University
stringclasses 19
values | Text
stringlengths 458
20.7k
|
|---|---|
Colorado School of Mines
|
achieve these two goals, the thesis will include:
Background Information
In Chapter 2 the physics governing electromagnetic (EM) radiation is explained
with a particular focus on the propagation properties and EM characteristics of a
subterranean mine. In addition an examination of recent search and rescue efforts
using teleoperated robotics will be addressed along with robotic systems utilizing
Wi-Fi technology.
Design and Construction Process
In Chapter 3 the original wireless control system for the Bobcat front-end loader
is analyzed for potential use in a subterranean environment and shown why it must
be modified. The new Wi-Fi based mesh network is explained. The Bobcat remote
control system is reverse engineered and converted into a Ethernet compatible device
for use with an ad hoc wireless system. The design process for the conversion covers
hardware signal interception, microprocessor configuration and communication, Eth-
ernet network configuration, and how to convert the existing signals to comply with
existing Bobcat electronics. In Chapter 4 the construction of the design elements is
shown to create a better understanding of how the design process works.
Experimentation and Analysis
In Chapter 5 the first set of experiments performed verify the propagation char-
acteristics of Wi-Fi EM radiation described by Chapter 2. Secondly, several tests
are performed working the way up in complexity towards a complete mock rescue
scenario utilizing a teleoperated Bobcat front-end loader and an autonomous vehi-
cle acting as a wireless relay. The first few steps tested the Wi-Fi mesh network
and the ability to navigate by signal strength alone. Successive experiments tested
the autonomous actions of the follower vehicle and its ability to navigate by signal
strength and proximity sensors. The Bobcat was tested individually to showcase and
verify the safety features necessary for full testing. Lastly, all systems were tested
4
|
Colorado School of Mines
|
CHAPTER 2
BACKGROUND
To understand and overcome the difficulties of this project, a thorough under-
standing of the subterranean mining environment is needed. This environment is
challenging not only because of its ability to destroy or disable vehicles but also be-
cause of the high attenuation of wireless signals. In this chapter, we will discuss the
physical challenges and the laws of physics that govern the communications and wire-
less propagation within an underground mine. Lastly, we will cover some of the past
work that has been done in the field of teleoperated robotics for search and rescue
within mines and also in the field of teleoperation with general Wi-Fi networks.
2.1 Subterranean Environmental Challenges
Undergroundminesarephysicallytaxingenvironments. Anyvehiclewithinamine
will likely encounter rocky terrain, mud, and water on a regular basis and might even
encounter rock fall, fire, low oxygen, flooding, thick mud, smoke, etc. in an emergency
situation. It may be impossible to build a vehicle that can withstand all the worst
case scenarios of a rescue mission, but it minimally should be ruggedized for minor
collisions, high ground clearance, sufficient traction, and water proofing. In addition
to all the vehicle’s physical challenges, a very large communications problem exists
within subterranean passageways. In the past many search and rescue robots had
to use cables for high bandwidth communications and for high power consumption.
Today, wireless systems are at a point where it is feasible,using commercial radio
hardware, to communicate within a subterranean mine, but there are still significant
difficulties.
7
|
Colorado School of Mines
|
2.2 Subterranean Radio Wave Propagation
Thepropagationofwirelesssignalswithinanundergroundmineortunneldepends
on many parameters including the cross-sectional shape of the mine, the composition
of the surrounding rock, the curvature of tunnels, the position of antennas, and the
signal frequency, among many others. In our project we use 802.11b/g Wi-Fi for the
high bandwidth communication that allows real time video for teleoperation. This
standard Wi-Fi protocol uses a frequency of 2.4 GHz (2.412 GHz for Ch 1), which
has a small, but not negligible wavelength (12.43 cm) in comparison to the dimension
of the mine (3 m to 10 m). This wavelength can by calculated from the relation
c 2.998×108
λ = = = 12.430cm. (2.1)
f 2.412GHz
2.2.1 Radio Wave Propagation Mechanisms
The three basic mechanisms that impact the propagation of radio waves are re-
flection, diffraction, and scattering [3].
Reflection occurs when an electromagnetic (EM) wave impinges upon an object
with dimensions that are much greater than the wavelength of the EM wave. Spec-
ularly reflected waves behave according to Snell’s law, where the incoming angle of
incidence is equal to the outgoing angle of reflection. A related mechanism is trans-
mission, where the non-reflected energy in the wave penetrates into a medium. In
the case of a subterranean tunnel, the amount of the radio wave that reflects and
the amount that transmits is governed by the the dielectric properties of air and the
surrounding rock, along with the angle of incidence.
Diffraction occurs when a wave impinges upon the edge of a larger object, and
creates secondary wavefronts to ultimately produce a bending effect of the original
wave according to Huygen’s Principle [4]. Diffraction allows the possibility of wireless
communications when an object blocks the line of sight between a transmitter and
8
|
Colorado School of Mines
|
receiver, depending on the geometry of the obstacle and the properties of the incident
wave.
Scattering is another important propagation mechanism and occurs when a wave
collides with an object that is physically smaller then the wavelength of the impeding
wave. This collision creates many low power waves, all propagating in random di-
rections. This is particularly important in a subterranean mining environment where
the walls are cut by explosions and mining tools, resulting in very jagged and random
surfaces. The jagged walls can be thought of as many small objects, each creating a
scattering surface for incoming waves.
The net effect of these three mechanisms on an radio wave inside a subterranean
tunnel can be complicated. As we shall see, these mechanisms can create attenuation,
bending around corners, and even concentrations of signal called waveguiding.
2.2.2 Radio Signal Attenuation
Radio waves lose power as they propagate from transmitter to the receiver. These
losses can be described by many models depending on distance from the transmitter
to receiver, the far field range of the transmitter, and the wavelength. These models
arise from theoretical equations, statistical models, and experimental adjustments.
In an open line-of-sight environment, radio waves encounter free space loss of
power from an isotropic transmitter to an isotropic receiver dependent upon Friis’
model for free space loss [3]
P λ2
r
PL(d) = = , (2.2)
P (4π)2d2
t
where PL(d) is the path loss, P is the transmitted power, P is the received power,
t r
λ is the wavelength, and d is the distance between the transmitter and receiver.
However, an underground mine is far from the ideal free space model and will re-
quire a more complicated modeling approach for describing path loss. A subterranean
9
|
Colorado School of Mines
|
is the distance between the transmitter and receiver, and (cid:104)κ2(cid:105) is the median electric
fieldamplitudegainoftheradiochannel, whereκvariesfrompoint-to-pointaccording
to a PDF like that shown in Figure 2.1 and described by Equation (2.3). Lastly the
path loss exponent n creates the inverse power function and is derived from empirical
measurements describing how fast energy is lost from the transmitter to receiver as
a function of distance. In free space, the path loss exponent is exactly two. In most
scattering environments, the statistically derived path loss exponent, “n,” is typically
above two but can be below two for subterranean tunnels, because of waveguiding
effects. Table 2.1 shows common measured path loss exponents and some path loss
exponents measured from the Edgar experimental mine [5].
Table 2.1: Path loss exponents for various environments.
Free Space/ Vacuum 2.0
City 3 - 6
Edgar mine straight tunnel 1.2 - 1.6
Edgar mine bent tunnel 2.9 - 3.3
2.2.3 Waveguiding
A waveguide is a pipe that focuses the energy of a spherical radiating source. This
focusing of energy is due to the reflection of EM radiation off the surrounding walls,
forcing the waves to bounce from wall to wall in a zig-zag pattern. Ultimately the
vast majority of the radiating energy is forced to propagate further into the hollow
pipe. The effectiveness of the waveguide is determined by the amount of reflectivity
it has, which is dependent on the dielectric constant and the smoothness of the walls,
relative to the wavelength of the EM radiation. Several measurements have been
taken to determine that many subterranean mines do exhibit waveguiding effects at
Wi-Fi frequencies [5]. When this occurs it places the path loss exponent below the
ideal value of two for free space. However, past research has shown the path loss
11
|
Colorado School of Mines
|
exponent reduces below two only for straight tunnel measurements. This is because
the waveguide effect needs shallow angles of reflection and the bending losses for for
sharp corners is very high. [5]
2.3 Robotic Rescue Efforts
There have been many endeavors to create robots suitable for search and rescue
purposes. Robots are typically built to specialize in a single purpose, and search
and rescue robots are no different. Snake robots have been designed for search and
rescue in hard to reach places, like the rubble of a collapsed building [6]. Larger
wheeled robots have been used to assist police and armed forces to seek out hostages
or victims in hostile buildings and other urban search and rescue locations. There
have even been a few attempts to use robots for search and rescue in subterranean
mining disasters.
For example, during the West Virginia Sago mining disaster in 2006, rescue work-
ers attempted to utilize a Remotec ANDROS Wolverine Robot, shown in Figure 2.2,
for search and rescue. This robot was designed for urban search and rescue and bomb
disposal, but had been modified for subterranean purposes. The 50 inch tall, 1200
pound robot had three cameras for navigation and surveillance, lighting, a manipu-
lator arm, air monitoring equipment, and up to 5000 ft of tethered cabling for data
transmission and power[1]. However, the robot at the Sago mine became immobilized
by mud and was no longer of assistance to the search and rescue team.
Another example was at the Crandall Canyon Mine in Utah where a mobile but
tethered robot built by Inuktun, shown in Figure 2.3, was sent down several vertical
boreholes to assess damage and look for victims. The robot was small, at about eight
inches wide, and used a tether for communications and power. It had two sets of
cameras; one fixed camera for guidance and another camera on a boom for greater
field of view and control.
12
|
Colorado School of Mines
|
Figure 2.2: Sago rescue robot: Remotec ANDROS Wolverine [1].
The robot encountered many problems including several scenarios where it could not
move past obtrusions in the borehole and water and debris fouling the camera lenses.
Ultimately the cable was severed by a boulder upon retrieval and the mission was
abandoned. However, the robot did produce some useful information and workers
where able to look around and determine where collapses had occurred in one section
of the mine. Unfortunately, it did not find any victims [7].
Other examples include the World Trade Center and most recently the mining
disaster at the Pike River coal mine in Greymouth, New Zealand.
2.4 Wi-Fi for Robotics and Disasters
Wi-Fi is very popular because of its availability, adaptability, and low cost. A few
commercial companies even make toys that utilize Wi-Fi control and video feedback
for teleoperated control of mobile toys [8].
There has also been a lot of active research using Wi-Fi for control over many
vehicles and robotic systems, including in disaster scenarios and especially urban dis-
aster zones[9]. Wi-Fi has many benefits that would be practical for disaster zones
13
|
Colorado School of Mines
|
Figure 2.3: Crandall rescue robot: Inuktun custom robot.
in which cables have been used in the past. Some benefits include already estab-
lished networks in urban zones and the benefits of not being tethered while still
having the bandwidth to stream video. Another active area of research is the use
of ad-hoc meshed Wi-Fi devices for rescue efforts. Many research systems alternate
between two Wi-Fi standards, 802.11a and 802.11g, to create a pseudo-meshing ef-
fect between wireless devices[10]. These pseudo-meshing systems are typically not as
robust as the products from commercial companies because commercially available
products can automatically adapt to changes in the environment and changes in net-
work configuration. In recent years commercially available Wi-Fi mesh products have
become available partially thanks to improved regulation of communications within
underground mines due to the Mine Improvement and New Emergency Response Act
(MINER Act) of 2006 from the Mine Safety and Health Administration (MSHA).
Wi-Fi has been one of the leading technologies that mining companies have placed
within mines thanks to the MINER Act. If the Wi-Fi architecture is already in place
inside a subterranean mine, then it is an extra resource for any search and rescue
14
|
Colorado School of Mines
|
CHAPTER 3
DESIGN
ThischaptercoversthedesignprocessforconvertingacommercialwirelessBobcat
front-endloadertoaWi-Fi/Ethernetcompatibleteleoperatedvehicle. Alongwiththe
conversion, this chapter discuses the Rajant Wi-Fi communication radios that were
chosen for the mesh network used in the MineSENTRY project, and some of the
design details of the Autonomous Mobile Radio (AMR).
3.1 Original Bobcat Remote Control System
Bobcat offers a remote control add-on for their line of front-end loader vehicles
shown in Figure 3.1(a). This remote control system is composed of a user inter-
face/transmitter unit and a receiver box. The receiver box plugs into a compatible
front-end loader and controls the vehicle via an onboard communications network.
The intended use of this system is for line-of-sight remote control, to protect the
driver in dangerous situations. An example can be seen in Figure 3.1(b).
The 2.4 GHz high frequency signal used by the Bobcat remote control system is
very highly attenuated around the corners of a subterranean tunnel. This is because
inside the corners of the mine there are extremely high bending losses, multifaceted
surfaces causing scattering, and a low coefficient of reflection for waves. A mesh or
ad hoc network is the only means for transmitting these high frequency waves far
into the mine. Unfortunately, the Bobcat remote control system is proprietary and
incompatible with Ethernet or Wi-Fi standards. Thus it was necessary to devise a
modification to the system so it would work over a mesh network.
17
|
Colorado School of Mines
|
so small that not much diffraction occurs around corners. Thus an ad hoc network
is needed to relay information throughout the mine environment, especially around
corners. There are several network devices available for creating ad hoc networks but
they lack easy adaptability and meshing abilities. As described in Chapter 1, mesh-
ing is the ability for one device to connect to multiple devices and for those devices
to connect to multiple devices, creating a redundant network to pass information
around physical breaks in the network. The company Rajant focusses on providing
Wi-Fitechnologiesforminingenvironments. Figure3.2showsRajant’smeshedWi-Fi
node call the Breadcrumb ME2. It is a highly adaptable Wi-Fi node capable of true
meshing and autonomously adapting to almost any physical network configuration.
Figure 3.2: Rajant Breadcrumb ME2 [11].
3.3 Bobcat Remote Control System Analysis
The general description of the original Bobcat remote control system is simple. It
is a transmitter and a receiver that communicate over a wireless channel. However,
the wireless communication protocol is very complicated because it uses a propri-
etary frequency hopping spread-spectrum transmission medium centered at 2.4 GHz.
This type of protocol is designed to prevent unwanted interceptions. The line-of-sight
range of the remote control system is about 500 meters. After the signal is captured
19
|
Colorado School of Mines
|
at the receiver, the data is interpreted and transferred to the vehicle using a standard
automotive wired network (CAN Bus). In the following sections we detail the anal-
ysis of each module within the original transmitter and receiver from Bobcat. This
description is included for the completeness of documentation and because signifi-
cant time and analysis was invested to determine the details of the original Bobcat
transmitter and receiver. Hopefully the information here allows follow up work to be
easier and more efficient.
3.3.1 Original Transmitter Analysis
The original remote control transmitter from Bobcat, seen in Figure 3.3, was
designed by the company Omnex.
(a) Remote control transmitter. (b) Transmitter internal circuitry.
Figure 3.3: Original remote control transmitter.
The system is designed in a modular fashion using parts previously created by Omnex
and is similar to other remote control systems by Omnex. The original remote control
transmitter can be broken down into the Main Board, Joysticks, Paddle Controls,
Switches, and Indicators. The interface between all five of these modules can be seen
in Figure 3.4.
