
(Equation 2.3).
⎛ N ⎞
⎛ X ⎞
⎜ ⎟
R= Exp −⎜ ⎟
⎜ ⎜ X ⎟ ⎟
⎝ ⎠
⎝ C ⎠
If X is the average size (X ), then the value of R is 0.5 (50% passing) as following;
⎛ N ⎞
⎛ X ⎞
⎜ ⎟
0.5= Exp −⎜ ⎟
⎜ ⎜ X ⎟ ⎟
⎝ ⎠
⎝ C ⎠
X
Thus, X = (2.7)
C (0.693)1/N
Cunningham suggested the Kuz-Ram model as demonstrate in Equation 2.7. The effect of
the uniformity index and the characteristic size on fragmentation distribution is such that
the characteristic size fixes the specific size in the size distribution curve, and the
uniformity index determines the shape of size distribution curves by having this
characteristic size.
15 |
Virginia Tech | Limitation of the Kuz-Ram
Cunningham assumed that the fifty percent (50%) passing size as the average size during
adjusting Rosin-Rammler model in Equation 2.7. The fifty percent (50%) passing size is
not the same as the average size of the fragments in the muckpile (Spathis, 2004).
Therefore, Equation 2.7 is modified by simply using the fifty percent (50%) passing size
(X ) instead of the average size (X ).
50
X
X = 50 (2.8)
C (0.693)1/N
In addition, the Kuz-Ram model has limits; the S/B ratio should not exceed two (2),
initiation and timing should be arranged to avoid misfires and cut-offs, the calculated
relative weight strength should be closed with yielded explosives energy, and the jointing
of the ground should be assessed carefully.
Kuz-Ram model is merely focused on the prediction of size distribution after blasting in the
muckpile. However, blasting engineers want to know the proper blasting pattern for optimal
blasting at any given blasting site and situation. Therefore, more practical usage of Kuz-
Ram model will be examined in this study using empirical specific explosives energy
prediction model.
2.3.2 Larsson’s model
In 1973, Larsson has proposed the equation for K50, 50% passing size. Namely, assessment
of blast fragmentation, 50% passing size, is predicted by using the model. The Equation 2.9
shows that model (Jimeno and Carcedo, 1995).
K = S'×e(0.58*lnB−0.145*ln(S/B)−1.18*ln(CE/c)−0.82) (2.9)
50
16 |
Virginia Tech | Chapter 3 Experiment
Experimental data were collected from four quarries, two in South Korea and two in the
United States. The “Bosung” and “Sanyang” quarry are located in Ul-san and Jin-hae,
South Korea respectively, and the “Pittsboro” and “Boxley” quarry are located in North
Carolina and Virginia, USA. From these four quarries, data from twelve blasts were
obtained. For each quarry blast, image sampling and analysis were conducted to obtain the
data for the reduction ratio from the bench face and the muckpile. Fracture toughness, K ,
IC
was obtained by using the END test on samples prepared from rock blocks.
3.1 Image Sampling and Analysis From a Quarry Blast
Figure 3.1 shows the importance of determining in-situ block size in the blasting. Many
blocks on the bench face are just released from in-situ block by the blasting energy.
Figure 3.1 The released rock from in-situ block of the bench in Warrenton,VA
Although an in-situ block size is an important factor in the blasting model, the in-situ block
size has typically been estimated by observation of rock mass and structural mapping
analysis both manually and partially.
19 |
Virginia Tech | Figure 3.2 The bench face image in Pittsboro
As seen in Figure 3.2, in practice it is impossible to measure the entire bench face to obtain
the in-situ block size manually because the bench face is both too high and too dangerous to
measure. Therefore, SPLIT was used to measure the whole bench face structure and to
obtain the in-situ block size in this research.
3.1.1 Image Sampling
Important issues in image sampling are: the location of the image, the image angle from the
surface of the muckpile, and the scale of the image. In order to obtain good images, which
are both capable of being analyzed and representative of the entire rock assemblage,
sampling strategies must be carefully considered.
The location of image taking is important, and there are two sampling methods, random and
systematic. Both methods are complex and must be considered the purpose of the
investigation. Another consideration is the angle of the surface being photographed. Ideally,
the surface should be perpendicular to the camera lens.
20 |
Virginia Tech | Consistent sampling from image to image is the main strategy in this research and one of
the most important factors in the sampling strategy. Analyzed data show large variations
from image to image, but as a whole, the data demonstrate consistency. In a muckpile after
blasting, merely remaining consistent in sampling may not be sufficient to show the real
size distribution. This strategy saves time in sampling and is convenient for the blasting
engineer to make a site-specific model for quarry blasting.
The Image Sampling on a Bench Face
To consider the whole bench face, the in-situ block size on a bench face will be obtained by
using SPLIT program.
A digital camera was used to get the image of the bench face, which will be used in SPLIT.
The maximum size image that can be processed using SPLIT is 1680*1400 pixels, so the
maximum size image needs to be considered during sampling images because image
editing may be required in SPLIT, and a larger image may not be opened in SPLIT without
such editing.
Figure 3.3 The bench face image from Bosung quarry
21 |
Virginia Tech | Image samples were obtained during charging explosives after drilling for blasting.
Approximately five to seven (5-7) pictures were taken at each blasting, and three to five (3-
5) appropriate pictures for analyzing in SPLIT were chosen. Figure 3.3 is an example of
image sample. An article of known dimensions, a scale material, must be in the picture in
order to provide scale. A white plate was used as a scale material on the bench face. The
same scale material must be used from image to image for analyzing all pictures in SPLIT
regarding each blasting. Also, the number of scale materials should be the same from image
to image for analysis. Typically, only one scale material was used for the bench face image
analysis in this research.
The Image Sampling From a Muckpile After Blasting
Fragmentation assessment can be achieved by analyzing scaled photographs taken of the
muckpile. The digital camera should be held such that the long axis of the photograph is
vertical. The image should be taken with the camera lens perpendicular to the muckpile
surface (JKMRC, 1996).
Figure 3.4 The muckpile image from Bosung quarry
To provide scale in the photograph, a tennis ball was used. If the slope of a pile needs to be
shown, then two scale materials can be used as shown in Figure 3.4. These materials should
22 |
Virginia Tech | neither be placed randomly on the muckpile nor in a horizontal line across the muckpile. In
addition, the same number of scale materials should be used from sample to sample in the
same blasting site to analyze together in SPLIT.
As previously mentioned, the maximum image size is 1680*1400 pixels. If the size is too
large to be analyzed in SPLIT, then image editing is required.
A more representative sample may be obtained by photographing the material being loaded
into a truck or as a truckload is being dumped because the outside surface of a muckpile
before digging cannot represent the material within the pile. That means obtaining images
of the entire exposed surface of the pile to avoid biased results, however it takes longer to
acquire these images.
The main focus of sampling of muckpile images in this research is consistency rather than
representation. Therefore, the image samples after blasting were obtained directly from the
muckpile and five to seven images of the muckpile were captured and analyzed at each
blasting. To show as much of the muckpile as possible, each image sample was obtained at
the different part of a muckpile and overlapping images were avoided.
Safety is tantamount during sampling images from a muckpile. Scale material was thrown
to the dangerous site and the zoom function of a digital camera was used for sampling
purposes. It is especially dangerous near the bench face after blasting.
3.1.2 Image Analysis Using SPLIT
SPLIT, image analysis program, will be used for analyzing a bench face and a muckpile.
The block structure of bench in a quarry before blasting may be obtained by analyzing the
bench face, and the result of blasting will be estimated by using SPLIT on the muckpile.
23 |
Virginia Tech | Imgae analysis on a bench face
Consistency is the primary focus in the initial trial of the image analysis program used on
the bench face to obtain information regarding the in-situ block size.