20
|
Colorado School of Mines
|
Figure 3.6: Joystick printed circuit board.
Each joystick communicates with the main electronics board via a six pin con-
nector. The connector can be connected in a daisy chain so that only one connector
attaches to the main board. In the chain configuration all wires except the data out
are connected in parallel. The data out line for each joystick is passed into each
microprocessor down the chain where the last microprocessor transmits the data for
all the joysticks. This only happens once because there are only two joysticks.
The six pins correspond to:
Pin 1: + 9.6 V
Pin 2: N.A.
Pin 3: GND
Pin 4: Data out
Pin 5: Reference Pulse
Pin 6: Clock Pulse
The data out is a digital two’s complement pulse code modulation signal created
by a microprocessor on each joystick circuit. This data signal is triggered out by a
clock and reference pulse train that is provided by the main board via the six pin
connector.
Paddle Controls: The two paddles control the actuation of peripheral devices
that could be attached to the Bobcat front-end loader. The paddles are simple elec-
22
|
Colorado School of Mines
|
Each switch is either a single pole single throw (SPST) or a single pull double
throw (SPDT). Each signal is digital, requiring either one or two bits each. Figure 3.9
shows two peripheral switches that differ from the other standard switches, however
their operation is the same. The emergency stop (E-stop) switch is a very critical
switch and is simply a single pull single throw switch. The ignition switch is a pair of
single pull single throw switches for turning on the remote and another for starting
the vehicle ignition.
(a) E-Stop (top). (b) E-Stop (bottom). (c) Ignition.
Figure 3.9: Peripheral switches.
Indicator Board: An indicator LED board is attached to the interior of the re-
mote. It provides LED indication for antenna connectivity, E-stop, and battery level.
An additional feature of this board is a tilt sensor which will trigger the emergency
stop in case the user falls over.
3.3.2 Original Receiver Analysis
The receiver box is a plug in module that can attach to most Bobcat loaders and
control it remotely. The actual electronics, shown in Figure 3.10(b), are stored inside
a ruggedized outer shell shown in Figure 3.10(a).
The receiver box controls the Bobcat vehicle through the wired J1939 CAN-Bus
network. The receiver has control over individual components of the vehicle with
this type of network. The specific J1939 CAN network is an extended version of the
original CAN-Bus, a protocol designed for vehicles. In the front-end loader network
24
|
Colorado School of Mines
|
(a) Outer shell. (b) Interior electronics.
Figure 3.10: Original receiver.
Bobcat added several features to prevent theft and for general safety. Bobcat’s safety
features include the rapid changing of messages, which follow a encrypted pattern.
If any bit is not received at the correct time and in the correct pattern then the
vehicle shuts down. Reproducing these CAN-Bus messages with the correct pattern is
extremely difficult if not impossible. A month of work was dedicated to intercepting
the raw data, transferring it over a Ethernet/Wi-Fi network, and recreating these
CAN-Bus messages, to no avail. The extremely small time window did not leave room
for any errors during transfer, which occasionally happens with a wireless network.
More time was dedicated to interpreting and recreating simulated CAN-Bus messages
with a microprocessor, however, we were unable to crack the encrypted pattern and
did not receive any help from Bobcat or Omnex.
3.4 Design Solution Outline
In order to use the Rajant Wi-Fi mesh network, the Bobcat remote electronics
need to become Ethernet compatible. Figure 3.11 shows the best location to intercept
information is at the data input of the main board, in other words at the output of
the physical switches and joysticks.
25
|
Colorado School of Mines
|
The joysticks communicate with the main electronics board via a pulse code mod-
ulated data signal along with a trigger/reference line and clock line. Figure 3.15
shows all three signals with the reference signal in both and shown at different time
scales. The reference rising edge pulse starts the communication and the individual
clock pulses each trigger a single bit from the joystick PCB.
Figure 3.15: High speed signal from joysticks.
Another complication for reading the joysticks is that the pulses are too fast for
the digital port of the microprocessor since each pulse is only a few microseconds
in length. A way around this is to latch the data into a shift register and send
the collected data with a high speed data bus, specifically I2C. In order to get the
data out of the joysticks, a fake reference and clock signal need to be created by the
microprocessor. After the microprocessor sends this data it needs to read the data
captured in the shift register through the I2C communication line. The interaction
between the microprocessor, the I2C shift register board, and the joystick electronics
is shown in Figure 3.16
The I2C board receives the reference pulse, clock pulse, data from the joystick,
and a latch pulse. It then passes it through a shift register, latches this data after all
16 bits have occurred, then passes it to a parallel-to-I2C chip which sends the data
along the I2C data bus to the microprocessor.
29
|
Colorado School of Mines
|
Figure 3.19: Outline of receiver.
3.6.1 RX Physical Design
ThePCBthatactsasabufferbetweenthemicroprocessorandtheoriginalBobcat
electronics was designed to provide all the miscellaneous tasks that could not be
provided by the microprocessor board. It is mainly composed of transistor array
chips which steps down the logic level voltage from 5V at the microprocessor to 3.3V
at the Bobcat electronics. LEDs also provide an easy interface to detect changes in
logic level for all the switch positions. A rather important note is the order of pull
up or pull down resistors for the logic level components on the Bobcat are oriented
to lock the parking brake and bucket of the Bobcat during a scenario with loss of
communication to the microprocessor. In order to simulate the analog voltages from
a paddle on the remote control, a DAC is required. Unfortunately, the microprocessor
does not include a DAC so one had to be added externally. This PCB includes a chip
that converts I2C serial data into analog voltages. Two of the analog pins where used
to simulate the two paddles on the remote control. Power distribution was also added
to this PCB at three voltage levels. The following voltages are provided:
1.) 9.0 V to power the original Bobcat electronics
2.) 5.0 V for microprocessor power and logic
3.) 3.3 V for the original Bobcat electronics logic
33
|
Colorado School of Mines
|
3.6.2 RX Software Description
The receiver microprocessor gathers all the control data from the TCP/IP channel
where it distributes that information across several areas including: digital I/O, an
analog I2C ADC, interrupt based joystick data simulation. The software starts out
in an initialization section, where all the ports are configured. It then passes to the
main loop where it looks for a client to connect to it. When it receives a connection
with a client, it goes into a second loop that that executes the following tasks:
Task 1.) Verify the connection with the client.
Task 2.) Collect data from the client according to a fixed pattern.
Task 3.) Pass the data along to the digital ports, I2C line, and output compare variables.
During task 2, the connection is verified a second time with a data stream that follows
a set pattern. First a pair of start bytes is sent, then nine bytes of data, then a pair of
stop bytes. If any of the patterns are incorrect, it is flagged and the number of errors
is incremented. If the pattern was correct the number of errors is nulled. Therefore,
after multiple errors have occurred sequentially, the program steps into one of several
safety stops.
Safety stage 1: 8 < Errors < 20 Stop bucket and wheels.
Safety stage 2: 20 < Errors < 30 Lock bucket and set parking brake.
Safety stage 3: 30 < Errors Shut down vehicle.
An output compare subroutine is always running independent of the main loops.
This subroutine takes the current variable for joystick data and triggers a pulse code
modulated signal with precision timing on a digital output pin of the microprocessor.
A full block diagram of the software is shown in Figure 3.20.
34
|
Colorado School of Mines
|
3.9 Autonomous Mobile Radio (AMR)
The AMR is an autonomous vehicle that carries a Rajant Wi-Fi node. This node
communicates with the vehicle telling it when and where to move based on radio
signal strength measurements. This section covers the software for the Rajant radio
communication interface and a summary of the vehicle hardware.
3.9.1 Vehicle
This section is included for context. Details of the systems described here are not
a contribution of this thesis. An EZ-GO TXT model golf cart, seen in Figure 3.22,
was chosen as the base platform of an autonomous mobile radio (AMR) because it is
a commercially ruggedized and reliable vehicle, the emission free electric motor, the
long lasting battery life, and the easily modifiable platform for electronic actuation.
Figure 3.22: AMR.
After acquiring the base platform, the golf cart was modified for electric actuation of
brakes, steering, and motor control. A showcase of the steering actuation can be seen
in Figure 3.23.
Secondly, microprocessors and a high level computer were added to provide low
37
|
Colorado School of Mines
|
capture the data using a Java API and to transfer the data to a client over a TCP
socket. The client program receives the data from the host and constantly updates
variables for the main program (written in Python) which uses the signal strength
information. The flow of data from the Java server to python client subroutine to
python main program can be seen in Figure 3.28,
Figure 3.28: Software flowchart for collecting signal strength data.
3.10 Conclusions
In this chapter we explained the Bobcat remote control system and its limitations
within a subterranean mine. We then covered the Rajant Wi-Fi mesh network system
and performed a thorough reverse engineering analysis of the Bobcat remote control
system. Next we detailed the design for converting the Bobcat remote control system
to be Ethernet/Wi-Fi compatible. A brief summary of the AMR was also covered,
with an emphasis on the radio signal strength navigation. The following chapter
builds upon the design elements and covers the construction of the Bobcat remote
control to Ethernet/Wi-Fi conversion.
41
|
Colorado School of Mines
|
CHAPTER 4
CONSTRUCTION
This chapter is an extension of the previous chapter and shows the construction
stages of all the electronic equipment required to convert the Bobcat front-end loader
to a teleoperated vehicle. Along with the Bobcat, there is also a description of the
final implementation of the AMR and Rajant Wi-Fi network.
4.1 Transmitter Construction
The modified transmitter, which can be seen in Figure 4.1, is composed of the
original shell with all its switches, the two paddles, and the two joysticks. The
following additions were made:
1.) A microprocessor with Ethernet abilities
2.) Two PCBs that convert the joystick data to I2C
3.) A power distribution and level conversion PCB
4.) A diagnostic LCD
5.) A Linksys Wi-Fi access point
4.1.1 Microprocessor with Ethernet
The microprocessor used is a common Arduino Mega board available from Spark-
fun.com (DEV-09152) [14]. This board uses the Atmega 1280 microprocessor, it has
54 digital I/O, one I2C, four UARTs, and a 16 MHz crystal. An Ethernet add on
called the Arduino Ethernet Shield allows the Arduino Mega board to connect to the
Rajant Ethernet/Wi-Fi network. The Ethernet shield is based on the Wiznet W5100
Ethernet chip providing connectivity through TCP or UDP.
43
|
Colorado School of Mines
|
CHAPTER 5
EXPERIMENTATION AND ANALYSIS
In this chapter, the results of several tests of individual subsystems are shown to
establish a good idea ofhow the entire system will perform. These subsystemsinclude
the Wi-Fi ad hoc Rajant network, the teleoperated Bobcat front-end loader, and the
autonomous mobile radio (AMR). In addition to each subsystem, the entire system
was tested to provide the final proof-of-concept and show the best comparison with
a real scenario. To achieve a direct correlation to a real subterranean environment,
several of the experiments took place underground in the Edgar experimental mine
in Idaho Springs, CO.
5.1 Rajant Network
Several experiments were designed and implemented to characterize the perfor-
mance of the Rajant network along with the propagation properties of Wi-Fi based
EM waves in a subterranean environment. The Rajant Breadcrumbs can output sig-
nal strength in terms of signal-to-noise ratio (SNR). In a underground mine, we can
make the assumption that noise levels will be equivalent throughout the environment
and thus the SNR will be dependent upon the signal strength and not the noise level.
Therefore SNR has a direct correlation with radio signal strength (RSS).
5.1.1 Hallway Signal Strength vs. Distance
To get an indication of how signal strength drops according to distance in a tunnel
environment, signal measurements were taken in a hallway. In Figure 5.1 twenty RSS
data points were gathered, averaged, and compared to corresponding distance points.
53
|
Colorado School of Mines
|
subterranean environments to have a lower value than that of free space [5]. Our
other data shows extremely high attenuation around corners, up to twenty times
higher than line of sight placement. This is due to the high bending loss from non
specular reflection off jagged walls, matching independent research [5].
These results ultimately show that LOS placement of equipment is typically the
optimum solution, but it is still possible to gather some usable signal strength around
corners (which would be impossible with shorter wavelength communication systems
like light).
5.2 AMR Signal Strength Navigation
Two types of tests are used to display the ability to navigate by radio signal
strength. One test uses humans and the other uses autonomous robots. Both types
of experiments are controlled by a computer that analyzes the radio signal strength to
adjacent communication nodes and determines the direction to go for either a human
or autonomous robot (also known as the AMR).
5.2.1 Human only Experiments
Before the AMR was completely ready, we performed two experiments simulating
how an autonomous robot would navigate and position itself for optimal relaying of
data. This was done by using a human with a Rajant radio node and a laptop which
observed the radio signal strength (RSS) between the nodes in front of and behind
the human AMR. An algorithm on the computer made computations and displayed
directions to the human; directing him/her to ”Move Forward,” ”Move Backwards,”
or ”Stop.” In the first experiment there was one leader node far away from the base
station. The AMR (a human directed by the algorithm) moves to position itself at
an equal RSS away from both the leader and the base station, acting as a relay for
data. Figure 5.6 shows that the system did command the AMR (human) to move
58
|
Colorado School of Mines
|
5.2.2 AMR RSS Positioning Without Wall Following
Another experiment used a single semi-autonomous robotic AMR. Here, semi-
autonomous means that a human performed steering, as autonomous wall-following
was not functional at the time of the test. As in the previous human based experi-
ments described above, directional motion of the AMR was commanded by the RSS
control algorithm and this command was executed autonomously. The only role of
the human was to steer during AMR movements. The AMR navigated from the base
station to the optimal position, guided by only the signal strengths to the leader node
and base station (follower) node. The AMR would gather radio signal data for 30
seconds and then move to an estimated position. It would then repeat the gathering
of data for 30 seconds and reposition again if necessary. The AMR repeated the gath-
ering of data and moving five times during the experiment, until it ultimately reached
its goal of equal signal strength between the leader and base station (follower). Fig-
ure 5.8 shows the resulting signal strength plots and distance versus time. Figure 5.9
shows the corresponding locations within the mine. Clearly, the algorithm was effec-
tive in achieving a balanced RSS between the AMR and the base and between the
AMR and the leader.
Figure 5.8: Semi autonomous RSS positioning plot.
60
|
Colorado School of Mines
|
Figure 5.9: Corresponding semi autonomous RSS positioning map.
5.2.3 Completely Autonomous AMR RSS Positioning
On April 2, 2010 we completed a successful test using the autonomous AMR with
wall following implemented. A leader node was placed at a far position within the
mine to simulate the path of a leading teleoperated vehicle venturing too far away
from radio communications. An AMR was given the task to relocate and relay in-
formation from the leader node back to the base station. In the experiment, the
AMR successfully relocated itself to find the optimal location for relaying informa-
tion. Secondly, in the same experiment, after the AMR had successfully positioned
itself at the optimal location, the leader then moved even further away. This caused
the signals strengths to become imbalanced and the AMR automatically took action
and repositioned itself again, this time in the new optimal location for relaying infor-
mation. The signal strength data for this successful test is seen in Figure 5.10 and
the corresponding map is shown in Figure 5.11.
Figure 5.10 shows the leader and base-station SNRs that are relative to the AMR.
These two SNRs directly correspond to the received signal strengths from the leader
and base station to the AMR. As can be seen at the beginning of Figure 5.10 the SNR
to the leader is low, while the SNR to the base-station is high. This is because the
AMR starts its journey near the base-station and moves toward the leader. As time
61
|
Colorado School of Mines
|
of providing uninterrupted radio communications between a teleoperated vehicle in a
subterranean mine with its driver at the base station.
5.3 Bobcat Experiments
The Bobcat front-end loader was independently tested in three ways: several tests
for operability and safety, a control time delay test, and a subterranean control test
within the Edgar mine.
5.3.1 Teleoperated Control
In order to safely test the full functionality of the modified Bobcat remote control
system, the Bobcat was placed on jack stands, as seen in Figure 5.12, and had an
emergency off switch manned at all times. The remote control system was tested
in several stages. First was control over a wired Ethernet cable. Secondly, was
control using the line of sight but wireless controller. Lastly, the system was tested
with the wireless Rajant Breadcrumb network, with several stages of hopping and
camera based operation. Teleoperated control is seen through webcams, as seen in
Figure 5.13.
TheBobcatremotecontrolsystemwastestedforcontrolofindividualcomponents,
responsiveness of control, and the fail safe conditions to prevent accidents. The
following scenarios were tested to ensure the operation of fail safe programming and
safety of operation:
1.) Ethernet loss of connection.
2.) Moving during a Ethernet loss of connection.
3.) Moving during a loss of power.
4.) Mainboard to microprocessor loss of communication.
63
|
Colorado School of Mines
|
Additional delay from more wireless hops was analyzed and showed no appreciable
delay. Zero delay is the ultimate goal, but this is a reasonable time delay to deal with
for our purposes.
5.3.3 Bobcat Subterranean Navigation
The last test, before a complete system mock disaster scenario test, was to see
how the Bobcat and Rajant network would perform in the Army section of the Edgar
mine. The Bobcat was outfitted with extra lights, so that the Ethernet-based camera
could see, and was sent into the mine, as depicted in Figure 5.15.
Figure 5.15: Bobcat underground.
The base station was set up to display the camera feed and to log data for the
experiment. The user then actuates the Bobcat controller from the base station,
while viewing the camera feed from a laptop display, as seen in Figure 5.16.
From the user’s perspective, one camera provides the viewpoint that an operator
might see from inside the cab of the Bobcat. Video quality was sacrificed to provide
the base station user a higher frame rate and better responsiveness. A sample video
feedback single frame is shown in Figure 5.17.
66
|
Colorado School of Mines
|
The point of the AMR was to create a seamless wireless network for the teleoperated
front-end loader and it performed its task well. The Bobcat traveled so far into the
mine that a single link would not have been sufficient for control. Thus, the AMR
was a necessary part of the system.
5.6 Conclusions
In this chapter the radio wave propagation characteristics, described in Chapter
2, are verified with several tests using the subterranean tunnels at the Edgar Ex-
perimental Mine in Idaho Springs, Colorado. Secondly, several tests were performed
on individual elements of the complete search and rescue caravan. The Bobcat was
tested first to ensure hardware and software operability along with safety procedures.
The Rajant network was tested to verify the ability to capture signal strength mea-
surements and the feasibility of navigation by those signals. The AMR was tested for
autonomous behavior in navigating obstacles and for following signal strength mea-
surements. Lastly all the elements came together for a complete system test where
the Bobcat front-end loader was teleoperated into the mine with an AMR following
it using only its sensors. The Bobcat was seamlessly controlled far into the mine, at
a point where a single direct wireless signal would not have been sufficient. Thus the
experiment worked with the autonomous vehicle acting as a relay to maintain the
wireless connection to the teleoperated Bobcat.
73
|
Colorado School of Mines
|
CHAPTER 6
CONCLUSIONS
One goal of this thesis was to show that wireless teleoperated control is possible
within an underground mine. However, in the absence of a fixed infrastructure it is
only made possible by utilizing relaying equipment with the ability to reposition. The
experiments from Chapter 5 have shown that a search and rescue robot could enter
a subterranean mine with its own independent communications equipment and still
communicate to the outside via several wireless relays atop autonomous vehicles. In
ourexperimentswesuccessfullyteleoperatedaBobcatfront-endloaderatamaximum
of about 350 ft (106 m) in a mine section containing a bend while utilizing a wireless
relay atop an autonomous vehicle. The autonomous vehicle positioned itself at the
optimal location for relaying signals to allow the best quality video feedback and
wireless range.
6.1 Recommendations for Future Work
There are a few things that would be beneficial for search and rescue robots in
subterranean mines. Ethernet-based air quality monitors would provide life-saving
information to rescue workers. A UHF radio-to-Ethernet interface for communicating
to existing leaky feeder communication lines would allow rescuers to possibly reestab-
lish communications to isolated sections of the mine. Lastly, the teleoperated vehicle
would benefit from the addition of better cameras and external sensors to help the
user with controlling the vehicle or by adding semi-autonomy.
6.2 Implications of Research
The implications of this research are two-fold. First, creation of an Ethernet
controlled teleoperated vehicle has almost limitless potential with existing global net-
75
|
Colorado School of Mines
|
APPENDIX A - A
The appendix provides references for reconstructing the project from hardware to
software.
A.1 Circuit Diagrams
The circuit diagrams here are for reference purposes, the files needed to recreate
these circuit boards are on the included CD.
A.1.1 RX
5 4 3 2 1
JJ1144
3.3V
12
PPoowweerr SSwwiittcchh
5V U 3U 3VV55 RReegg U 5U 5VV66 RReegg U 9U 9VV77 RReegg
D 1234567121234567
RR JJJ IJ
IJJ
gg
116622
iigg
1234567121234567
nnhh iitttt iioo SS nnww &&iitt cc EEhh ssee ttooss pp 12 123412RR LLJJ JJJJ ee44 7755iigg12 123412 fftthh PPtt aaPP ddaa dddd lldd eellee 12345678I2 VVVVVGVVoorDr ooNe eC uu uuUU D DDDf ftt ttH LAB CD11
AA
CC66I LD 55O DS 77SV 33A D ACAAAA d IIO CA PPd123 L WWC ((1 11 91111 DD3 12 0456
AACC))
RRRR11 Q2Q2NN1166665599
1234J IJ I2211 CC55
,, IICC,, OOCC R PR POO4466 TT RRRR4455 CCCC22 AArr1234CCCC dd..11 .. JJ &&11
1234
33 MMPPoo
aaJJ
ww
iinn11
ee
BB66
rr ((ss PPuu oopp ww ((11 ee22 rr))VV))
D
LLeefftt SSwwiittcchheess JJooyyssttiicckkss
C C
Level Conversion
UU22
11111111121234567890 LLeeBBBBBBBBOV vv87654321
CE eeC
ll
ccG ooN
N
nnAAAAAAAA vvD C12345678 1 1234567890
RRRR55 RRRR44 RRRR33 RRRR22 RRRR66
MMJJ
ii1234567
ss11 cc11
..
SSNRLHfAAl
wwoP
DDCP
oa iir 22 tttn cc24
hheess
B
1111111112 11111111121234567890 1234567890U LU
L U LU Lee
eeBBBBBBBBOV BBBBBBBBOV33
44vv
vv87654321
C 87654321 CE Eee eeC Cll
ll
CC
CCG
Goo
ooN
NN
NAAAAAAAA
AAAAAAAAnn
nnD
DC
C12345678
12345678vv
vv
1
11
123456789 234567890
0
RRRR RRRRRRRR88
2211 3344 RRRR
RRRRRRRR
11
2277
33
22
RRRRRRRR 1111 2200
RRRRRRRR
RRRR
2211
11
0011
66 RRRR1177 RRRR1188 RRRR1199
RRRRRRRR 1199
55
JJ
1234567
123456711
J LJ
LRR00
99 eeiigg fftthh
AAAAAAA
AAAAAAA
SSttDDDDDDD
DDDDDDD
wwSS2222223
3333333ww
iitt6835790
2461357
cciitt hhcc eehh sseess B
A A
RRRR4444 RRRR4433 RRRR4422 RRRR4411 RRRR4400 RRRR3399 RRRR3388 RRRR3377 RRRR3366 RRRR3355 RRRR3344 RRRR3333 RRRR3322 RRRR3311 RRRR2299 RRRR2288 RRRR2277 RRRR2266 RRRR2255 RRRR2244
D LD LEE22 DD11 D LD LEE22 DD00 D LD LEE11 DD99 D LD LEE11 DD88 D LD LEE11 DD77 D LD LEE11 DD66 D LD LEE11 DD55 D LD LEE11 DD44 D LD LEE11 DD33 D LD LEE11 DD22 D LD LEE11 DD11 D LD LEE11 DD00 D LD LEE99 DD D LD LEE88 DD D LD LEE66 DD D LD LEE55 DD D LD LEE44 DD D LD LEE33 DD D LD LEE22 DD D LD LEE11 DD TTTiiitttllleee MMMiiicccrrrooocccooonnntttrrrooolllllleeerrr IIInnnttteeerrrfffaaaccceee BBBoooaaarrrddd
SSSiiizzzCCCeeeuuussstttooommm111DDDooocccuuummmeeennnttt NNNuuummmbbbeeerrr RRReee111vvv
5 4 3 2 DDDaaattteee::: SSSaaatttuuurrrdddaaayyy,,, AAAuuuggguuusssttt 111444,,, 222000111000 SSShhh1eeeeeettt 111 ooofff 111
Figure A.1: RX Interface Board
79
DTUNN OGI
1 32
DTUNN OGI
1 32
tjud nOA I
1 32
|
Colorado School of Mines
|
ABSTRACT
The use of passive systems for the treatment of acid mine drainage has been
extensive over the past decade. However, the basic understanding of how different
organic material characteristics affect a passive treatment system’s (PTS) ability to
remove toxic metals and actively reduce sulfate has been absent from the literature. This
investigation performed elemental (C and N), physical (moisture and organic fractions),
nutritional (e.g., neutral detergent and water soluble), sorption and leaching analyses on
twelve organic materials commonly used in PTS. The organic materials were grouped
into three categories: agricultural and industrial products (alfalfa pellets, sugar beet pulp
pellets, brewery waste, corncobs, and walnut hulls), woods (maple, oak, pine, poplar, and
walnut) and inoculum sources (dairy manure and wetland sediment). Metal sorption of
manganese to the twelve organic materials and inorganic leaching were examined in
batch experiments. The effects of organic substrate characteristics on metal removal and
sulfate reduction rate (SRR) were evaluated for pine, oak, alfalfa and corncobs in column
experiments receiving synthetic mine water.
The highest overall sulfate reduction rate (SRR) as well as the highest sustained
SRR for 70 days of column operation were observed for the column (alfalfa) with the
highest neutral detergent soluble compounds (NDSC), water soluble fraction (WSF) and
ethanol soluble fraction (ESF) with r2 = 0.95-0.96 for a linear correlation. NDSC, WSF
and EST appear to be good measures of the relative degradation rates of the organic
material. Organic carbon and total organic content had inverse relationship to the
observed SRR and thus were poor measures of the relative degradation rates of organic
material.
A correlation exists between manganese batch sorption studies and the initial
removal of manganese in the column tests. However, the batch tests severely
|
Colorado School of Mines
|
ACKNOWLEDGEMENTS
I’d especially like to thank my advisor Dr. Linda Figueroa whose lab experience,
problem solving, bouncing ideas, and encouragement. Diane Ahamann for her help in
understanding the relationships between the different microbial communities present in
PTS. Thomas Wildeman for always having an open door, and listened patiently to the
latest problems and shared his ideas about what could be happening in the columns and
sharing all of his many years of experience with us. The EPA Rocky Mountain Regional
Hazardous Substance Research Center Region 8 Section 3 provided the funding for this
project.
I would like to thank the following people who helped me throughout my time at
CSM. Barb Harvey for all of her help in the lab, especially with the ICP-MS machine and
for continuously fixing such a temperamental beast. Jim Horan for never running the
other way when he saw Miranda and I coming with a question or a request, and for
always taking the time out of his busy schedule to hunt for an obscure piece of lab
equipment. Thanks to Marie, Stephanie, and Pascale for their tireless help in processing
countless samples and very positive out look on life. John Albert whose knowledge,
successes, and continuous assurance inspired us all. Miranda who shared every success
and failure throughout this adventure in higher education. Who was a never tiring
sounding board for ideas, problems, frustrations, and ultimately their solutions. Couldn’t
have hoped for a better partner
To the support from my family, I wouldn’t be the person I am today without their
guidance, support, and nudges out the door. My amazing and lovely wife who showed me
that you can make your dreams come true with hard work and determination. For
supporting and understanding me and my dreams. Moreover, for being the best
|
Colorado School of Mines
|
1
CHAPTER 1
INTRODUCTION
Acid mine drainage (AMD) forms through the chemical and biological oxidation
of pyrite (FeSz) and metal sulfide minerals found in mine waste rock and tailings. This
reaction results in low-quality water characterized by low pH (2-4) and elevated
concentrations of sulfate and other dissolved metals. On United States Forest Service
lands there are between 20,000 and 50,000 mines generating acidic drainage (UDSA
1993). An estimated 500 billion gallons of AMD are produced annually, affecting around
16000 to 17000 km of streams (Lyew et al., 1994). The continuous pollution of streams
has been largely due to the impracticalities of conventional treatment technologies in
remote mountain areas where these numerous mines are located.
Conventional treatment includes the addition of lime (Ca(OH)2 and CaO) thus
raising the pH and precipitating the metals as hydroxides. This process produces a waste
sludge of gypsum and metals hydroxides that is highly contaminated with heavy metals
and must first be dewatered before disposal in a landfill (Lyew et al., 1994).
Conventional AMD treatment typically carries high costs, involves continuous
maintenance, and requires accessibility and infrastructure not currently available in
remote areas. As a result, passive treatment systems such as constructed wetlands and
permeable reactive barriers have received a great deal of attention for remote sites
(August et al., 2002).
Metals can be removed from AMD in the anaerobic biozone of a passive
treatment system (PTS). Several types of PTS employ an anaerobic biozone to remove
metals including constructed wetlands, bioreactors, and permeable reactive barriers. PTS
are a low cost alternative because they require no power and no chemicals after
|
Colorado School of Mines
|
2
construction (Gusek and Wildeman 2002). The lack of continuous upkeep allows PTS to
be implemented in areas where conventional treatments are not financially practical.
Wieder et al., (1990) reported, that the effectiveness of PTS have been both extremely
variable and to date generally unpredictable.
The key metal removal mechanisms of anaerobic biozones are sorption and
precipitation. Sorption of metals (including manganese) is limited to the availability and
affinities of sorption sites, and is therefore considered a short-term mechanism.
Anaerobic precipitation of metals occurs when a metal binds with sulfide, and produces
an insoluble compound. The sulfide in an anaerobic biozone is produced by sulfate
reducing bacteria (SRB). An adequate production of sulfide can only be maintained if the
proper quantity and quality of substrate is available to the SRB community.