Following is the important and specific setting at the step of Editing and Compute Size in
SPLIT for analyzing the bench face.
1. Only the scale material should be edited.
2. Zero percent (0%) of percent fines adjustment.
3. Rosin-Rammler was used for fines distribution.
Bench face images were not edited much in this study. Only the scale material and the other
parts of image except the bench face were edited. Therefore, the part of the bench face
image was not edited except for the actual scale material. If the bench face image is edited,
then the editing image is not sufficient because the crack on the bench face is often very
difficult to see. Although non-edited images analysis cannot reflect the real size distribution
of the block on the bench faces and contains errors, non-edited images revealed more
consistent analysis result than edited ones.
On the bench face, the fines are not considered because the bench face image analysis is for
obtaining information regarding the in-situ block size. Therefore the setting for fines
adjustment is zero percent (0%).
For the prediction of fines in SPLIT, there are three options - Schuhmann, Rosin-Rammler,
and best-fit. Any prediction model may be used, but the same model should be used from
image to image. The Rosin-Rammler model was chosen for this research.
At the step of Graphs and Outputs, this setting depends on the user. The setting of graphing
is “Cumulative”, and size axis and percent axis is “Linear” in this research. Figure 3.5
shows the bench face image analysis in SPLIT.
24 |
Virginia Tech | Usually one or two scale materials were used on the muckpile. If the angle of muckpile
surface needs to be considered then two scale materials were used. Of course, the number
of scale materials depends on the sampling situation.
Muckpile image analysis has the following limitations regarding fines estimation, so the
percent fines adjustment was set to “Medium.” The percent fines adjustment percentage
may be changed to twenty (20%), forty (40%), or sixty percent (60%), but the percentage
should be consistent from image to image. In this research, fifty percent (50%) was chosen
as the percent fines adjustment because this is the usual setting for muckpile image analysis
in SPLIT. Figure 3.6 shows the muckpile image and analyzed image in SPLIT.
Figure 3.6 The muckpile image and analyzed image in SPLIT
In addition, Rosin-Rammler model was used for prediction of fines. Users may choose the
prediction model. The situation is the same with the bench face image analysis.
Consistency of the model choice should be kept from image to image.
The setting of graphing is “Cumulative”, and size axis and percent axis is “Linear” at the
step of Graphs and Outputs in this research. This setting also depends on the user(s).
26 |
Virginia Tech | Table 3.1 The data from the size distribution curve of the bench face and muckpile
F20 F50 F80 P20 P50 P80
Bosung1 165 410 810 46 255 493
Bosung2 245 586 937 23 214 428
Bosung3 196 420 821 91 226 431
Bosung4 158 285 467 72 209 335
Bosung5 178 330 551 93 252 499
Pittsboro 158 281 483 40 100 187
Boxley 514 929 1512 69 161 304
Sanyang1 102 192 334 55 179 350
Sanyang2 102 196 382 88 192 290
Sanyang3 114 211 360 24 89 199
Sanyang4 50 91 149 21 82 163
Sanyang5 81 164 309 31 144 265
F20, F50, and F80 in Table 3.1 are from the size distribution curve of a bench face. F20
means the twenty percent (20%) passing size in the size distribution curve of a bench face,
and F50 and F80 are the same as F20.
P20, P50, and P80 are similar with F20, F50, and F80. These are twenty (20%), fifty (50%),
and eighty percent (80%) passing size in the size distribution curve of the muckpile.
F50 and the Mean In-situ Block Size
The reason for the bench face analysis using SPLIT is to obtain information about the block
size. In Pitttsboro blasting, the mean in-situ block size was measured manually, and the size
is 0.2 meters. In Table 3.2, the scale of F50 (or F20) and the mean in-situ block size is
similar. Although more research will be needed, it shows the possibility that F50 may be
used as a new index of mean in-situ block size. In addition, reasonable results of data
analysis were shown with these data in Chapter 4.
Table 3.2 The data from the bench face, and the manually measured block size
in Pittsboro
Location F20 F50 F80 Mean in-situ block
Pittsboro 158 281 483 200
28 |
Virginia Tech | 3.2 Calculation of K
IC
Standardized experimental testing has not been developed to determine the fracture
toughness of rock. The critical value of stress intensity factor, K , may be determined
IC
experimentally in different ways. Most methods involve intense sample preparation and
are then loaded under very specific conditions.
Donovan proposed a relatively easier test (Donovan and Karfakis, 2004). This test uses an
edge notched disk (END) sample loaded on a wedge of set geometry. The sample is then
loaded uniaxially until failure.
The wedging device used in END test consists of hardened steel. The wedge angle (α) is
11o. The experimental set-up is shown in Figure 3.8.
Figure 3.8 Test set-up for END wedge test
The experimental set-up is shown in Figure 3.8, and using MTS 810, the axial force is
applied with 8996N load cell. The obtained data (load, load-line displacement) is saved in a
PC. Loading rate of 0.003mm/sec for the load line displacement was used for the test
(Donovan, 2003). The peak load, P is recorded for K calculation using Equation 3.1.
V IC
29 |
Virginia Tech | ⎛ ⎛α⎞⎞
⎜ 1−µtan⎜ ⎟⎟
Κ =2⋅ D ⋅⎜ Ρ ν * ⎝2⎠⎟ ⋅⎜⎛ a + 1 ⎟⎞ .1 (3.1)
IC 2a ⎜⎜ 2tan⎜⎛α ⎟⎞ 1+µcot⎜⎛α ⎟⎞ ⎟⎟ ⎜ ⎝0.355715(D−a)3/2 0.966528(D−a)1/2⎟ ⎠ t
⎝ ⎝2⎠ ⎝2⎠⎠
Where: P = the applied peak load
v
α = the wedge angle
µ = friction coefficient
a = the notch length
D = specimen diameter
The peak load and friction coefficient are used in Equation 3.1 to determine the fracture
toughness.
The sample properties and geometric values used in this study were taken from experiments.
The angle of wedge is eleven degrees (11o). A tilt test was used to determine friction
coefficient (µ) on hardened steel. The sliding angle is φ and tanφ is equal toµ. The
friction coefficients for the four quarry rocks are given in Table 3.3.
Table 3.3 Values of φandµof four quarries rock
Friction Coefficient (µ)
Rock Type
φ(degree) µ
Pittsboro 24.0±1.11 0.445
Boxley 24.8±3.03 0.462
Bosung 30.1±2.14 0.580
Sanyang 27.2±1.67 0.514
Table 3.4 shows mode I fracture toughness, tensile strength, and specific gravity values for
the rocks in the four quarries.
30 |
Virginia Tech | Table 3.4 Fracture toughness, K , Tensile strength, σ, and Specific Gravity
IC t
Location K IC(MPa*m^0.5) σ t(MPa) Specific Gravity (t/m3)
Pittsboro 1.539±0.100 16.94±1.92 2.69±0.06
Boxley 1.724±0.193 17.45±2.57 2.73±0.01
Bosung 1.889±0.398 22.48±4.70 2.58±0.05
Sanyang 1.230±0.216 14.72±1.23 2.49±0.07
3.3 Blasting specific explosives energy
Specific energy for fragmentation is the explosive or mechanical energy required to
fragment a unit of volume or mass of rock (Rustan, 1998). The Specific Explosives Energy
(E ) in this research represents the blasting energy required to fragment a unit of mass of
SE
rock. Therefore the unit is “wh/tonne”. This value is affected by a blasting pattern
(explosives amount, bench height, burden, spacing, hole diameter, rock specific gravity,
and the type of explosives), consequently E can be assumed to represent the blasting
SE
pattern and can be determined using Equation 3.2.
Explosives Energy per hole(wh)
E = (3.2)
SE Height×Burden×Spacing×S.G.(ton)
Explosives energy per hole is affected by the diameter of the hole, bench height, and type
of explosives. The explosives amount per hole was estimated as the average explosives
amount per hole in this research.
Powder factor is the quantity of explosives used per unit of rock blasted (Kg/ton or Kg/m3).
An accurate prediction of powder factor in blasting is needed for optimal blasting to reduce
operation costs (drilling, blasting, loading, haulage and crushing). Powder factor is one of
the most important tools used to design the blasts (Jimeno and Carcedo, 1995). Since E is
SE
conceptually same with powder factor, E prediction will be evaluated for optimal blasting
SE
31 |
Virginia Tech | pattern by using an empirical equation model in this study. Blasting pattern for Pittsboro
and Specific Explosives Energy (E ) are tabulated in Table 3.5.
SE
Table 3.5 Blasting pattern and E in Pittsboro blasting
SE
Blasting pattern and E Pittsboro Unit
SE
Bench Height 19.8 m
Burden 4.57 m
Spacing 4.57 m
Rock Specific Gravity 2.69 t/m^3
Hole diameter 165 mm
Explosives Hydromite4400
Explosives amount per hole 429 kg
Explosives energy per gram 863 Kcal/Kg
Explosives energy per hole 431 Kwh
Specific Explosives Energy (Ese) 387 wh/tonne
3.4 The Main Issues in SPLIT for the Research (Norton, B., 2005)
There are various issues in the use of SPLIT. The concepts, P50, fines, the fines percent
adjustment in a muckpile, and the size of images, were considered as described below.
An Average Size and P50 in SPLIT
The P50 is not necessarily the average size, but is fifty-percent (50%) passing size by
weight. That means the P50 is not the mean size in terms of dimension. It is the mean size
by weight. Half of the volume or weight is less than this size particle. It is assumed the
particles have all the same density, so the terms of volume and weight can be used
interchangeably.
Larger or Smaller Image
When the muckpile was evaluated, the smaller image would definitely be more efficient
because it is smaller and closer to the size of the material. A larger image would be less
32 |
Virginia Tech | efficient as it takes longer to edit. However, a larger image provides more particles for the
sample, so there is a tradeoff. When the bench face was evaluated, the small and large
images showed similar results of image analysis. Once again, the main focus is keeping
scale consistent from image to image.
Fines
Smaller particles are hidden under the larger particles and are not visible. The resolution of
the image is such that the software can only measure down to a certain point. Therefore the
empirical model, either Rosin-Rammler or Schumann, below the point of the size
distribution curve, predicts the size distribution. The curve color is a little changed at that
point from that at which the size distribution is predicted by the empirical model in SPLIT.
The Selection of the Fines Percent Adjustment in a Muckpile
The key is not to change the setting from sample to sample. The default setting is fifty
percent (50%) and this is what many people use ninety-nine percent (99%) of the time
when analyzing muckpile images. This is because we do not have sieve results and we want
to be able to compare curves knowing the same setting was used to generate them. The
evaluation of images from a blast muckpile is particularly difficult due to its size, depth and
internal variations.
To obtain information of the in-situ block size on the bench face, the default setting is zero
percent (0%) because fines analysis on the bench face are not needed for this study, and the
key is keeping the setting consistent from image to image.
33 |
Virginia Tech | Chapter 4 – Data Analysis
Donovan has developed a model to predict the specific comminution energy, E , in jaw
C
crushers using the HECT system. Following is one of the Donovan’s models (Donovan and
Karfakis, 2004).
[ ]
E = −0.511+0.511RR K (1≤ RR<1.5)
c IC
E
=[ 0.215RR0.425]
K (RR ≥1.5)
[ kwh/t]
c IC
The unit of E is kilowatt-hours per metric ton and reduction ratio is defined as the particle
c
size divided by the closed side set of a jaw crusher.
Conceptually E and Donovan’s E are similar with each other. Thus, the prediction
SE C
model,E =a RRb K c, is assumed, and data analysis is performed using this proposed
SE IC
equation.
For the assessment of rock blasting, four factors should be considered (Cunningham, 1987)
and (JKMRC, 1996).
(1) Rock density
(2) Mechanical strength (UCS)
(3) Elastic properties (Young’s modulus)
(4) Structure (In-situ block size)
Mechanical strength is related with the Uniaxial Compressive Strength (UCS), and UCS
has been used as mechanical strength in usual blasting model. However, the Uniaxial
Tensile Strength (UTS), measured by the Brazilian test, has a better correlation with rock
fracturing. Especially, fracture toughness has a strong relationship with the tensile strength
of rock, and further, a good correlation with energy consumption for rock fragmentation in
a crusher (Donovan, 2003). Thus, the rock properties, mechanical strength and elastic
34 |
Virginia Tech | modulus, will be replaced by the K , fracture toughness, in the empirical blasting model in
IC
this study.
Assumed energy prediction equation form contains following meanings:
- E is affected by a blasting pattern (Bench height, burden, spacing, hole diameter, and
SE
explosives amount) and rock density, so E will represent a blasting pattern in the model.
SE
- RR reflects both whole benches face structure (In-situ block size, F) and the desired
fragment size (P) after blasting.
- K represents rock properties for prediction of explosives energy consumption.
IC
Therefore, assumed equation form contains the necessary factors for assessment of rock
fragmentation and includes target fragment size, P80.
4.1 The Equation Model
Table 4.1 summarizes the blasting and rock property data from twelve blasts in the four
quarries. The bench face and the muckpile information are represented by “F” and “P”
respectively. The data in Table 4.1 is analyzed using the proposed equation,
E =a RRb K c.
SE IC
Table 4.1 The obtained data of blasting
E K F20 F50 F80 P20 P50 P80 RR RR RR
SE IC 20 50 80
Bosung1 103 1.539 165 410 810 46 255 493 3.6 1.6 1.6
Bosung2 103 1.539 245 586 937 23 214 428 10.8 2.7 2.2
Bosung3 107 1.539 196 420 821 91 226 431 2.2 1.9 1.9
Bosung4 107 1.539 158 285 467 72 209 335 2.2 1.4 1.4
Bosung5 110 1.539 178 330 551 93 252 499 1.9 1.3 1.1
Pittsboro 387 1.724 158 281 483 40 100 187 4.0 2.8 2.6
Boxley 268 1.230 514 929 1512 69 161 304 7.4 5.8 5.0
Sanyang1 268 1.889 102 192 334 55 179 350 1.8 1.1 1.0
Sanyang2 292 1.889 102 196 382 88 192 290 1.2 1.0 1.3
Sanyang3 336 1.889 114 211 360 24 89 199 4.8 2.4 1.8
Sanyang4 222 1.889 50 91 149 21 82 163 2.4 1.1 0.9
Sanyang5 210 1.889 81 164 309 31 144 265 2.6 1.1 1.2
35 |
Virginia Tech | 4.1.1 Specific Explosives Energy, K and RR
IC 80
Following data in the table shows that E is correlated with RR and K value.
SE 80 IC
Table 4.2 E with given K and RR
SE IC 80
Location E (Wh/tonne) K (Mpa*m1/2) RR
SE IC 80
B o s u n g 1 1 0 3 1.539 1.6
Bosung2 103 1.539 2.2
Bosung3 107 1.539 1.9
Bosung4 107 1.539 1.4
Bosung5 110 1.539 1.1
Pittsboro 387 1.724 2.6
Boxley 268 1.230 5.0
Sanyang2 292 1.889 1.3
Sanyang3 336 1.889 1.8
Sanyang5 210 1.889 1.2
In Bosung and Sanyang quarries, each blasting location was close to the other, so the rock
property, K value was assumed to be the same within the same quarry blasting in this
IC
study. Therefore, although END test was just conducted for K value from Bosung1 and
IC
Sanyang1 blasting, the other blasted rock in Sanyang and Bosung were assumed as the
same with Bosung1 and Sanyang1 at each.