1.1 Hypothesis
• The sorption capacity of organic substrates used in anaerobic biozones can be
selected to provide for long-term removal of manganese.
• Operationally defined measures of organic composition are correlated to sulfate
reduction rates (SRR) in anaerobic biocolumns.
1.2 Objectives
• Evaluate the utility of batch sorption experiments to estimate maximum sorption
capacity and theoretical breakthrough times in anaerobic biocolumns.
• Use relatively simple operationally defined methods to characterize organic
compositions of organic materials used in anaerobic biozones
• Correlate the above organic fractions (i.e. percent organic, carbon, protein, simple
sugar, etc.) to the rate of sulfate reduction in anaerobic biocolumns.
|
Colorado School of Mines
|
3
CHAPTER 2
BACKGROUND
2.1 History and Literature Review of PTS Designed to Treat AMD
In the late 1960s, researchers from Ohio State University documented that AMD
flowing into a wood bog showed reductions in metals and sulfate (Tuttle et ah, 1969). In
1978, researchers at Wright State University documented that a natural Sphagnum bog
showed the ability to treat waters with low pH and high metal concentrations (Huntsman
et al., 1978). Similar observations were noted for water being treated by the Tub Run Bog
by a group at West Virginia University (Lang et al., 1982). These studies sparked the
field of AMD treatment through use of mechanisms found in natural systems.
The concept behind passive treatment is to allow the naturally occurring chemical
and biological reactions that aid in AMD treatment to occur in the controlled
environment of the treatment system. Passive treatment conceptually offers many
advantages over conventional active treatment systems. The use of chemical addition and
energy consuming treatment processes are virtually eliminated with passive treatment
systems, as are the operation and maintenance requirements.
Several examples of full scale AMD PTS can be found in the literature. In 1993,
the 16 acres Fabius Coal Mine “Hard Rock” Wetland was constructed capable of treating
over 600 gallons per minute of AMD and is currently operating as designed with
reportedly little maintenance. The West Fork Sulfate Reducing Bioreactor System was
designed and implemented in 1995. This 4.5 acre, multi-celled system can treat 1,200
gpm of zinc and lead laden mine water down to discharge limits of 30 ppb lead (Gusek
and Wildeman 2002). One year after installation, the Nickel Rim reactive barrier’s
effluent concentrations of Fe decreased from 740 - 1000 to < 1 - 91 mg/L and alkalinity
|
Colorado School of Mines
|
4
values increased by a factor of 12 (from 60 - 220 to 850 - 2700 mg/L CaCOs) (Blowes et
al., 2000). As the above examples illustrate, PTS offer a legitimate alternative to costly
conventional AMD treatment.
2.2 Dominant Metal Removal Mechanisms Found in PTS
PTS are able to removal both the acidity and metals from AMD, through a series
of ecological and geochemical reactions. These include processes that directly remove
contaminants as well as chemical reactions that cause the precipitation of metal
compounds (August et al., 2002). The direct removal of metals consists of physical (i.e.
filtration of suspended and colloidal metals), biological (i.e. uptake by plants and
microorganisms), and sorption processes. Chemical removal mechanisms include the
formation and precipitation of metal oxides and sulfides.
2.2.1 Bio/phvto-remediation
During the early investigations in the PTS, both bioremediation and
phytoremediation were matters of extensive study for the treatment of AMD.
2.2.1.1 Bioremediation
Metals found in AMD are essential micronutrients to algae. Algae have been
commonly found in AMD with high concentrations of iron and manganese. The
bioaccumulation of manganese is particularly of interest due to its persistence in PTS.
Kepler (1988) reported that algae have been shown to accumulate up to 56,000 mg Mn /
|
Colorado School of Mines
|
5
kg plant tissue (dry wt). However, later experiments by Clayton et al., (1998) and
Richardson et ah, (1988) show that algae predominantly aid in the removal of manganese
through precipitation. Richardson et al., reported that the much of the manganese was
being removed from solution extracellularly as manganese precipitates. Clayton et al:,
concurred with Richardson et al., and stated that the algae (via photosynthesis and the
resulting increase in pH) only facilitate the formation and precipitation of manganese
oxides.
2.2.1.2 Phvtoremediation
Spratt and Wieder (1988) investigated the ability of Sphagnum plants to
accumulate metals in an AMD wetland. Spratt and Wieder state that while Sphagnum
plants can achieve an uptake value of 29 g Fe/m2/yr, these concentrations were ultimately
fatal. In a natural wetland receiving AMD, August et al., (2002) report that the total mass
of manganese accumulated by wetland vegetation (1-3 mg Mn/g) was six and 11 times
higher than that of iron and zinc respectively. However, this metal uptake by wetland
vegetation was not a permanent process. During the winter months, vegetation would die
and release the metals back into the water column. August et al., concluded similarly to
Sencindiver and Bhumbla (1988) that while wetland vegetation possess the ability to
uptake metals, the major mechanism for metal removal is their ability to stimulate
microbial processes.
2.2.2 Sorption
Sorption is a broadly defined term for the transfer of ions from the solution phase
to the solid phase. The partitioning of heavy metals between solid and aqueous phases is
|
Colorado School of Mines
|
6
controlled by properties such as surface area, surface charge (induced by the formation of
organic coatings on the surface), pH, ionic strength, and concentration of complexing
ligands (Petrovic et al., 1999). Adsorption refers to the attachment of ions to soil
particles, by either cation exchange or chemisorption. Chemisorption represents a strong
and more permanent form of bonding than cation exchange, and can form where
electrostatic interactions oppose adsorption. Chemisorption of a species depends more on
specific chemical interactions between it and a surface than surface charge attraction. A
number of metals and organic compounds can be immobilized in the soil via
chemisorption with clays, iron (Fe) and aluminum (Al) oxides, and organic matter
(DeBusk 1999, Benjamin 2002).
Cation exchange involves the physical attachment of cations to the surfaces of
clay and organic matter particles in the soil. This is a weaker attachment than chemical
bonding, and therefore a less permanent removal mechanism. The attractive forces
causing physical adsorption are primarily via electrostatic interactions. The capacity of
soils for retention of cations, expressed as cation exchange capacity (CEC), generally
increases with increasing clay and organic matter content (DeBusk 1999, Benjamin
2002).
An adsorption isotherm is any equation that relates the amount of material sorbed
at the surface to that in solution in systems that have reached equilibrium. Isotherm
models can be fit to the experimental data and are used to predict sorption values over a
large range of concentrations. While numerous equations for isotherms are available in
the literature, the most common adsorption isotherms for application to engineered
treatment systems are the Langmuir and Freundlich (Richards and Reynolds 1995). The
popularity of the Langmuir and Freundlich isotherms is due in large part by their ability
to accurately describe a wide variety of sorption data. These isotherms have simple
equations and their adjustable parameters are easily estimated compared to the three or
more found in more complex isotherms (Kinniburgh 1986).
The Freundlich isotherm uses a specific application of the power function,
|
Colorado School of Mines
|
7
f(x) - axk, to model the distribution of chemicals between solution and solid phases. One
form of the Freundlich equation is described by Eq. 2.1 :
Eq. 2.1
Where qe is actual mg sorbed/ gm solid, Ce = equilibrium concentration in mg/L, and the
K and n are empirically derived constants (Reynolds and Richards 1995). The Freundlich
isotherm fits experimental sorption data well, but does not provide information about the
actual sorption mechanism (Wolt 1994).
The Langmuir isotherm was originally developed to describe the adsorption of
gases by solids. It assumes that there is a limited area available for adsorption, the
adsorbed solute material is only one molecule in thickness, adsorption is reversible, and
an equilibrium condition is achieved. Once an equilibrium condition is achieved, it is
assumed that the rate of adsorption is equal to the rate of desorption. The Langmuir
isotherm is described by Eq. 2.2 (where qe = actual mg sorbed/ g solid, Ce = equilibrium
concentration in mg/L, a = maximum mg sorbed/ g solid, and K = experimental constants
in L/mg) (Reynolds and Richards 1995).
aKCe
Eq. 2.2
The maximum sorption capacity of the solid phase can be estimated by fitting a
Langmuir isotherm to the experimental data and then solving for “a”.
|
Colorado School of Mines
|
8
2.2.3 Precipitation
The dominant metal removal mechanisms in PTS are the oxidation/hydrolysis of
metals and metal sulfide precipitation.
2.2.3.1 Oxidation and Hydrolysis Precipitation
Iron and manganese can be removed as precipitates from metal laden waters via
oxidation and hydrolysis reactions. Equations 2.3 and 2.4 show ferrous iron becoming
oxidized and then the ferric iron undergoing hydrolysis and precipitating.
4 Fe2+ + 02 + 4 H+ -* 4 Fe3+ + 2 H20 Eq. 2.3
4 Fe3+ + 12 H20 — 4 Fe(OH)3 J + 12 H+ Eq. 2.4
The extent of metal removal depends on dissolved metal concentrations, dissolved
oxygen content, pH and net alkalinity of the mine water, the presence of active microbial
biomass, and detention time of the water in the wetland. The pH and net acidity/alkalinity
of the water are particularly important because pH influences both the solubility of metal
hydroxide precipitates and the kinetics of metal oxidation and hydrolysis. As shown in
Equation 2.4, metal hydrolysis produces considerable amounts of acidity. Alkalinity
present in water can buffer the pH and allows metal precipitation to continue. Inorganic
oxidation reaction rates decrease a hundred-fold with each unit drop in pH (Skousen
1999).
Manganese oxidation occurs more slowly than Fe and is sensitive to the presence
of Fe2+, which will prevent or reverse Mn oxidation. Consequently, in aerobic net
alkaline water, Fe and Mn precipitate sequentially (not simultaneously) with the practical
|
Colorado School of Mines
|
9
result that Mn precipitation occurs only after all Fe is precipitated (Skousen 1999).
Abiotic manganese oxidation occurs at pH > 8 and therefore, Mn oxide formation is
essentially controlled by pH. However, algae have been reported to facilitate abiotic Mn
oxidiation by maintaining a high pH and oxygen environment (Clayton et al., 1998,
Richardson et al., 1988).
2.2.3.2 Metal Sulfide Precipitation
The reduction of sulfate to sulfide by sulfate reducing bacteria (SRB) indirectly
precipitates metals as metal sulfides in the reducing environment found in anaerobic
microbial PTS. Research has shown that SRB obtain energy and nutrients by oxidizing
low molecular weight organic carbon compounds (i.e. lactate) and using sulfate as an
external electron acceptor. SRB reduce sulfate to sulfide in a metabolic pathway known
as the dissimilatory reduction of sulfate. Meaning, SRB have the ability to reduce sulfate
without the assimilation of the sulfur into cellular material. Equation 2.5 shows the
products of SRB converting lactate into acetate. This reaction produces hydrogen sulfide
and bicarbonate. Bicarbonate will help minimize the potential for acid generation and
hydrogen sulfide readily complexes with metals to form metal sulfides (See Eq. 2.6). The
majority of metal sulfides are insoluble in water (Stumm and Morgan 1996).
2CH3CHOHCOO' + S042' 2CH3COO' + 2HCO3 + H2S Eq. 2.5
Me2+(aq) + H2S(aq) MeS(s) + 2H+ Eq. 2.6
-*
The activity of SRB in the PTS is approximated by monitoring the sulfate
reduction rate (SRR). SRR is expressed as ?moles of sulfate/m3/day and its calculation is
|
Colorado School of Mines
|
10
presented in equation 2.7 (where Sm and S0ut = sulfate concentration of the influent and
effluent respectively, Q = flow, and Vp = packed volume).
SRR = Eq. 2.7
The initial studies investigating the use of SRB in anaerobic PTS concluded that
the identification of inexpensive organic carbon sources, that produce high rates of
sulfate reduction, are essential for making anaerobic AMD PTS cost effective.
Subsequent research discovered that mixtures of various organic materials produced the
highest SRR. Gilbert et al., (1999) reported that the highest SRR was produced from a
mixture of 25% dairy manure, 10% sawdust, and 10% alfalfa. This mixture is similar to
the 20% poultry manure, 30% leaf compost, 3% wood chips mixture used by Cocos et al.,
(2002), and the mixture of 10 % sewage sludge, 10% sheep manure, 5 % leaf compost,
12.5% wood chips, and 12.5 % sawdust reported by Waybrant et al., (1998).
Attempts at understanding the reasons behind why some mixtures produced
higher SRR than others were made by a number of researchers. Gilbert et al., (1999) and
Pinto et al., (2001) tried to link an easily measured attribute (pH) to the actual nutritional
characteristics such as organic acid content and C:N ratios. Kuyuck et al., (1994) reported
a mixture of lactic acid, ammonium sulfate, and potassium biphosphate salts possessed a
C:N:P ratio of 110:7:1 and promoted the highest sulfate reducing activity. A mixture of
various wastes products (wood pulp, manure, and brew dried grain) was then produced
that equaled this C:N:P ratio and yielded a SRR of 0.3 moles S/m3/day. Waybrant et al.,
(1998) also suggested that SRR may be depended upon the nitrogen and phosphorus
contents found within the organic substrates. While these analyses are a step in the right
direction, they failed to recognize that sustained rates of sulfate reduction are dependent
on SRB as well as the other microbial communities found within a PTS.
|
Colorado School of Mines
|
11
What was unknown at the time was that once all these simple organics are
consumed, SRB become reliant on other microbial communities to degrade complex
carbohydrates and supply the simple organic fractions. Without materials comprised of
different structurally complex organics, the dependent relationships between the bacterial
communities will collapse resulting in system failure. There have been some references
in the literature expressing a need for long-term performance studies with respect to SRB
and the other communities found in PTS (Cocos et al., 2002, Eger and Wagner 1995,
Blowes et al., 2000). Waybrant et al., (1998) suggests that a mixture of organic sources
could potentially last longer than treatment systems containing just manure or other
individual substrates. Eger and Wagner demonstrate that by the end of a four-year study
all but one of mixtures tested stopped actively removing sulfate. They report that
although the calculations based on the use of total carbon indicate the systems could last
for over 20 years, the carbon that can be effectively used within a PTS may only be a
small percentage of the total amount. The results of these studies show that there is a
need for research investigating the balance between carbon sources, SRB, and other
microbial communities present in PTS.
2.3 Anaerobic Microbial Ecology and Carbohydrate Degradation
The literature associated with the treatment of AMD shows a lack of research
examining the interactions of SRB and other microbial communities. In anaerobic
environments, SRB will consume readily available organic acids and simple sugars and at
the same time reduce large amounts of sulfate. As the scarcity of these foods sources
increases, the SRB become more reliant on other communities of bacteria to degrade
monomers and amino acids into simpler compounds.
Figure 2.1 illustrates the many forms of carbohydrates found in plants. The
diagram ranges from structurally simple to complex carbohydrates (left to right) (Hall
|
Colorado School of Mines
|
14
Hemicellulose (other wise known as xylan) is the next most abundant and widely
distributed carbohydrate after cellulose. The hemicelluloses have a polymerization
number of only 30-100 units and are not structurally related nor do they contain the same
building blocks as cellulose. They are more rapidly degraded, and by a larger number of
microorganisms than cellulose. When enzymatically degraded by xylanase, hemicellulose
tends to produce oligomers containing (i.e. arabinose, glucose, galactose) components.