There is the reduction ratio for 80% passing for blast 1 and 4 at the Sanyang quarry, but
consequently the data was omitted from the analysis. The reason is that although explosives
energy was given, breakage was not realized and the entire blasting energy was consumed
to release the block on the bench.
In Table 4.2, regression analysis in EXCEL and SAS was fit to the proposed model, and
both programs gave the same result.
E = 11.7 RR 1.202 K 4.14 (4.1)
SE 80 IC
Where:
E is specific explosives energy (wh/ton).
SE
36 |
Virginia Tech | The lower part of the muckpile should be investigated to determine fines or small sizes of
rock distribution in the muckpile. Thus, a sieving test is needed to analyze RR or P20 in
20
the muckpile, or image sampling may be done from a hauling truck, although it will take
considerable time.
- Blasting data are too short to investigate the real relationship and to determine some
models.
The sampling method and analysis tool in these blastings is not good enough to obtain fines
in this research.
Although predicted E is not in agreement with given RR and K , the result is
SE 20 IC
reasonable because there is the limitation of SPLIT usage and image sampling method in
this research. The disagreement of relationships among RR , K , and E is rather
20 IC SE
reasonable.
4.1.3 P50 and P80
The two factors, P50 and P80 are in proportion as the change of uniformity index, N from
Rosin-Rammler model.
1/N
⎛ln0.5⎞
P50 =⎜ ⎟ P80 (4.4)
⎝ln0.8⎠
Where:
P50: 50 % passing size from the size distribution curve in a muckpile after blasting
P80: 80% passing size from the size distribution curve in a muckpile after blasting
N: The Uniformity index from Kuznetsov (Kuznetsov, 1973)
The Equation 4.4 can be derived from the following Rosin-Rammler equation.
x
( )n
R = e− x
C
42 |
Virginia Tech | L in the bottom charge length in m,
B
L is the column charge length in m,
C
L is the total charge length in m,
H is the Bench height in m.
If a staggered blasting pattern was employed, then the uniformity index will be increased
by ten percent (10%).
In the quarry blasting with similar bench height and hole diameter, the change of total
charge length, L, and the standard deviation of drilling accuracy, W, is limited.
Therefore, the range of (log0.5/log0.8)1/N is just (+/-) 0.1 as (+/-) 2 meters change burden
and spacing with the same bench height and hole diameter in the blasting pattern from
Equation 4.4 and 4.5.
Derived Equation 4.4 shows that the relationship between P50 and P80 will be in
proportion in quarries blasting which have the similar bench height and hole diameter.
Bosung (1,2,3,4) and Sanyang blasting (4,5) have similar bench height, 11m, and hole
diameter, 76mm. Table 4.4 is the blasting data of P50 and P80 in Bosung and Sanyang
quarry.
Table 4.4 P50 and P80 in Bosung and Sanyang
Location P50(mm) P80(mm)
Bosung 1 255 493
Bosung 2 214 428
Bosung 3 226 431
Bosung 4 209 335
Sanyang 4 82 163
Sanyang 5 144 265
The data of “P50 and P80” from Table 4.4 was analyzed using EXCEL regression, and the
following relationship was obtained.
P50=0.5P80 (4.6)
44 |
Virginia Tech | Chapter 5 Discussion of Results
5.1 Improvements Revealed by the Research
1. SPLIT was used successfully to examine the bench face.
Using SPLIT on the bench face was effective and produced reasonably consistent results.
F50 in SPLIT on the bench face can be a new index as the mean in-situ block size. Actually,
the manually measured mean in-situ block size and F50 in SPLIT on the bench face share
the similar scale in the Pittsboro quarry.
2. The Kuz-Ram model was adapted as a practical means for predicting size distribution in
the muckpile.
In the Kuz-Ram model, the average size of rock in the muckpile and the size distribution
curve were predicted using given factors in the blasting site. However, in this research,
because the products (P80) were what we wanted, the desired P80 was taken for granted,
and P50 predicted by the empirical model; proper burden and spacing were predicted by
using the empirical model with the given blasting factors. In addition, the size distribution
curve in the muckpile was predicted with P50 and proper burden & spacing based on the
given factors. P80, the eighty percent (80%) passing size, was assumed to be the optimum
size desired by most model users and the representative size of the muckpile.
3. An attempt was made to use Mode I fracture toughness as the rock property in the
blasting model instead of UCS (Uniaxial Compressive Strength) and E (Young’s modulus)
in Kuz-Ram.
Although tensile strength has a strong correlation with rock breakage, only UCS and E are
used in Kuz-Ram. In our study Mode I fracture toughness was used in the blasting model
because tensile strength has strong relationship with K .
IC
46 |
Virginia Tech | Mode I is the most commonly encountered mode of crack deformation in rock blasting
applications. However, although the relationship between K and the size distribution of
IC
blasting result could be determined, the data was insufficient to figure out the correlation.
Additional sieving tests seemed reasonable and justifiable. In addition we discovered that
K may especially correlate with fines and the small size of rock in the muckpile's whole
IC
size distribution. However, further research is needed regarding these issues because the
data is insufficient to make a final determination and an image analysis program has the
disadvantage of underestimation of fines in the muckpile.
4. Although this model is not specific to any blasting area because of the shortage of
blasting data, it may be adapted by blasting engineers to particular mines for the purpose of
making a site-specific model.
The empirical model was derived step by step as follows:
• Collect blasting image data (the bench face and the muckpile image)
and rock sample after blasting.
• Conduct the END test with sampled rock.
• Analyze collected data using image analysis.
• Conduct regression using SAS or EXCEL with suggested equation form.
• Adjust the empirical equations for specific blasting sites.
• Use this adjusted empirical equation model in the simulation model based on
Visual Basic. NET program.
Thus, it would seem that an organized blasting design with the desired consistent result,
P80, may be possible.
47 |
Virginia Tech | 5.2 The Bench Face Structure and the Blasting Design for Desired Consistent Results
Optimized blasting in mining is a complex issue because the entire detailed process must be
considered. Optimized blasting is different at each blasting site because there are different
crushers, drilling machines, hauling machines, and bench height considerations. However,
whether or not blasting is optimized, organized blasting for consistent results based on a
simple and flexible blasting model will greatly reduce energy waste.
If there is a given bench height and diameter of the drilling hole at the blasting site, then
burden and spacing are two of the most important factors in the blasting model because
these factors are easy to manipulate, whereas other factors are more difficult to control.
Therefore, application of the model has been focused on obtaining proper burden and
spacing to obtain the desired blasting result by using the site-specific blasting equation
model consistently.
To obtain consistent blasting results, one of the greatest obstacles in a given blasting site is
accurate determination of changes in the bench face structure. Although the rock properties,
bench height, and hole diameter are the same from blasting to blasting in a given blasting
site, the bench face structure is almost always changed during each blasting.
In this research, the change of bench face structure was measured easily and quickly using
the image-processing program. As with the change of the bench face structure (size ‘F’),
the burden and spacing in a blasting pattern were likewise adjusted for each blasting site.
5.3 P80 for the Optimal Blasting in a Quarry (Kojovic, 2006)
Practically, for a typical quarry, waste is below 0.005m, the desired final product size
ranging from 0.005 to 0.025m. That means everything above 0.025m has to be reduced.
This is typically done via blasting followed by multi-stage crushing. The amount of
48 |
Virginia Tech | blasting will clearly impact the amount of crushing required, so the trick is to find the
optimum level of blasting to maximize the yield of 0.005~0.025m product, at the lowest
overall cost (of drilling, blasting, and crushing, including the cost of energy, liners and
maintenance). Too much blasting can lead to too many fines, which may well reduce the
amount of crushing required at the expense of lowering the overall yield. Too little initial
crushing might require more crushing later in the process, which means more electrical
energy and more wear and tear on the equipment, with the result that costs go up once again.
Ideally, each quarry will maintain an optimum balance between the blasting and crushing
required to achieve the best possible outcome. Because rock conditions will dictate what
these balances should be, it is not enough to target just one size (P80). However, in
practice, targeting a P80 of 0.2m in the blast might be great in terms of the reduced
crushing requirements, but the blast would have to be too energetic, resulting in too much
waste, that is to say, anything less than 0.005m. We are therefore looking for the best
balance to achieve the optimum yield and lowest overall cost. To consider blasting as an
isolated process is to ignore the possible negative impact downstream.
P80 may be just a convenient yardstick on the level of coarseness in the blasting model, but
the predicted size distribution curve was obtained by using P80, resulting in proper burden
and spacing in the research model. Thus, a closed design for optimization in a quarry
process may be obtained by manipulating the blasting design (burden, spacing, and desired
P80) in the model. In the mine to mill optimization project (Virginia Tech and JKMRC
2004~2006), the optimized blasting came up with a P80 range of 0.2~0.3m. This range may
be used as the P80 size for optimized blasting in a usual quarry.
5.4 Generalized Blasting Model
A generalized blasting model is difficult to devise because rock properties and bench face
structure vary greatly, and obtaining blasting data is both expensive and time-consuming.
Actually, a generalized model may be unattainable because every blasting model must be
49 |
Virginia Tech | adjusted for the given blasting site. With this in mind, in constructing the present model just
twenty or thirty pieces of blasting data were used per given blasting site. This model may
be more powerful and efficient than any other blasting model within a given blasting site.
Using the method described in this thesis, will assist blasting engineers in achieving more
efficient blasting in their quarries and provide a much-needed reduction in the amounts of
energy used in the mining process.
5.5 Application for the Simulation Model
Using the obtained empirical equation models and adjusted Kuz-Ram model, the simulation
program can be made with Visual Basic.NET program.
Although the following simulation program is not truly sufficient for definitive conclusion,
this practical trial will be useful for making a site-specific blasting model simulation for
blasting engineers.
1. Given Factors in the model
*Blasting pattern
- H,D : Bench height(m)/ Hole diameter(mm)
- Ex. : Explosives amount (Kg, per hole)
- L : Charge length (m)
- W : Drill accuracy Standard Deviation(m)
*Rock properties
- SG : Rock Specific Gravity (tonne/m3)
- K : Fracture toughness (Mpa*m0.5)
IC
*The block structure on the bench face
- F80 : The in-situ block size (m)
*The goal of the blasting
- P80 : Wanted P80 in the muckpile size distribution (m)
2. The empirical model from the data analysis using regression
- E = 11.7 RR 1.202 K 4.14
SE 80 IC
- P50 = 0.5 P80 P50
50 |
Virginia Tech | 3. The adjusted model from the empirical model
⎛ 1 ⎞ ⎛ Ex. ⎞
B×S =⎜ ⎟×⎜ ⎟ B*S
⎜ ⎟
⎝11.7×(RR )1.2 ×(K )4.1 ⎠ ⎝S.G.×H ⎠
80 IC
4. The uniformity index (with above burden and spacing)
⎛ B⎞⎛ S/B⎞0.5 ⎛ W ⎞⎛ L −L ⎞0.1 L
N =⎜2.2−14 ⎟⎜1+ ⎟ ⎜1− ⎟⎜ B C +0.1⎟ ×(1.1 or 1.0) B*S
⎜ ⎟
⎝ D⎠⎝ 2 ⎠ ⎝ B ⎠⎝ L +L ⎠ H
B C
5. Characteristic size (with above P50)
P50
X = P50
C (0.693)1/N
6. Percent passing(%)
X
P =100(1−R) =100(1−exp(− )N) Predicted Size Distribution
P X
C
The previous flow chart shows the mechanism of the following simulation program in
Figure 5.1. Using this mechanism, the following simulation can be made.
The suggested application is to obtain the proper burden and spacing with given hole
diameter, bench height, explosives amount, and the kind of explosives as the change of the
rock and the bench structure.
Above used equation in second blank of the flow chart, E = 11.7 RR 1.202 K 4.14 was
SE 80 IC
obtained from 10 blasting data analysis, but another equation, (P50) = 0.5 (P80) is just from
6 blasting data analysis, both models were derived from the different number of blasting
data. However, the above application is just for showing how to make the model practical
using an empirical equation model. In addition, especially, the given factors (explosives
51 |
Virginia Tech | Chapter 6 Conclusion and Future Work
6.1 Research Summary
The following sequence was followed in the research:
1. Collect blasting image data and rock sample.
Bench face image sampling was conducted before blasting, and muckpile image sampling
was done after blasting. For sampling images of the bench face and the muckpile, same
scale materials were used from image to image. In addition, the angle between the bench
face and the camera was kept perpendicular, and rock sampling was conducted after
blasting. The most important factor for sampling image is consistency.
2. Conduct the END test with sampled rock.
Using suggested END test by Donovan (2003), the mode I fracture toughness value, K
IC
could be obtained easily and quickly. The obtained K value is as follows:
IC
Table 6.1 Fracture toughness, K , and Tensile strength, σ
IC t
σ(MPa)
Location K IC(MPa*m^0.5) t
Bosung 1.539±0.100 16.94±1.92
Sanyang 1.724±0.193 17.45±2.57
Pittsboro 1.889±0.398 22.48±4.70
Boxley 1.230±0.216 14.72±1.23
3. Analyze collected data using image analysis, SPLIT.
The setting in SPLIT depends on the users, but there should also be consistency from image
to image analysis. Following is the summarized data from image analysis of the bench face
and the muckpile, and specific explosives energy from each blasting pattern.
53 |
Virginia Tech | Table 6.2 Blasting data from image analysis, END test, and the blasting pattern
E K F20 F50 F80 P20 P50 P80 RR RR RR
SE IC 20 50 80
Bosung1 103 1.539 165 410 810 46 255 493 3.6 1.6 1.6
Bosung2 103 1.539 245 586 937 23 214 428 10.8 2.7 2.2
Bosung3 107 1.539 196 420 821 91 226 431 2.2 1.9 1.9
Bosung4 107 1.539 158 285 467 72 209 335 2.2 1.4 1.4
Bosung5 110 1.539 178 330 551 93 252 499 1.9 1.3 1.1
Pittsboro 387 1.724 158 281 483 40 100 187 4.0 2.8 2.6
Boxley 268 1.230 514 929 1512 69 161 304 7.4 5.8 5.0
Sanyang1 268 1.889 102 192 334 55 179 350 1.8 1.1 1.0
Sanyang2 292 1.889 102 196 382 88 192 290 1.2 1.0 1.3
Sanyang3 336 1.889 114 211 360 24 89 199 4.8 2.4 1.8
Sanyang4 222 1.889 50 91 149 21 82 163 2.4 1.1 0.9
Sanyang5 210 1.889 81 164 309 31 144 265 2.6 1.1 1.2
4. Conduct regression using SAS or EXCEL with suggested equation form.
Using the assumed equation form to predict the specific explosives energy, regression was
examined using SAS and EXCEL and the same constants were obtained. Actually, there
were about 80 kinds of analysis trials, from which the following results were selected.
E = 11.7 RR 1.202 K 4.14
SE 80 IC
E = 15.0 RR 0.86 K 4.02
SE 50 IC
E = 49.0 RR 0.20 K 2.19
SE 20 IC
Where:
E is specific explosives energy (wh/ton).
SE
RR is the reduction ratio based on 80% passing.
80
F80 is 80% passing size from the bench face image analysis (mm).
P80 is 80% passing size from the muckpile image analysis (mm).