(Schlegel 1996)
Lignin (by mass) is quantitatively the most dominant component of plants after
cellulose and hemicellulose. Lignin is the slowest component in plants to degrade and is
therefore a major source of slowly decomposing organic substances. The formation of a
three dimensional complex structure between lignin, cellulose, and hemicellulose renders
the latter less accessible to microorganisms due to physical barriers (Pareek et al., 1998).
Structurally, lignin is non-uniform and very complex, with basic units generally
derivatives of phenyl propane (Schlegel 1996). Although the degradation of lignin has
been extensively studied, under anaerobic conditions lignin had been thought to be
refractory. Studies by Benner et al., (1984) and Pareek et al., (1998) proved that the
degradation of lignin did occur under anaerobic conditions. Pareek et al., (1998) even
reported that the formation of depolymerization products (i.e. glucose and xylose) of
lignin degradation occur more rapidly under sulfate reducing conditions than
methanogenic.
Proteins are long chains of amino acids that are enzymatically degraded by
proteolytic bacteria. These bacteria excrete proteases to cleave the peptide bonds holding
the amino acids together. The proteases are grouped based on their active pH ranges.
Alkaline proteases (active in the pH range of 8-11) are unspecific and cleave all peptide
bonds. Acid proteases tend to cleave only peptide bonds between certain amino acids.
The resulting amino acid and low-molecular weight peptides are actively taken up and
utilized for the growth of many microorganisms. Amino acids are typically degraded
|
Colorado School of Mines
|
15
along two pathways (oxidation and deamination) to produce a multitude of organic acids
(Gottschalk 1986).
Phosphorolysis, hydrolysis, and transglycosylation are the three pathways used by
cellulolytic bacteria for the degradation cellobios and oligomers. Phosphorolases are used
to convert polysaccharides to glucose-1 -phosphate intracellularly. Extracellularly,
Amylases are used to break apart macromolecules during hydrolysis. Transglycosylation
results in starches being converted into small rings of glucose (Schlegel 1996).
Once the complex macromolecules (proteins, lignins, celluloses) have been
broken down into monomers, in the processes described above, fermenters continue the
degradative reactions. Fermentative bacteria have the ability to enzymatically transform
monomers into a wide variety of organic acids, alcohols, and gases. The SRB can then
obtain energy and nutrients by oxidizing these low molecular weight organic carbon
compounds (lactate, acetate) by using sulfate as an external electron acceptor (Herbert et
al., 1998). Therefore, in order to obtain an active and healthy SRB community a balance
must be maintained between all of the processes and microorganisms illustrated in Figure
2.2.
2.4 Long-term Performance of PTS
PTS have a finite capacity for the treatment of AMD. A PTS’s longevity is related
to its initial chemical and physical properties as well as changes affecting these properties
though the life of the system. Physical properties of a PTS include the total mass of
reactive material within the barrier, the design/ layout of the treatment system, and its
flow capacities, porosity and permeability. Chemical characteristics include the desired
removal mechanism, rates of reactions, and the sorption and composition properties of
the reactive materials.
|
Colorado School of Mines
|
16
Blowes et al., (2000) reports a number of potential processes (both chemical and
physical) that can lead to a decrease in system’s rate of reaction or longevity. Physical
processes such as clogging of pores could create preferential flow channels, which would
decrease the residence time as well as the amount of reactive surface area. Chemical
reactions are also limited by precipitating secondary minerals decreasing porosity,
causing preferential flow, and reducing reactive surface area.
The consumption and depletion of the reactive material will also affect the
reaction rate in PTS over time. Many systems contain sufficient reactive material to
theoretically remove target contaminants for hundreds of years. These calculations are
based upon total carbon within the reactive mixture and its consumption by SRB.
However, the reactive material within the system will be consumed by microorganisms
other than SRB resulting in a decrease in longevity.
The results found in the literature of full scale PTS has been mixed. Some PTS
have been successfully removing metals from AMD for multiple years (Gusek and
Wildeman 2002) while others have encountered problems. After only 6 weeks of full-
scale operation, the loading capacity of the West Fork Mine anaerobic PTS was half of its
original design. The H^S gas produced by active SRB was being retained within the
treatment cells and preventing full design flow (Gusek et al., 1998). Three years after the
Nickel Rim barrier was installed the rate of iron removal by sulfide precipitation has
declined 45% from an initial rate of 2100 to 1000 mg L*1 year-1. The decrease in removal
rates has been potentially linked to preferential flow reducing the reactive surface area
(Blowes et al., 2000). Both studies suggest that these problems may have been identified
if proper treatability studies had been performed.
|
Colorado School of Mines
|
18
3.1.1.2 Percentage of Organic Material
A 1” ceramic crucible was washed and placed into the muffle furnace set at 550° C
for 1 hr and allowed to cool in a desiccator for another hour. The crucibles’ weights were
then recorded. 0.300 g of the dry sample was then placed into these pre-weighted
crucibles and burned over a Bunsen burner by setting it at around a 45 angle on a burning
triangle and stand. Once all flames and smoke stopped, the crucible was placed into the
muffle furnace set at 550° C for a minimum of 8 hours. The pre-muffle burning step
prevents any ash or organic material from escaping the crucible while in the muffle
furnace. After 8 hrs in the furnace, the crucibles were allowed to cool for 5 min before
being placed into the desiccator for 1 hour. This step prevents the plastic trays of the
desiccator from melting.
3.1.1.3 Total Carbon and Nitrogen / Protein Content
The total carbon and nitrogen percent of each organic material was performed
using an Exeter Analytical, Inc. C, N, H CE-440 Elemental Analyzer at the USGS
Boulder facility (contact Deborah Ropert). 1 gram samples were dried overnight in a 105°
C drying oven and stored in a 15 ml falcon centrifuge tubes until analyzed. The samples
were weighed (3000 pg ± 1000 pg) and run in accordance to the USGS’s standard
operating procedure.
|
Colorado School of Mines
|
19
3.1.2 Neutral Detergent-Soluble Carbohydrates Nutritional Relevance and Analysis
(NDSCNRA3
The NDSCNRA uses differences in solubility to separate and quantify
carbohydrate fractions found in organic materials. The 80% ethanol extraction is able to
solublize low molecular weight mono- and oligosaccharides, organic acids, and a small
amount of crude protein found in organic material. The saccharide content of the 80%
ethanol solution is measured by a colorimetric assay and the organic acid fraction can
then be estimated by difference. Starch can be measured in the ethanol insoluble residue
by an enzymatic/colorimetric assay. The neutral detergent soluble fiber (NDSF) (fmctans,
pectic substances, (3-glucans, and other non-starch polysaccharides) is estimated by
subtracting the ethanol insoluble and starch fraction from the neutral detergent soluble
carbohydrates (NDSC).
A neutral detergent/ a-amylase extraction will solublize all carbohydrate fractions
except hemicellulose, cellulose, and lignin. By performing an acid hydrolysis on the
neutral detergent insoluble fraction, one can fractionate the lignin from the hemicellulose
and cellulose.
3.1.2.1 80% Ethanol Soluble Fraction Analysis
0.2 ± 0.01 g of a dried sample was weighed and placed into a 50 ml culture tube
with a Teflon screw cap. 40 ml of 80% ethanol was added, the tube capped, and shaken
on a shaker for 24 hrs at room temperature. The entire content of the culture tube was
then filtered under a vacuum through a Whatman 54 filter. The culture tube and cap were
washed with 80% ethanol and the rinse was poured through the filter. The filter residue
was rinsed twice with 20 ml of 80% ethanol. The entire volume of extraction solution
used was measured in a graduated cylinder and stored in a sealed container for later
|
Colorado School of Mines
|
20
analysis. The filtering apparatus was then placed back on the vacuum flask and the filter
residue was rinsed twice with 20 ml of acetone. The used acetone solution can be
discarded. The filtering apparatus was carefully taken apart and the filter plus the residue
was placed into a pre-weighed 2” aluminum tray. The sample plus tray was then dried at
105° C overnight and then cooled in a desiccator for at least 1 hr before weighing. The
filter residue was then carefully scraped from the filter paper into the same pre-weighed
tray and weighed.
3.1.2.2 Total Ethanol Soluble Carbohydrates (Phenol - Sulfuric Acid Test)
100 ml of lOg/L (10 mg/mL) sucrose stock solution was prepared. 10 ml of 80%
ethanol was then placed into sucrose standards of 0, 25, 50, 75, and 100 mg/L. A standard
curve was run with each group of samples. The standard stock solution and individual
standard solutions were prepared fresh on the day they were to be used. All samples were
prepared in triplicate and compared to the sucrose standard curve.
0.5 ml of each sample was added to an acid washed 10 ml Hach vial. 0.5 ml of the
5% phenol solution was added, followed by the addition of 2.5 ml of the concentrated
H2SO4. The vial was capped and vortexed for 5 seconds. After all samples were prepared,
they were vortex again and allow to sit for 1 hr at room temperature. At a wavelength of
490 nm, the absorbance of all samples were read and recorded.
|
Colorado School of Mines
|
21
3.1.2.3 Starch Gelatinization and Hydrolysis Method
In this procedure, the starch in the sample was hydrolyzed into glucose and the
glucose content was measured. Therefore, all of the glucose and simple sugar fractions
must first be removed from the organic material through the 80% ethanol extraction.
The reagents required for this method are: Heat-stable -amylase
(Sigma-Aldrich Company, St. Louis, MO), 0.1 M sodium acetate buffer (pH ~ 4.5), and
Amyloglucosidase (Sigma, A-3514 from A. niger in ammonium sulfate. Sigma-Aldrich
Co.**). (**Note: The amyloglucosidase used in this method was no longer stocked by
Sigma-Aldrich, so an attempt was made to replicate this solution.)
3.1.2.4 Glucose Analysis via Glucose Oxidase — Peroxidase (GOP! Method
The glucose analysis found in Hall (2000) includes two steps. The first is a starch
gelatinization and hydrolysis procedure. The second is a GOP analysis.
3.1.2.4.1 Starch Gelatinization and Hydrolysis Procedure
Duplicate 100 mg sub-samples of the EISF were weighed, recorded, and placed in
100 ml beakers. 20 ml of DIH2O and 0.1 ml of heat-stable-a-amylase (A-3306, Sigma-
Aldrich Company, St. Louis, MO) were added and stirred with a magnetic stir bar. The
beakers were covered with aluminum foil and placed in 93° C water bath. After 1 hour,
the beakers were removed and allowed to cool on the bench for 15 minutes. The solution
in each beaker was then filtered through glass wool plugs in funnels into 100 ml
volumetric flasks. The funnels and glass wool were rinsed and the filtered solutions were
|
Colorado School of Mines
|
22
adjusted to volume with DW. Through repeated inversion and shaking, the solutions were
mixed.
A 1 ml aliquot of each sample was then pipetted into individual 50 ml volumetric
flasks. 8 ml of 0.1 M sodium acetate buffer (pH ~ 4.5) along with 50 pL of
amyloglucosidase (A-3514 Sigma-Aldrich Company, St. Louis, MO) were added and
gently swirled to mix. The flasks were incubated in 60° C water bath for 30 minuets, and
gently swirled every 10 minutes during this procedure. The samples were then brought to
volume with DW and were ready for glucose analysis.
3.1.2.4.2 GOP Procedure for Glucose Analysis*
(*Note the glucose analysis method required a GOP solution and a glucose
standard solution whose preparations can be found in Hall 2000).
Duplicate 0.5 mL aliquots of each sample and of the glucose standard solution
were pipetted into acid washed 10 Hach vials. 2.5 ml of GOP was added to each vial,
then sealed and vortexed for 5 seconds. The vials were then placed into a rack in a 35 -
40° C water bath. After 45 minutes, the vials were allowed to cool for 10 minutes in the
dark. The absorbance of each vial was measured at X = 505 nm.
3.1.2.5 Neutral Detergent Soluble Fiber fNDSFl Analysis
The NDSF analysis uses a neutral detergent solution that was made up of the
ingredients can be found in Hall (2000). The solution was allowed to equilibrate
overnight and the pH adjusted the following morning to 6.9 - 7.0. Duplicate 1.000 (± .01)
grams dry wt. samples of the raw organic material were accurately weighed into 600 ml
beakers. 100 ml of the neutral detergent solution and 0.2 ml heat-stable -amylase were
then added to the beaker and swirled gently. The beakers were then covered with
|
Colorado School of Mines
|
23
aluminum foil and placed on a reflux burner set at high until the solution was brought to a
boil. Once boiling, the burner was turned down so that the solution was kept at a slow
boil. The solution is kept boiling for 1 hour and carefully swirled every 15 min. After 1 hr
from the time when boiling began, the solution was filtered through a Whatman 54 filter.
After all of the solution has been poured through the filter, the beaker was rinsed
completely with boiling water and poured onto the fiber mat. The fiber mat was rinsed
twice with boiling water followed by two rinses with acetone. The filter plus the mat was
placed onto a pre-weighed 2” aluminum pan and then into the 105° C drying oven
overnight. Once dry, the fiber mat was carefully scraped off the filter back into the pan.
The fiber mat weight ((weight of mat + pan) - weight of pan) was then weighed and
recorded.
3.1.2.6 Acid Hydrolysis Lignin Analysis
The Acid hydrolysis method uses 72% H2SO4 to dissolve all the organic fractions
of the organic material except lignin. After the NDSF analysis was performed, duplicate
0.300 g (dry wt.) of the neutral detergent insoluble fiber (NDISF) was weighed out into a
3” disposable test tube. 3 ml of 72% H2SO4 were pipetted and then mixed with a
disposable glass stir rod. Every sample had it’s own stir rod to prevent sample loss and
cross contamination. The test tubes were placed into a rack and in a 40° C water bath for
1 hour. The samples were mixed thoroughly every 5 min during this hour. After 1 hour
the sample were poured into a 100 ml glass serum bottle. The test tube and stir rod were
thoroughly rinsed with 5 ml increments of DW. A total of 40 ml of DW must be added to
the serum bottle (this value included the volume used for rinsing). The resulting
concentration of H2SO4 was then 4%.
The serum bottle were sealed with a septa and metal crimp. All serum bottles
were then placed into an autoclave for 1 hour set at 212° C. After 1 hour the solution were
|
Colorado School of Mines
|
24
filtered through a pre-weighed fritted bottom crucible (Coors porous bottom crucibles,
medium porosity VWR # 23857-047 - Coors # 60531). The entire solution as well as any
remaining material in the serum bottle, were washed with DW into the filtering crucibles.
The filter mats was rinsed with DW twice and then the crucibles were placed into a 105°
C drying oven overnight.
3.1.2.7 Compositional Analysis Calculation
The neutral detergent soluble fiber fraction (NDSF), organic acid fraction (OA),
and cellulose/hemicellulose fraction (CHF) of the organic materials were calculated as
the difference in mass among residues of known composition. The CHF was calculated
by subtracting the lignin fraction from the NDISF. The NDSF and OA fractions were
defined by the following equations (all fractions are shown in percents):
OA = (% ethanol solubility - TESC - (% total protein - EISF % Protein)) Eq. 3.1
NDSF = ((EISF % organic - EISF % protein) - (NDISF % organic
- NDISF % protein) - Starch) Eq. 3.2
A combination of NDSF + Starch fraction was estimated (due to problems quantifying
the starch fraction) from the percent solubility of organic materials in ethanol and NDS.
NDSF + Starch = (% NDS solubility - TESC - OA - (% total protein
- EISF % Protein)) Eq. 3.3
|
Colorado School of Mines
|
25
3.1.3 Water Soluble Fraction Analysis
Water solubility is not used by the NDSCNRA because it is not as precise as the
80% ethanol. Water will solublize the same fractions as the 80% ethanol but also extracts
polysaccharides such as pectic substances and fmctans.
The water solubility analysis was performed using the ethanol soluble fraction
analysis method. DI water was substituted for the 80% ethanol throughout the method.
3.2 Batch Manganese Sorption Experiments
The following experiment was performed to determine the manganese sorption
characteristics of all the organic materials. Duplicate 1.000 gram (dry wt. ± .005 g)
samples of each organic material were mixed with 100 ml solutions of varying
concentrations (0, 10, 25, 50, 75, 125, 175, and 250 mg/L) of Mn in 125 Erlenmeyer
flasks. The different concentrations were produced by diluting a 1000 mg/L stock
solution of Mn made up of 2.74 grams of MnS04 into a 1 L volumetric flask. The
mixtures were then covered and shaken in the dark for 24 hours. Table 3.1 shows all the
organic materials tested as well as the concentrations of Mn used. All samples were
filtered, acidified, and then run on a Perkin Elmer ICP-MS. All QA/QC protocols were
followed when running the ICP-MS analysis.
3.3 Column Experiments
Predominantly filled single organic material column experiments were performed
to observe each material’s initial sorption manganese removal capacity as well as the
SRR in a flow through system.
|
Colorado School of Mines
|
26
Table 3.1 Manganese concentrations used in sorption study
Organic material Concentration mg/L of Mn
Alfalfa 0, 25, 50,75, 125,and 175
Brewery Waste 0, 25, 50, 75, 125, and 175
Beet Pulp 0, 25, 50, 75, 125, and 175
Corncob 0, 25, 50, 75, 125, 175, and 250
Wetland 0, 25,50,75,125,and 175
Manure 0, 25, 50, 75, 125, 175, and 250
Maple Wood 0, 25, 50, 75, 125, and 175
Oak Wood 0, 25, 50, 75, 125,and 175
Pine Wood 0, 10, 25, 50, 75, 125, and 175
Poplar Wood 0, 25, 50, 75, 125, and 175
Walnut Hulls 0, 25, 50, 75, 125, 175, and 250
Walnut Wood 0, 10, 25, 50, 75, 125, and 175
3.3.1 Column Design
30 cm long by 5 cm diameter glass columns were used during the column
experiments. Each column had four 1 cm threaded side sample ports located at 6 cm
increments from the base of the column. All sample ports, except the effluent port, were
sealed with a rubber gasket and a screw-on cap. A 3” rubber stopper served as the top to
the column as well as aided in the release of gas. The stopper was drilled with a 1/8” hole
and a 4” long 1/8” outer diameter glass rod was placed through the stopper. A 1/8” tube
was then placed at the end of the outside-facing glass tube. To the other end of the 1/8”
tube, a one-way check valve was added. A plastic cap fit over the bottom of the columns
and had rubber o-rings to make sure the covers air and watertight. A metal screen was
placed on the inside of each end-cap to prevent the flow of bulk material exiting the
column. A threaded (female) lurer-lock fitting was screwed into each end-cap.
An Isometric peristaltic pump was set at level 7 and using a tube diameter of 0.59
mm provided a sustained flow of 30 ml/day. A 1/8” tube and fittings connected the pump
|
Colorado School of Mines
|
28
3.3.2 Column Mixture and Packing Procedures
150 grams of reactive mixture was weighed (dry wt. basis) for each column and
mixed in separate 1-gallon Ziploc bags. The reactive mixture was comprised of 45% # 8
mesh silica sand, 5% #10 mesh limestone, 10% fresh dairy manure, and 40% selected
organic material (either alfalfa, corncobs, oak, or pine). Once all of the components were
placed in the Ziploc bag, between 100 - 300 ml of 1000 mg/L SO4 solution was added to
aid in the homogenization processes. The bags were sealed and mixed by hand until fully
homogenized.
Duplicate columns were constructed for each selected organic material. All
columns are packed wet using a solution of 1000 mg/1 SO4. The reactive mixtures were
packed as tight as possible to reduce the potential formation of air pockets. All columns
were packed to below the second highest sampling port (average pack volume = 380 cm3
± 90 cm3). A metal screen was placed on top of the mixture to prevent the flow of bulk
material. The difference between the total and packed volume was then filled with silica
sand to prevent the mixture from expanding as well as acting as a filter for the effluent.
Once packed, a 3 cm by 1 cm piece of plastic mesh screen was folded and stuffed into
effluent side port to prevent clogging.
The packed columns were allowed to sit for 7 days with no pumping, and then an
influent of 1000 mg/1 SO4 was pumped through the columns for another 7 days. The
influent composition and flow rates for the column experiment are shown in Table 3.2.
This delay allows the establishment of the different microbial communities prior to
receiving high concentrations of heavy metals. The effluent was collected in sample
bottles and stored in a 4° C refrigerator prior to analysis. Table 3.3 shows the chemical
analyses run on all effluent samples.
|
Colorado School of Mines
|
30
CHAPTER 4
Results
4.1 Organic Material Characterization
The purpose of the organic material analysis was to use relatively simple
operationally defined methods to characterize organic materials, and then correlate the
resulting organic fractions to SRR, The different organic materials chosen to investigate
can be categorized into three groups: agricultural and industrial products (AIP), woods,
and microbial inoculum sources. The AIP includes: alfalfa, brewery waste, corncobs,
sugar beet pulp, and walnut hulls. The different woods tested were maple, oak, poplar,
walnut, and pine. The two inoculum sources were fresh dairy cow manure (manure) and
the sediment from a constructed wetland (wetland).
4.1.1 Physical and Elemental Analysis
The percent moisture, percent solids, and the percent organic fraction related to
raw sample weight are shown in Table 4.1. The AIP and woods average moisture was
relatively low at 7.2% ± 2.6%, while the inoculum sources are comprised of mostly water
with percent moisture averaging over 50% ± 14.9%. The deviation of the % solids is the
same as that of the % moisture.
|
Colorado School of Mines
|
4.1.2 Ethanol and Water Solubility Analysis
The solubility in ethanol and water of each organic material is shown in Table 4.3.
The water can dissolve polysaccharides (i.e. pectic substances and fmctans) in addition
the components dissolved by ethanol (Hall 2000). The water soluble fraction is larger
than the ethanol soluble fractions as expected
Table 4.3 Soluble fraction characteristics relative to dried sample
ESP WSF
Sample ID % %
Alfalfa 23.1 ± 1.27 32.1 ±0.09
Brewery Waste 19.4 ± 0.34 26.8 ± 0.70
Beet Pulp 12.3 ± 0.95 24.7 ± 0.80
Corncob 12.5 ± 1.59 17.2 ±2.67
Wetland 3.7 ± 0.28 10.1 ±0.64
Manure 3.7 ± 2.67 8.9 ± 1.16
Maple 0.8 ± 0.75 3.4 ± 3.43
Oak 5.7 ± 1.16 3.7 ± 1.21
Pine 2.3 ± 0.77 3.7 ± 0.22
Poplar 0.0 ± 0.00 1.7 ±0.25
Walnut Hull 0.5 ± 0.25 4.0 ± 0.06
Walnut Wood 4.0 ± 0.54 2.9 ± 1.93
4.1.3 Neutral Detergent-Soluble Carbohydrates Nutritional Relevance and Analysis
(NDSCNRA3
The nutritional analysis data presents (Table 4.4) (from left to right) the
carbohydrate fractions in order of lowest (i.e. organic acids) to highest (i.e. lignin)
structural complexity. The % starch values could not be analyzed following the method
found in the NDSCNRA because the amyloglucosidase required is no longer in
production. A substitute amyloglucosidase was used but it contained 10-12% starch and
ARTHUR LAKES LIBRARY
COLORADO SCHOOL OF MINE
GOLDEN, CO 80401
|
Colorado School of Mines
|
39
Figure 4.7 Sulfate reduction rate vs time for pine and oak columns
SRR for Pine & Oak vs. Time
0.9
i
c 0.7
f
0.6
% Pine 1
| 0.5 Pine 2
V) 0.4
$
Oak 2
_ 0.3
0.2
JS
S
0 10 20 30 40 50 60 70 80 90 100
Days
Both the corncob and alfalfa columns show a high variability during the initial
stages of column operation. The average and maximum SRR and standard deviations
(before and after day 45) for all columns are shown in Table 4.7. After day 45, the
average SRR was 0.60 and 0.54 mol S / m3 / day for the alfalfa and corncob columns
respectively. The average post 45 day SRR for oak and pine columns was 0.3 mol S/m3/
day for both. The effluent zinc concentrations for the pine and oak columns are shown in
Figure 4.8 and alfalfa and corncobs are displayed in Figure 4.9. The graphs for effluent
manganese for the pine and oak columns are shown in Figure 4.10 and alfalfa and
corncobs are displayed in Figure 4.11. All columns ran with an influent of just 1000
mg/L SO4 and no metals for the first 8 days. During this time, all columns showed some
leaching of zinc and manganese.
|
Colorado School of Mines
|
45
%OC %Protein %ESF %WSF %NDSC %HC+C %Lignin
Pine
50.3 2.0 2.25 3.65 11.36 64.46 22.82
Oak
50 2.6 5.66 3.67 16.78 59.72 20.74
Corncobs
42.8 6.9 12.48 17.23 25.45 56.38 11.48
Alfalfa
45 23.8 23.1 32.07 47.42 28.52 8.16
5.1.1 Percent Moisture and Organic Analysis
The percent moisture and organic faction are useful measurements for organic
material for the use in a passive treatment system (PTS), because both water and
inorganics can represent a significant fraction of the raw material weight and impact the
estimate of organic material added. Manure and wetland sediment are commonly used as
inoculum sources in PTS and average percent moisture of 55% and an organic fraction of
11%.
Manure has typically been placed in systems by either volume or as a weight
percent of the entire mixture (Kuyuck et al., 1994; Benner et al., 1999, Gusek et al.,
1998). With percent moisture values of up to 65%, the actual mass of solids may be as
low as 35% of the total mass of manure added. The organic fraction (an estimate of
available organic substrate) of manure was 10-20% by weight. Thus, as little as 10% of
the total mass of raw manure is potentially available to microorganisms. The amount of
organic material added could be underestimate by 90% if manure was added into a PTS
based on the raw sample weight.
|
Colorado School of Mines
|
46
5.1.2 Organic Carbon vs SRR
Organic carbon content of mixtures has been used as an estimate of available
substrate to the microbial communities. Waybrant et. al, (1998) found that in anaerobic
batch studies mixtures with the lowest carbon content (9-11%) had the lowest sulfate
reduction rates (0.14 mg L^day^g"1). Mixtures with the highest carbon content (35-47%)
were found to also have the highest sulfate reduction rates (SRR) (1.52-4.23 mg L'May"1
g"1). However, one mixture with 40% carbon did not follow the pattern of higher SRR
with higher organic content. This suggests that the SRR may also be dependent on other
nutritional characteristics.
A graph of the SRR of each column at day 20, 49 and 70 versus the percent
organic carbon is shown in Figure 5.1. The data presented in Figure 5.1 shows that higher
SRR is not correlated to higher organic carbon content. In fact, both alfalfa and corncob
(% organic carbon of 42.8 and 45, respectively) had higher SRR values than pine and oak
with 50% organic carbon. Therefore, other characteristics of the organic materials were
investigated.
|
Colorado School of Mines
|
47
Figure 5.1 SRR versus organic carbon content
0.7000
O Day 20
0.6000 ♦ Day 49
□ Day 70
0.5000
m
□ ♦
E
0.4000
(/) « F
o
□
| 0.3000 -O-
ûf
ûi 0.2000
V)
8
0.1000
8
" "
0.0000
42 44 46 48 50 52
% Organic Carbon
5.1.3 Protein vs SRR
Waybrant et. al, (1998) observed that one mixture studied that had high organic
carbon content and lower SRR values that other mixtures studied, also had a low total
nitrogen level. A comparison was also performed between the protein values of the
organic materials and the SRR from the column studies (see Figure 5.2). The data
presented in Figure 5.2 suggests that in this experiment, the protein content is
proportional to SRR. On day 49, the organic materials over 6% protein have similar peak
and day 49 SRR values, which suggests that at that time protein is not a limiting nutrient.
However, on day 70, the SRR is 1.5 times higher for the alfalfa than corncobs (23% and
6% protein respectively) suggesting that long-term SRR sustainability may be correlated
to protein content (r2 = 0.908 for a linear corrrelation).
|
Colorado School of Mines
|
49
The SRR results from the anaerobic biocolumn experiments (Figure 5.3) also show that a
correlation exists between the ESF and SRR (r2 = 0.956 for day 70 rates).
Figure 5.3 SRR vs ethanol soluble content
0.7000
0.6000
■o
m 0.5000
E □
S) 0.4000
♦
□ o Day 20
| 0.3000
♦ Day 49
□
□ □ Day 70
g| 0.2000
(/) 8
0.1000
"8“
0.0000 -Q —
0 5 10 15 20 25
% Ethanol Soluble Fraction
5.1.5 Water Solubility vs SRR
The WSF is important because water is the actual solvent for operational PTS and
was performed in the same manner as the ESF analysis. Another correlation can be drawn
between the WSF of the organic materials used in the column experiments and the
resulting SRR (Figure 5.4). As the WSF increases in each column so too does the peak,
day 49, and day 70 SRR (r2 = 0.957 linear correlation for day 70 rates).
The WSF of organic materials provides more information than just the percent
readily degradable organic substrates. The water solubility also can actually reduce the
amount of available food to microbial communities. The more soluble an organic
material is, the more likely that it is to be flushed from the system. In batch experiments,
|
Colorado School of Mines
|
50
this phenomenon is not a problem but in a column or full-scale flow through system, a
large portion of the organic material could potentially be washed from the system. Thirty
two percent of the mass in a system packed with alfalfa could potentially be flushed from
the system. A system comprised of wood would see only 4% removed. This loss of
material due to water solubility will decrease the total organic fraction available to
microorganisms in a PTS’s and thus decrease a system’s longevity.
Figure 5.4 SRR versus water soluble content
0.7000
0.6000
0.5000
0.4000
ODay 20
0.3000 ♦ Day 49
□ Day 70
0.2000
0.1000
0.0000
10 20 30 40
% Water Soluble Fraction
5.1.6 NDSC vs SRR
The NDSC analysis offers a less extensive and time consuming, but still valuable,
alternative to performing all the tests required for the NDSCNRA. The NDSC can be
performed in any lab with a hot plate, standard reagents, and a vacuum filter apparatus.