RR and RR are same as RR .
20 50 80
F50 is 50% passing size from the bench face image analysis (mm).
K is the Mode I fracture toughness of rock (MPa*m1/2).
IC
In addition, the relationship between P50 and P80 was derived from the Rosin-Rammler
model.
54 |
Virginia Tech | 1/N
⎛ln0.5⎞
P50 =⎜ ⎟ P80
⎝ln0.8⎠
P50 = 0.5 (P80)
5. Use the empirically obtained equation models for making the site-specific blasting model.
In this research, the following two equation models were selected for this application.
E = 11.7 RR 1.202 K 4.14
SE 80 IC
P50 = 0.5 (P80)
6. Make the simulation model based on Visual Basic.NET program.
For convenience of usage, the language program VB.NET was used and a simulation
program was created. With this simulation program, proper burden and spacing can be
predicted for the target fragment size, P80, and P50 is predicted automatically. In addition,
the size distribution is predicted for the given blasting factors. Using this predicted size
distribution in reverse allowed us to adjust burden and spacing and target size, P80, in the
muckpile to achieve the desired size distribution.
For an example of the program usage,
1. At the given site, a blasting engineer wants to have the specific target fragment size, P80,
in the muckpile which is based on previous blasting data analysis at this site.
2. The rock properties and the bench face structure are totally different at each blast, but the
bench height and hole diameter are usually the same at each blast within the given blasting
site. Therefore, the blasting engineer uses the same blasting pattern (hole diameter, boring
depth, explosives, and explosives amount) except burden and spacing at each blast.
3. The blasting engineer conducts END wedge splitting test to obtain K , and uses the
IC
image analysis program on the bench face to obtain the in-situ block size information.
55 |
Virginia Tech | 4. The equation models, obtained by the current research, are used to get the proper burden
and spacing for the target fragment size, P80, in the muckpile.
Using this program, it is important to note that the blasting parameters, with the exception
of burden and spacing, are set equal to those used previously at the site.
6.2 Conclusion
Two main ideas were combined in this study. First, the whole bench face structure
information may be obtained by using an image-processing program, and this information
can be used for the blasting model. Secondly, mode I fracture toughness, K , can be used
IC
to represent the rock properties in the blasting energy prediction model.
In the research, these two ideas were investigated to determine whether these ideas are
practical. The largest drawback in this investigation was the small amount of blasting data.
To create a new empirical blasting model and to test the new ideas in this blasting model,
twelve pieces of blasting data from four quarries was not enough.
However, if we consider that obtaining blasting data requires much time and expense, the
data, although insufficient, may be used to determine certain possibilities.
Basically, if we consider the concept of RR, E , and K , in the blasting energy prediction
SE IC
model, then the following relationships are reasonable and may be hypothesized:
- E and RR will be in proportion.
SE
- E and K will also be in proportion.
SE IC
- The in-situ block size, F and E will be in inverse proportion because of leaking energy
SE
of the explosives.
56 |
Virginia Tech | There follow the main empirical equation models obtained from the blasting data analyzed
in this research:
E = 11.7 (RR )1.202 K 4.14
SE 80 IC
P50 = 0.5 (P80)
The above equations were not generalized because of the lack of data. However, the
research results are consistently reasonable, especially the findings (1) that E and K
SE IC
with given RR are in proportion and (2) that E and RR are in proportion with given K ,
80 SE IC
and (3) that P50 and P80 are in proportion within a given bench height and hole diameter.
Using SPLIT on the bench face to get in-situ block sizes (F) was the first challenge. At each
image, the result of the SPLIT analysis reveals significant variability, but it also
demonstrated consistency on the whole, and it is a reasonable challenge because the
equation model is reasonable based on data analysis. Furthermore, in a blasting model, the
actual properties of the rock may be replaced by the rock fracture toughness, K to predict
IC
the energy consumed in blasting.
Thus, although blasting data is insufficient to determine a model and verify any conclusion,
the above hypotheses are supported by twelve (12) blasting data analyses, and are both
reasonable theoretically and practical. The data shows at least the possibility that the
proposed ideas are valid.
Application
Burden and spacing are two of the most important factors in the blasting model because
these factors are easy to manipulate, whereas other factors are harder to control with given
bench height and hole diameters. Therefore, application of the model focused on getting the
proper burden, spacing for target fragment size, and, further, size distribution prediction of
the muckpile after blasting.
57 |
Virginia Tech | Using the above equation models, proper burden and spacing for the desired target
fragment size was obtained. The following equation is adjusted to show this:
⎛ 1 ⎞ ⎛ Ex. ⎞
B×S =⎜ ⎟×⎜ ⎟
⎜ ⎟
⎝11.7×(RR )1.202×(K )4.14 ⎠ ⎝S.G.×H ⎠
80 IC
In the above equation, RR is F80/P80. Thus the desired target fragment size, P80 was
80
considered, and F80 was obtained from the bench face image analysis. In addition, rock
property, K , and the amount of explosives per given bench height were considered in this
IC
equation. Size distribution using Kuz-Ram was predicted given this burden and spacing
automatically in the simulation program.
This simulation program is not generalized because of the shortage of data, but it is a site-
specific model program. Although it will take time and effort to determine, this kind of site-
specific model program may be more powerful than the generalized blasting fragmentation
prediction program, within a given blasting site, due to variable rock properties and
structure in the site.
Following are the characteristics of the described simulation program in this research.
- Mode I fracture toughness was used as rock property for blasting energy prediction.
- Changed bench face structure was measured easily and quickly by SPLIT at each blasting.
- Proper burden and spacing was obtained for the target fragment size, P80, in each given
blasting site, not just for the prediction of rock fragmentation, but for practical reasons
appreciated by blasting engineers.
The author hopes to show in the future uses to which the obtained empirical equation
models might be put, so the research is not completed. The suggested simulation program
merely demonstrates the capability of the obtained empirical blasting equation models to
obtain proper burden, spacing, and size distribution predictions after blasting.
58 |
Virginia Tech | 6.3 Limitation in the Research and Future work
Muckpile image Sampling
In the muckpile image sampling, although the main sampling strategy in this research is
consistency and this method is not time consuming, the findings are statistically inadequate
accurately to represent the muckpile. To have better muckpile image samples to
characterize the entire muckpile, obtaining images from the hauling truck is better, but this
will add to the amount of time needed to perform the sampling.
DATA
In this research, there are initially three (3) givens in the empirical blasting model:
1. The use of K in the blasting model.
IC
2. The use of conceptually new factors, E and RR.
SE
3. The use of SPLIT for getting the in-situ block size of a bench face.
However, the empirical model consisted of just twelve (12) blasting data. More blasting
data is needed to verify the findings regarding blasting energy prediction and to suggest
additional aspects of the new empirical blasting model.
P80
To obtain consistent blasting results, the author chose the target fragment size, P80.
However, this one size, P80, alone is not enough to represent the blasting result. The
muckpile should be considered in its entirety to estimate size distribution.
However, finding P80 in the muckpile image analysis may be related with optimal blasting
results. In a usual quarry blasting, if P80 is between 0.2~0.3m, then it would be excellent.
However, the result ultimately depends on the situation present in each quarry, i.e., each
quarry’s particular variables. Additional research is needed to demonstrate this to the fullest
extent.
59 |
Virginia Tech | P50 and P80
In the Rosin-Rammler model, the relationship between P50 and P80 is in proportion, as
follows:
1/N
⎛ln0.5⎞
P50 =⎜ ⎟ P80
⎝ln0.8⎠
As the change of uniformity index, N, P50 is in proportion with P80. The empirical
equation model of P50 and P80 is also in proportion to the Rosin-Rammler model.
Therefore, the empirical model derived from the collected blasting data may be used in the
research model. Further research using larger amounts of blasting data is required to resolve
this issue and determine a generalized equation.