The NDSC fraction is a combination of the ESF, starch, and detergent soluble
|
Colorado School of Mines
|
51
polysaccharides fractions. The test separates the carbohydrates into the fractions that are
degraded through hydrolysis from the non-hydrolyzed components. Because hydrolysis is
the slowest degradation step, the NDSC can be used to estimate an organic material’s rate
of degradation.
Organic materials like alfalfa and sugar beet pulp have close to 50% of their
carbohydrates in the more highly degradable fraction. Woods are on the opposite end
with a NDSC fractions ranging from 8% for maple to 17% for oak. Walnut hulls,
corncobs, and brewery waste (24, 26, and 32% respectively) could provide the highly
degradable fraction initially for adequate SRB growth, but will also have a larger quantity
of the more slowly consumed carbohydrate source than the alfalfa and sugar beats.
Benner et al., (1984b) studied the anaerobic degradation rates of organic materials
having high and low NDSC fractions (grass and hardwoods). After 294 day, 30 % of
polysaccharides found in grass were mineralized, while only 4.1% of polysaccharides
were mineralize for hardwoods. This study showed that for the organic materials tested,
high NDSC was correlated to a higher extent of degradation. Benner also observed
differences between the lignin degradation rates of organic materials. Over the same time
period, 1.5% of the hardwood’s lignin was mineralized compared to 16.9% of grass’s
lignin. Benner was able to demonstrate that the biodégradation rates of the lignin
component of organic materials with high NDSC fractions are much higher than those
with low NDSC fractions. This suggests that even the slowly degraded fractions of
organic materials with high NDSC fractions are consumed faster than those materials
with low NDSC fractions.
Based upon the percents of mineralization over the 294 days and assuming a zero
order degradation rate, the lignin of the grass would be completely mineralized in 4.8
years. The wood, however, would provide a steady (albeit slow) source of degraded
carbohydrates for 53.7 years. Therefore, systems composed primarily of organic
materials with high NDSC fractions will not sustain microbial communities for long
|
Colorado School of Mines
|
54
5.2 NDSCNRA
The NDSCNRA offers an in-depth look at an organic material’s carbohydrate
characteristics and can contribute to a detailed comparison of organic materials
composition and their effects on initial and long-term SRR. The results of the NDSCNRA
(Table 4.4) correspond with Figure 2.1 and (from left to right) show the lowest to highest
structural complexity.
An understanding of the microbial communities, their sources of food, the
degradation products and rates, as well as, the complex symbiotic relationships within a
treatment system are vital when designing for long-term sustainability. With this
understanding, the NDSCNRA can be used as a tool to design a treatment system that
provides an ample SRR to precipitate metals and increase the alkalinity during startup,
but maintain an adequate SRR for the continuous treatment of metals and pH. Batch to
full-scale treatment studies, performed in the literature, found that the use of a mixture of
AIR and wood organic materials has provided the highest sustainable SRR (Pinto et al.,
2001; Gilbert et al., 1999; Waybrant et al., 1998; Kuyuck et al., 1994; Wildeman et al.,
1994). These studies conclude that mixtures that balance of organic materials containing
high organic acid and saccharides fractions (quickly consumed by SRBs) with those
comprised of mainly lignocelluloses (slowly degraded source of carbohydrates) result in
a sustained SRR. The NDSCNRA results confirm that a mixture of AIP and wood
materials would have such a mixture capable of sustaining an adequate SRR.
The NDSCNRA was a useful tool in the determination of the organic materials’
organic composition. A comparison of the actual % organic (tested in a muffle furnace)
and the sum of the carbohydrate totals can be found in Table 4.4. For the wood species,
(from which these analyses were derived) the percent differences between the actual and
calculated % organic were all below 2.7 percent. The AIP had more variability but all
were within a 7% difference except for corncob. The corncobs at 11.5% difference and
the inoculums (manure = 21% and wetland = 8%) had a high fines content that interfered
|
Colorado School of Mines
|
55
with some of the analyses. When compared to a batch of sugar beet pulp tested and
reported by Hall (2000), our own sugar beet pulp analysis resulted in difference of less
than 2% for the Summed Total and the % Organic (see Table 4.4). The disparity in the
other fractions can be attributed to seasonal, lot, and manufacturing differences between
the two batches of sugar beet pulp tested. Based upon these results the NDSCNRA
proved to be an effective method used in assessing the different carbohydrates fraction of
organic materials typically found in PTS.
5.3 Sorption
Batch sorption provides an estimate of the maximum amount of metal sorption
that can be expected, and thus allows for an approximation of time that sorption may be
the dominant removal mechanism. Manganese is commonly found in AMD but is not
readily removed from solution in an anaerobic biozone due to the pH greater than nine
required for insoluble manganese (II) hydroxides, carbonates, and sulfides formation
(Stumm and Morgan 1981, Wildeman and Updegraff 1997).
The manganese sorption capacity of manure is approximately 3.5 times higher
than alfalfa (27.0 and 7.6 mg/g respectively). Clayton et al., (1998) observed similar
affinities for manganese in identical field reactors containing either manure or alfalfa.
Over a 7 week period, the manure reactor exhibited greater than 97% removal of
manganese, while the alfalfa mixture showed no removal capacity. While it was found
true in the study performed by Clayton et al., high metal removal does not necessarily
signify that a material has a high sorption capacity. If there is a large volume of material,
even a low mg/g capacity will still be able to remove some metals initially.
The sorption capacity (mg of Mn sorbed/ gram organic material) for each column
was estimated at the Ce = 50 ml/L Mn for columns containing 60 grams of organic
material and 15 grams of manure. Based on an influent concentration of 50 mg/L Mn and
|
Colorado School of Mines
|
56
a flow rate of 30 mL/day, the theoretical maximum breakthrough time was estimated
(Table 5.2).
Table 5.2 Sorption longevity
Estimated*
qe
sorption Estimated
Sample (mg Mn/g
capacity Breakthrough
ID substrate)
(mg Mn) for Time (days)
for Ce = 50
entire column
Alfalfa 2.45 340 227.2
Corncobs 11.23 868 579
Oak 1.45 281 187
Pine 1.62 291 194
Manure 12.93
* based on 60 g primary substrate and 15 g manure
The average actual breakthrough time for manganese for all columns was 15 days.
The dissimilarity between the actual and theoretical breakthrough times are likely due to
differences in experimental conditions. The manganese batch sorption tests were carried
out by adding the organic matter to DI water containing only manganese and sulfate and
without adjusting the pH. The equilibrium pH of the organic materials tested ranged from
3.5 - 9.0, such a range will effect the solubilization or precipitation of metals. The second
factor diminishing the breakthrough of manganese was the competitive sorption of other
metals. The column influent contained equal concentrations of manganese and zinc. By
having a fixed number of sorption sites, the affinity of the organic material to either of
these metals would determine the final removal capabilities.
Kemdorff and Schnitzer (1980) examined different metals and their sorption to
humic acid at varying pHs and concentrations. As pH increases from a pH of 3.7 to 5.8
the percent sorption of manganese increases from 3 to 36%. Therefore, the organic
material’s equilibrium pH will greatly affect the extent of sorption. The spike in effluent
zinc observed in only one of the alfalfa columns was associated with lower pH.
|
Colorado School of Mines
|
57
Wieder (1990) looked at the sorption characteristics between metals (Al, Fe, Cd,
Ni, Zn, and Mn) and organic materials {Sphagnum peat and sawdust) typically found in
AMD treatment systems. Wieder determined that there is preferential sorption of
particular metals by organic materials. While the affinities of the different metals to
Sphagnum peat are ranked as:
Al3+ > Fe2+ > Cd2+ > Ni2+ = Zn2+ = Mn2+,
the affinities to sawdust are:
Zn2+ = Al3+ = Fe2+ > Mn2+ = Cd2+ = Ni2+.
The maximum manganese sorption capacities for both materials are similar to the average
of 3.6 mg Mn/g for the woods found in our analysis (Peat “a” = 4.27 mg Mn/g and
Sawdust “a” = 1.79 mg Mn/g). Wieder also used solutions containing equal
concentrations of all metals. These tests showed that the manganese sorbed was half that
of zinc (on a mass basis) for peat (0.3 and 0.6 mg/g respectively)
Machemer et al., (1992) performed batch sorption studies between fresh
mushroom compost and AMD. Their results were the same as those reported by
Kemdorff and Schnitzer (1980) where the amount sorbed for Fe, Cu, Zn, and Mn is as
follows:
Fe = Cu » Zn > Mn.
It was also observed that as the total metal concentration increased, the percent of Mn and
Zn adsorbed decreased and adsorption of Fe and Cu increased.
Machemer et al., then performed sorption experiments on a field-scale treatment
system for Fe, Cu, Mn and Zn. An average of 1.0 L/min AMD (40 mg Fe/1, 32 mg Mn/1,
9.2 mg Zn/1, and 0.6 mg Cu/1) flowed through a 5.6 m3 cell filled with fresh mushroom
compost. All metals were almost completely removed for the first 30 days. After day 30,
the average removal for Mn and Zn was only 20% while Fe and Cu continued to show
greater than 80% removal for another 70 days.
Batch sorption experiments can provide important information regarding the
sorption capacities of different metals to various organic materials. These sorption
|
Colorado School of Mines
|
58
capacities can be used to calculate theoretical breakthrough times. The accuracy of the
calculated breakthrough times will depend largely on the experimental conditions used.
Sorption capacities of organic materials are greatly effected by variations in pH and the
presence of multiple metals in the influent.
5.4 Column Studies
Four organic materials were selected for comparative column studies. Alfalfa has
the highest ethanol and water-soluble fraction, and second highest NDSC fraction.
Alfalfa has been recommended by a number of investigations, as an important component
for producing successful mixtures (Pinto et al., 2001; Tombre et al., 1996; Clayton et al.,
1998; Bechard et al., 1994; Gusek et al., 1998). Corncob was chosen primarily due to its
high sorption capacity of manganese. Oak was selected because out of the wood materials
tested it has the highest soluble fraction. Pine was selected due to its wide availability and
it is a soft wood.
5.4.1. Leaching of Inorganics
The four organic materials used in the column study were analyzed for leaching
using a DW control sample from the batch sorption experiments. A correlation between
the organic fraction and the leaching of inorganics is seen for the AIP (Figure 5.8) with a
second order polynomial (r2 = 0.974). Oak and pine have an inorganic fraction of 0.1%
and 0.0% respectively and close to no inorganics leached during the batch experiment.
These findings were supported during the startup of the column studies, when very little
zinc or manganese was found in the initial effluent. Alfalfa and corncobs have a higher
inorganic fraction (10 and 16.3 % respectively). This information along with their high
|
Colorado School of Mines
|
59
initial conductivity readings correlates qualitatively to the leaching of inorganics by the
alfalfa and corncob columns.
Figure 5.8 Mass of Fe. Mn. and Zn leached per unit mass of substrate versus inorganic
content
Total Fe, Mn, and Zn Leached v=ooooV»omixt oocee
vs % Inorganic
0.14
0.12
~o 0.10
£
Ô 0.08
8 0 06
I 0.04
pi
0.02
0.00
0 2 4 6 8 10 12 14 16 18
% Inorganic
The leaching of inorganics from the batch study may seem small (only 5.2 mg of
S/g of corncob), but when multiplied by amount of material placed in these systems the
effects can be quite significant. The effect of leaching can be seen in the initial column
effluents (Figure 4.4, 4.5, 4.9, and 4.11). The addition of just 60 grams of corncobs in our
column study more than doubled the initial effluent concentration of sulfur from 350
mg/L to near 800 mg/L. The alfalfa columns increased the effluent sulfur concentration to
550 mg/L but also added 7.0 mg/L Mn and 2.8 mg/L Zn. This initial increase in metals
can result in effluent emissions larger than is allowed by a system’s discharge limit.
The addition of sulfur through leaching can have profound effects on the design
and monitoring of treatment systems. Batch studies are typically used as a fast and easy
method for the comparison and selection of substrates to be used in PTS (Gilbert et al.,
1999; Pinto et al., 2001; Waybrant et al., 1998; Thombre et al., 1996 ). These batch
|
Colorado School of Mines
|
60
studies monitor how quickly the influent sulfur is reduced and a SRR is associated to
each of the organic materials tested. The release of significant quantities of sulfur by
organic materials will increase the concentration of sulfur in solution, resulting in
erroneous values used to calculate the SRR. The organic materials that leach sulfur are
inaccurately assessed a poorer SRR value and may not be considered for further
investigation. Therefore, it is important to know the approximate amount of inorganics
leached by an organic material before one can accurately measure its SRR.
The problem of sulfur leaching from organic materials was seen in the columns
containing corncobs and alfalfa. As stated earlier the average initial effluent sulfur values
for the corncob and alfalfa columns were 780 mg/L and 550 mg/L respectively. To
estimate the amount of sulfur leached by each organic material the mass of substrate in
the column is multiplied by the results found in Table 4.6. This calculation results in the
theoretical leaching of an additional 340 mg S and 130 mg S over the first few weeks for
corncobs and alfalfa respectively.
The SRB populations in the corncob and alfalfa columns may have been actively
reducing sulfate before day 10 and 20 respectively. The leaching of sulfur resulted in a
negative initial SRR value for both columns. To accurately calculate the initial SRR
values for these columns, or any others packed with sulfur leaching organic material,
identical columns would have to be run without any sulfate reduction taking place. By
comparing the effluent sulfur of the non-sulfate reducing and active columns, it could be
possible to calculate an estimated initial SRR for columns packed with sulfur leaching
organic material. The other alternative for calculating SRR is to simply wait for the
excess sulfur to be flushed from the system.
The leaching of sulfur into the effluent prevented an accurate calculation of SRR
for both the corncob and alfalfa columns during startup. The values of the SRR during
startup of the corncob and alfalfa columns were negative until day 10 and 20
respectively. Even though these systems could have been actively reducing sulfate, it is
only after these dates that positive SRRs can be calculated. Because of these problems, it
|
Colorado School of Mines
|
61
is not possible to identify when and at what rate did sulfate reduction actually begin for
these columns. However, the sulfate reduction occurring in these columns after the initial
flushing of inorganics is a valuable tool in addressing what effects varying substrate
characteristics have on short- and long-term SRR.
The quantity of Zn and Mn leached from each column during the initial 8 days of
operation is calculated by multiplying the effluent concentrations by the flow rate and the
day intervals (Actual Mn and Zn values Table 5.2). The estimated metals leached from
the column (Estimated Mn and Zinc values Table 5.2) were derived by multiplying the
mass of organic material found in each column (60g organic material and 15g manure) by
the amount of Zn and Mn leached from these materials during the sorption studies (Table
4.6).