P20 and Fines
P20 in the size distribution of the muckpile was used to determine the relationship between
rock properties and fines. However, this trial failed because in general the image processing
system underestimates fines, and the available data is insufficient to provide a firm
conclusion. As a result, obtaining good data about fines is not possible using available
image processing systems. To resolve this issue, a sieving test of the muckpile must be
done after blasting, and more investigation is needed.
60 |
Virginia Tech | References
Adel, G., Kojovic, T., Thornton, D. (2004) Mine-to-Mill Optimization of Aggregate
Production-Annual report Department of Mining and Mineral Engineering,Virginia Tech
and JKTech-JKMRC Commercial Division, The University of Queensland.
Adel, G. (2004) Abstract of Project “Mine-to-Mill Optimization of Aggregate
Production” Department of Mining and Mineral Engineering, Virginia Tech.
Allis-Chalmers. (1985) Allis-Chalmers High Energy Crushing Test System User Manual.
Version 1.0
Bearman, R.A., Pine, R.J., and Wills, B.A. (1989) Use of fracture toughness testing in
character the comminution potential of rock. Proceedings of MMIj/IMM Joint
Symposium, Kyoto, 161~170
Chong, K.P. and Kuruppu, M.D. (1984) New Specimen for Fracture Toughness
Determination of Rock and Other Materials, International Journal of Fracture, Vol. 26, pp.
R59-R62.
Chung, S. H. & Katsabanis, P.D. (2000) Fragmentation prediction using improved
engineering formulae. International Journal of Blasting and Fragmentatioin.
Clark, G. B. (1987) Principles of Rock Fragmentation John Wiley and Sons, New York,
Chichester, Brisbane, Toronto, Singapore.
Cunningham, C. (1983) The Kuz-Ram Model for the Prediction of Fragmentation from
Blasting, Proceedings of the international Symposium on Rock Fragmentation and Blasting,
Lulea, Sweden.
Cunningham, C. (1987) Fragmentation Estimations and the Kuz-Ram Model – Four years
on, Proceedings of the second international Symposium on Rock Fragmentation and
Blasting, Keystone, Colorado.
Cunningham, C. (1996) Optical fragmentation assessment - A techanical challenge
Proceeding Measurement of Blast Fragmentation, Balkema, Rotterdam
Dahlhielm, S. (1996) Industrial applications of image analysis – The IPACS system
Proceeding Measurement of Blast Fragmentation, Balkema, Rotterdam
Donovan, J. G. (2003) Fracture Toughness Based Models for the Prediction of power
consumption, product size, and capacity of jaw crushers Doctor of philosophy dissertation,
Mining and Mineral Engineering, Virginia Tech, Blacksburg
61 |
Virginia Tech | Kojovic, Toni (2006) Private communication with an expert in JKMRC
Kuznetsov, V.M. (1973) The mean Diameter of the Fragments Formed by Blasting Rock,
Soviet Min. Sci..
Lilly, P.A. (1986) An empirical method of assessing rock mass blastability Proceeding
Large Open Pit Mining Conference, pp. 89~92.
Lim, I. L., Johnston, I. W., Choi, S. K., Boland, J. N. (1994) Fracture testing of a Soft Rock
With Semicircular Specimens Under Three Point Bending International Journal of Rock
Mechanics, Mining Sciences, and Geomechanics Vol. 31, pp.185~197
Liu, Q., Tran, H. (1996) Comparing system- Validation of Fragscan, Wipfrag and Split
Balkema, Rotterdam
Maerz N. H. (1996) Image sampling techniques and requirements for automated image
analysis of rock fragmentation Proceeding Measurement of Blast Fragmentation, Balkema,
Rotterdam
Maerz, N. H., Palangio, T. C., Franklin, J. A. (1996) WipFrag image based granulometry
system Proceeding Measurement of Blast Fragmentation, Balkema, Rotterdam
Murakami, Y. (1987) Stress intensity factors handbook. Pergamon Press
Norton, B (2005) Private communication with an expert in SPLIT engineering
Ozdemir, K., Kahriman, A., Karadogan, A., and Tuncer, G. (2003) Blasting Fragmentation
Assessment and control using the split digital image analysis system International
Conference on Earth Sciences and Electronics.
Rustan, Agne (1998) Rock Blasting Terms and Symbols A.A. Balkema, Rotterdam,
Brookfield
Schleifer, J., Tessier, B. (1996) FRAGSCAN: A tool to measure fragmentation of
blasted rock Proceeding Measurement of Blast Fragmentation, Balkema, Rotterdam
Schmidt, R.A. (1980) A microcrack model and its significance to hydraulic fracturing and
fracture toughness testing Proceedings of the 21st US Symposium on Rock Mechanics,
pp.581~590
Schmidt, R.A and Rossmantihm, H.P (1983) Basics of Rock Fracture Mechanics, Rock
Fracture Mechanics
63 |
Virginia Tech | MODELING OF CO SEQUESTRATION AND ENHANCED GAS
2
RECOVERY IN COMPLEX UNCONVENTIONAL RESERVOIRS
Foteini Vasilikou
ABSTRACT
Geologic sequestration of CO into unmineable coal seams is proposed as a way to
2
mitigate the greenhouse gas effect and potentially contribute to economic prosperity
through enhanced methane recovery.
In 2009, the Virginia Center for Coal and Energy Research (VCCER) injected 907
tonnes of CO into one vertical coalbed methane well for one month in Russell County,
2
Virginia (VA). The main objective of the test was to assess storage potential of coal
seams and to investigate the potential of enhanced gas recovery. In 2014, a larger scale
test is planned where 20,000 tonnes of CO will be injected into three vertical coalbed
2
methane wells over a period of a year in Buchanan County, VA.
During primary coalbed methane production and enhanced production through CO
2
injection, a series of complex physical and mechanical phenomena occur. The ability to
represent the behavior of a coalbed reservoir as accurately as possible via computer
simulations yields insight into the processes taking place and is an indispensable tool for
the decision process of future operations. More specifically, the economic viability of
projects can be assessed by predicting production: well performance can be maximized,
drilling patterns can be optimized and, most importantly, associated risks with operations
can be accounted for and possibly avoided.
However, developing representative computer models and successfully simulating
reservoir production and injection regimes is challenging. A large number of input
parameters are required, many of which are uncertain even if they are determined
experimentally or via in-situ measurements. Such parameters include, but are not limited
to, seam geometry, formation properties, production constraints, etc.
Modeling of production and injection in multi-seam formations for hydraulically
fractured wells is a recent development in coalbed methane/enhanced coalbed methane
(CBM/ECBM) reservoir modeling, where models become even more complex and
demanding. In such cases model simulation times become important. |
Virginia Tech | ACKNOWLEDGMENTS
There are no words to express my gratitude to all the people who have supported me in this
challenging period of my life. I sincerely thank Professor Michael Karmis for his guidance and
advice in all aspects of my educational and professional life. Without his support I wouldn’t be
here.
I would like to thank Dr. Nino Ripepi for his support during the past three years. He has
been an excellent advisor and teacher and helped me overcome many difficulties along this
journey.
I cannot thank enough Professor Zach Agioutantis for his skill, teaching, motivation,
guidance and patience. Professor Agioutantis has been an amazing mentor.
I would like to express my gratitude to Dr. Kray Luxbacher for her advice and guidance, and
to Professor Gerry Luttrell who was the first person to welcome me to the Mining Engineering
Department and guide my studies.
A special thank you to Steve Schafrik, whose exceptional technical and coding skills were of
utmost importance in this effort.
I would also like to thank Dr. Cigdem Keles for her valuable help and support with the
reservoir modeling.
Finally, a big “Thank you” to everyone at the Virginia Center for Coal and Energy Research
who has contributed with their support in this work and especially my dear friend and advisor
Dr. John Craynon.
I would also like to express my love for my amazing mother Alkioni, who never left my side
and crossed the Atlantic Ocean numerous times to be by my side; to my cousin, Dr. Demetrios
Vavvas, who has been a father, an advisor and a friend to me; and to my brother Ioannis for his
constant love and support.
I am thankful to my friends Christos, Constantinos and Korina for always being there for
me.
v |
Virginia Tech | PREFACE
This research effort comprises the four major tasks summarized below. These tasks are
addressed in a set of scholarly works which are either published or will be submitted for
publication.
Task 1. Investigation of dynamic evolution permeability models
The main objectives for Task 1 are to:
a) Critically review permeability change models which are proposed in the literature for
coalbeds with respect to primary and enhanced production.
b) Implement a permeability change model in single well models by coupling a
reservoir simulator and appropriate geomechanical code.
Task 2. Sensitivity analysis for numerical models
The main objectives for Task 2 are to:
a) Perform sensitivity analyses to study the behavior of the reservoir models when
production and injection occur in multiple zones.
b) Examine the response of the reservoir models with respect to varying input
parameters that affect the volumetrics properties.
c) Validate well and reservoir characteristics in the models that have an effect on the
production mechanism.
Task 3. Comparison of different approaches that model well stimulation in reservoir
simulations
The main objectives for Task 3 are to:
a) Develop models where a negative skin factor is assigned.
b) Explicitly model hydraulic fractures in reservoir models.
c) Study and compare enhanced flow properties and the effect on injection depending
on the well stimulation approach.
Task 4. Predict enhanced methane recovery, CO flowback and breakthrough
2
The main objectives for Task 4 are to:
xiii |
Virginia Tech | THE APPLICATION OF CONSTITUTIVE LAWS TO MODEL THE
DYNAMIC EVOLUTION OF PERMEABILITY IN COAL SEAMS FOR
THE CASE OF CO GEOLOGIC SEQUESTRATION AND ENHANCED
2
COAL BED METHANE RECOVERY1
Foteini Vasilikou, Virginia Center for Coal and Energy Research, Virginia Tech,
Blacksburg, VA
Nino Ripepi, Virginia Center for Coal and Energy Research, Virginia Tech,
Blacksburg, VA
Zach Agioutantis, Virginia Center for Coal and Energy Research, Virginia Tech,
Blacksburg, VA
Michael Karmis, Virginia Center for Coal and Energy Research, Virginia Tech,
Blacksburg, VA
Abstract
Injection and storage of carbon dioxide (CO ) in deep unmineable coalbeds decreases
2
anthropogenic greenhouse gas emissions and presents a financially viable solution by enhancing
recovery of coalbed methane (ECBM). Coalbeds are commonly characterized by a dual porosity
system, which is comprised of a network of natural fractures (cleats) and matrix blocks of coal
exhibiting highly heterogeneous porosity. The gas transport through the cleat system is governed
by Darcy's Law. This study reviews and critically evaluates available models for describing
coalbed permeability that can be applied to calculate gas flow in such systems. In addition, the
1 The Application of Constitutive Laws to Model the Dynamic Evolution of Permeability in
Coal Seams for the Case of CO Geologic Sequestration and Enhanced Coal Bed Methane
2
Recovery. F. Vasilikou, N. Ripepi, Z. Agioutantis, M. Karmis.
This paper has been published already in the proceeding of the 29th Pittsburgh Annual Coal
Conference in 2012.
Foteini Vasilikou researched and prepared this manuscript, with Nino Ripepi, Zach
Agioutantis and Michael Karmis providing technical and editorial input.
1 |
Virginia Tech | potential of using geomechanical models to better account for the physical processes that occur
during coalbed methane production and CO injection and storage is also investigated. The
2
results of this review can be used for evaluating modeling approaches when employing reservoir
simulators to simulate injection and storage in ECBM cases.
Introduction
Historically, recovery of natural gas or coalbed methane (CBM) is described as early as
1858. By 1953 degasification of the Pittsburgh seam was underway for safety reasons. By 1980,
companies began profiting from deep coal seams where natural gas was of interest as a clean and
domestic energy source, whether the coal was mineable or not (Bodden and Ehrlich, 1998;
Ayers, 2002; Zarrouk, 2008; Liu, 2011). Coal seams are termed as unmineable where mining is
considered to be infeasible given foreseeable technology, costs, sales prices, inadequate seam
thickness, poor areal continuity, adverse geology (steeply dipping, rolls, faults), poor coal
quality, excessive depths and other reasons. Typical depths for unmineable coal seams are 300-
900 m (Siriwardane, 2008). The most methane that could be recovered by the pressure depletion
method is not anticipated to be larger than 50% of gas-in-place, even after several decades of
production (Puri and Yee, 1990). Hence, in the 90’s enhanced coalbed methane (ECBM)
production via injection of CO in the unmineable coal seams was suggested as a more efficient
2
way of extracting a larger fraction of methane in place without having to overly reduce reservoir
pressure. In 2005 under the Kyoto Protocol, an international agreement linked to the United
Nations framework convention on climate change, a cap on reduction of greenhouse gas
emissions 5% below the 1990 greenhouse levels, further increased the interest in sequestering
CO in coalbeds. Thus, enhanced coalbed methane has the twofold benefit of enhanced
2
production and redeeming greenhouse gases pollution. In Australia, Canada, China, Poland and
the USA, a number of ECBM projects are ongoing (Hamelinck et al., 2003; Law et al., 2002;
Damen et al., 2005; Zarrouk, 2008).
Large ‘natural gas’ or methane reserves are existent all over the world such as in Canada,
Russia, China, the United States and Australia.
There exist a number of physical processes which occur during production of coalbed
methane and injection and storage of CO in coal seams. Coalbeds are characterized by a dual
2
porosity system, which is comprised of a network of natural fractures (cleats) and matrix blocks
2 |
Virginia Tech | of coal exhibiting highly heterogeneous porosity. The gas transport through the coal matrix
micropores (primary porosity), driven by the concentration gradient, is controlled by Fickian
diffusion. The flow of gas and water through the cleat system (secondary porosity) is governed
by Darcy's Law. Methane is initially stored in the adsorbed state, within the porous structure of
the coal matrix blocks, which is usually described by a Langmuir-type isotherm. During coalbed
methane production, reservoir pressure is decreased through dewatering, allowing methane
molecules to desorb from the internal coal matrix (Fickian diffusion) and travel through the cleat
structure (Darcy Flow). The opposite process occurs during CO injection. This results in an
2
inverse diffusion process, where CH molecules are desorbed from the coal matrix and replaced
4
by CO molecules. The combination of reservoir dewatering and its associated depressurization,
2
CH desorption and CO adsorption causes alterations in the stress regime acting on the coal
4 2
matrix. Gains or losses in water and gas-relative permeability can be noted, but vary greatly in
accordance with the geological characteristics of the coal.
Coalbed Characteristics
Coal is a highly heterogeneous porous medium that contains micropores (<2 nm), mesopores
(between 2 and 50 nm), macropores (>50 nm) and natural fractures formed during coalification.
(Wang 2009; Shi and Durucan, 2005). In the literature, coalbeds or coal seams are characterized
by a dual-continuum system comprised by the porous coal matrix and cleats (fractures) (Liu and
Rutqvist, 2009). Cleats are considered to be a system of densely spaced, orthogonal fractures;
cleats are comprised of face and butt cleats. Face cleats are long, well-developed, almost planar
fractures that extend parallel to each other. Butt cleats, run at an angle of roughly 90 degrees to
the face cleats and usually terminate at them (Figure 1). Face and butt cleats are approximately
perpendicular to the bedding planes (Liu 2011; Gu 2009).
3 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.