Table 5.3 Metal leaching from column experiments
Metals leached during column studies (mg]
Actual Estimated* Actual Estimated*
Sample ID
Manganese Manganese Zinc Zinc
Alfalfa 2.07 ± 0.01 1.3 0.61 ±0.18 1.4
Corncob 1.06 ±0.26 0.6 0.59 ± 0.43 5.2
Oak 0.33 ± 0.07 0.3 0.02 ± 0.01 0.2
Pine 0.35 ± 0.05 0.4 0.01 ± 0.002 0.3
Dased on batch studies
Experimental design differences between the sorption and column analyses and
characteristics of the two metals likely contributed to either the over- or underestimation
of leaching. The batch sorption experiments were performed using a control of DI H2O,
were run for only 24 hours, and were agitated. The column experiments were wet packed
with a 1000 mg/L SO4 solution, underwent no pumping for the first 7 days, and for the
following 7 days a 1000 mg/L SO4 solution was pumped at a rate of 30 ml/day.
The estimated zinc leached from all columns was overestimated on average by a
factor of the 12, while manganese was underestimated on average by a factor of 1.3.
Unlike manganese, zinc’s readily formed metal sulfide complexes are insoluble at a
|
Colorado School of Mines
|
62
neutral pH. The lack of sulfate and the 24-hour timeframe limits the likelihood of sulfate
reduction interfering with the batch sorption/leaching test. However, the column study
was designed to minimize the sulfate reduction’s startup phase. Therefore, it is possible
that the overestimation of zinc leached from the columns was due to sulfate reduction
activity. Manganese sulfides remain soluble at neutral pHs and its leaching would not be
affected by active sulfate reduction.
Preferential sorption by organic material may also explain why the estimate and
actual leaching values are different. Sorption studies in the literature have shown that
manganese is typically the least preferentially sorbed metal by various organic materials
(Kemdorff and Schnitzer 1980, Machemer et al., 1992, Wieder 1990) Machemer et al.,
(1992) also reported that as metal concentrations increase, manganese is less competitive
for adsorption sites than other metals. During the batch sorption test, organic materials
were tested and reached equilibrium in isolation. The column studies contain a mixture of
manure, limestone and sand, in addition to the organic material. The metals leached from
these additional materials could possibly out compete manganese for sorption sites and
therefore may have cause the underestimation for manganese.
While not completely accurate, the batch sorption study was able to approximate
the leaching of manganese from the column system. The differences in experimental
design and chemical characteristics likely caused the overestimation of zinc and the
underestimation of manganese. If the experimental design of both systems were more
closely matched (i.e. similar duration, testing solutions, and organic material mixtures
were used in both experiments), then a better estimation of metal leaching may be
possible.
|
Colorado School of Mines
|
63
5.4.2 Sulfate Reduction Rates
The duplicate columns of both the pine as well as the oak columns started
reducing sulfur around day 20 (See Figure 5.9). The slope of the oak columns’ initial
increase in SRR was approximately 0.032 SRR/day for the first 10 days. A high SRR of
over 0.5 moles S/m3/day was sustained for the following 10 days. After the peak in SRR,
a declined was seen for 36 days at a rate of -0.0086 SRR/day. The final 10 days of testing
show the SRR stabilizing at a rate of 0.2 moles S/m3/day. This overall trend of a rapid
startup, a shortly sustained peak, followed by a steady decrease, and then a stabilizing of
the SRR has been seen repeatedly in the literature (Seyler et al., 2003; Wildeman et al.,
1994; Waybrant et al., 2002; and Eger and Wagner 1995).
The pine columns SRR curve differs from that of the oak and other studies, in that
it lacks a dramatic initial increase. The pine columns have an initial increase in rate of
only 0.006 SRR/day for 20 days. The maximum SRR value achieved by either column is
0.38 moles S/m3/day. An average SRR of 0.3 moles S/mVday is maintained from day 40
to day 60. After day 60, an average decline of -0.0036 SRR/day continues for the
remainder of the experiment.
ARTHUR LAKES LIBRARY
COLORADO SCHOOL OF MINES
GOLDEN, CO 80401
|
Colorado School of Mines
|
64
Figure 5.9 Changes in SRR for Oak and Pine Columns
SRR vs Time for Oak and Pine
Oak< 40 Days
0.6
y= 0.032k- 0.5888
R2= 0.9375
0.5
Oak> 50 Days
0.4 y=-0.0086k+0.8396
R2= 0.916
0.3 Fine < 50 Days
y= 0.0064k-0.0101
R2= 0.8584
0.2
Fine > 50 Days
0.1 y =-0.0036k + 0.5029
R2= 0.3564
0
0 100
20 40 60 80
Days
♦ Pine <50 day ♦ Pine >50 days » Oak< 40 days ■ Oak> 50 days
The differences seen between the two wood species can be attributed to both
chemical and carbohydrate attributes. The oak exhibits a similar SRR pattern as those
found in other studies. However, the oak columns fail to reach or maintain a SRR as high
as that seen in a similar column study using a mixture of organic materials (Seyler et al.,
2003). Oak’s limited source (5.7 % ESF) of easily degraded carbohydrates appears to be
sufficient to result in a moderate initial SRR. It is hypothesized that once these simple
carbohydrate sources were all consumed by the microbial community, sulfate reduction
became a rate-limited process. The SRB became reliant on other microbial communities
within the column to provide energy sources through the degradation of complex
carbohydrates. The decrease in SRR occurred in the oak columns because the rate of
degradation of the complex carbohydrates was not sufficient to sustain the initial high
SRR.
Pine is comprised of even less easily degraded carbohydrate sources than oak
(2.3% ESF). Pine also has been described as having antimicrobial properties that could
affect SRR (Eger and Wagner 1995). Therefore, it comes as no surprise that the SRR of
|
Colorado School of Mines
|
65
the pine columns did not obtain the same SRR as the oak or as that found in the literature.
The lack of an initial SRR peak however, was different than that seen in the studies
previously reported. The SRR in the pine columns slowly increases for 38 days until it
reaches that of the oak columns. From this point forward the SRR of the two materials
are almost identical. The ability of the pine columns to eventually attain and then equal
the oak columns suggests that the antimicrobial substances found in pine may have
prevented the initial peak in SRR and that only after these substances are leached from
the column could uninhibited sulfate reduction occur.
The results from both the oak and pine column studies suggest that woods lack an
adequate supply of structurally simple carbohydrate fraction to maintain a high initial
SRR. Pine may also possess antimicrobial toxins that inhibit the growth of the microbial
communities that support the sulfate reduction process. The final level of SRR (0.2 moles
S/m3/day) for the oak and pine columns is less than the long-term rates found in the
literature (Seyler et al., 2003, Waybrand et al., 2002; Eger and Wagner 1995; Wildeman
et al., 1994). This low rate of SRR is likely being sustained by the degradation of the
lignocellulose fraction found in the woods. The quantity of lignocellulose in these
organic materials and the rate at which it is being degraded could potentially maintain the
present SRR for an extended period of time.
The duplicate columns packed with alfalfa had the highest average and maximum
percent difference of the four organic materials tested (Figure 5.10). Both columns show
a positive SRR after day 20. For the following 10 days, column 2 shows the highest
initial increase in SRR of any column with a slope of 0.06 SRR/day and a peak value of
0.7 moles S/m3/day. During this same period, column 1 remains at a SRR of less 0.1
moles S/m3/day. Starting on day 33, column 2 exhibits the typical decrease in sulfate
reduction at a rate of -0.018 SRR/day. Unlike the SRR seen in the wood columns or
preliminary column and literature studies, a second increase in SRR (0.0048 SRR/day)
occurred between day 44 and the end of sampling. Column 1 begins its initial increase in
sulfate reduction 10 days after column 2, and maintains a steady increase (0.027
|
Colorado School of Mines
|
67
Figure 5.11 Changes in SRR for Corncobs Columns
SRR vs Time for Corncobs Corncob 1A
y = 0.0129x+0.0821
0.9 R3 = 0.7819
0.8 Corncob 1B
y = 43.0213x+0.9452
0.7 R2 = 0.9587
Corncob 1C
5 0.6
y = 0.014x-0.1856
0.5 R?= 0.9186
0.4
Corncob 2A
y = 0.0521 x-0.4935
0.2 R2= 0.6319
Corncob 2B
0.1 y = -0.0339X +1.5136
R3 = 0.9067
0.0
Corncob 2C
0 10 20 30 40 50 60 y = 0,014x-0.0066
R' = 0.9483
Days
A Corncob 1A a Corncob 1B ♦ Corncob 2A Corncob 2B
• Corncob 2C • Corncob 1C
The two alfalfa columns showed large disparities in SRR. Although the reasons
behind these differences are not fully understood, there are trends within the data that
correlate effluent pH to SRR and metal removal. On day 21, the SRR and effluent pH,
Mn, and Zn values were statistically the same for both columns. From day 23 to 32, the
pH of column 2 begins a steady increase to an average of 6.5, while column 1 remains at
a pH of 5. During this period of increasing pH, column 2 shows its initial SRR peak,
alkalinity greater than 1100 mg/L CaCOs, and the complete removal of Zn in the effluent.
Column 1 shows no initial SRR peak, alkalinities of 600 mg/L CaCOs, and a rapid
increase in effluent zinc and manganese concentrations. On day 35, column 1 finally
exhibits a simultaneous increase in effluent SRR, alkalinity, and pH, as well as, a
decrease in zinc and manganese. A higher correlation exists between SRR vs pH for
column 1 than column 2. A correlation can also be made between column 1 ’s SRR and
the removal of both Zn and Mn. The differences seen in the two alfalfa column effluents
may be affected by a number of factors (i.e. heterogeneity of the inoculum source, short-
|
Colorado School of Mines
|
68
circuiting in the column, competition, etc...). While the reasons for column 1 ’s lag in
SRR is not clearly understood, this lag has allowed for the relationship between pH and
SRR as well SRR and metal removal to be examined.
The unique slopes in SRR of the alfalfa 2 and corncob columns (Figures 5.10 and
5.11) are consistent with the ideas found in the conceptual microbial process model
(Figure 2.2). Alfalfa and corncobs had the highest quantities of organic acids and low
levels of simple saccharides. The initial SRR peak seen in all three columns could be
associated with the consumption of these readily available organic acids. The following
decline in SRR suggests that energy sources of the sulfate reducers had become limited
and were not sufficient to sustain the high rate of sulfate reduction. As the population of
fermentative bacteria increases, so to will the rate at which these microbes can degrade
monomers to energy sources accessible to SRB. The second increase seen in the column
effluent SRR is likely the result of this process providing an additional energy source to
SRB.
A lack of simple/soluble substrate availability is one possible scenario explaining
alfalfa column 1 not exhibiting two separate SRR peaks. During days 23 through 35,
alfalfa column 1 sustained a positive but low rate of SRR (< 0.1 moles S/m3/day). By the
time an increase in SRR became active on day 35, alfalfa column 1 had undergone 12
additional days of flushing. This additional flushing may have removed some of the
soluble organic acids responsible for the initial SRR peak seen in the corncob and other
alfalfa columns. The rate at which the initial SRR increased in alfalfa column 1 is also
lower than the SRR/day rate seen in column 2. Therefore, the additional flushing
experienced by alfalfa column 1 is likely to be responsible for the lack of two SRR/day
peaks and a lower initial SRR/day rate.
Although the trend in SRR measured in alfalfa column 2 is not replicated by
column 1, it is very similar to those seen in both of the corncob columns. These trends are
consistent with the idea that organic substrates with low C:N ratios and very high organic
acid/NDSC fractions will produce a very rapid initial rate of SRR/day. The high SRR
|
Colorado School of Mines
|
69
maintained by these columns to date is expected to last as long as the rate of degradation
can provide the adequate energy sources for the SRB. Alfalfa, with the lowest hemi-
/cellulose fraction and the second lowest lignin fraction, is expected to be degraded very
rapidly. The rate at which alfalfa is degraded will likely maintain a high SRR until all the
easily degradable carbohydrates are consumed. It is presumed that this will cause a
dramatic reduction in SRR. Corncobs will likely not maintain as high of a SRR as the
alfalfa columns due to its higher hemi-/cellulose and lignin content. This higher complex
carbohydrate fraction, is expected however to provide the corncobs columns with a lower
but longer lasting SRR. By understanding an organic material’s carbohydrate
composition, rate of degradation and the relationship between microbial communities
found in RTS, one can approximate its the short and long term effects on the SRR.
5.4.3 Metal Removal Capabilities
The metal removal capabilities of the columns are consistent with the results of
the manganese sorption analysis and the column effluent pH. Corncobs and alfalfa were
found to have higher sorption capacities than pine and oak in the sorption experiment, but
show conflicting results in their ability to remove manganese in a RTS. Shortly after the
Zn and Mn were added on day 8, all columns except corncob 1 show an increase in
manganese effluent concentration. The two alfalfa columns initially show the worst
manganese removal capabilities, but once the reduction of sulfate begins the manganese
removal improves. Oak and pine initially mimic the alfalfa columns but from day 23 to
58 they maintain an average effluent concentration of less than 15 mg Mn/L. Since day
58, however, the effluent manganese concentration for oak and pine has steadily
increased. The corncob columns do show the ability to removal manganese initially.
From day 11 to 21 corncob column 1 shows complete removal of manganese. Column 2
does not completely remove manganese but maintains the effluent concentration below 5
|
Colorado School of Mines
|
71
CHAPTER 6
SUMMARY
6.1 Conclusions
The highest overall sulfate reduction rate (SRR) as well as the highest sustained
SRR for 70 days of column operation were observed for the column with the highest
neutral detergent soluble compounds (NDSC), water soluble fraction (WSF) and ethanol
soluble fraction (ESF) with r2 = 0.95-0.96 for a linear correlation. NDSC, WSF and EST
appear to be good measures of the relative degradation rates of the organic material.
Organic carbon and total organic content had inverse relationship to the observed SRR
and thus were poor measures of the relative degradation rates of organic material.
A correlation exists between manganese batch sorption studies and the initial
removal of manganese in the column tests. However, the batch tests severely
overestimate the maximum manganese removal capacity. Differences in water
characteristics (pH and other ions) and column composition between batch and column
experiments were probably factors in the inability to predict the column sorption behavior
from batch data.
The potential for organic materials to leach inorganics targeted for removal in
PTS (e.g. zinc and manganese) can be estimated by batch incubation of organic materials
with water. Estimates of manganese leaching from batch tests were 60 to 100% of the
measured values. Estimates of zinc leaching from batch tests were 100 to 1000% of the
measured values. The difference between manganese and zinc may be related to the
higher sorption of zinc relative to manganese for the sand and limestone fractions of the
packed columns.
|
Colorado School of Mines
|
72
6.2 Future Work
Future work of this project should include the continued monitoring of the current
columns for SRR and metal removal until failure of all columns. The information gained
from the final SRR should be reanalyzed for correlations between the long-term SRR and
the organic materials’ compositions. Similar long-term column studies should be
performed on organic materials having high %NDSC fractions but protein fractions to
determine if a correlation between protein content and SRR exists.
A more complete sorption analysis should be performed for all organic materials.
This analysis should include a mixture of metals to identify the effects of competitive/
preferential sorption for all materials. An additional sorption analysis should also be
performed on the actual organic mixtures as are found in the column test. These analyses
may provide a better approximation of theoretical breakthrough times and leaching of
zinc and manganese
Based on the long-term results of the column experiments and the sorption
analyses, a mixture of organic materials should be selected that can provide a high but
sustained SRR. Variations in the percent of the different organic materials should be
tested in similar column studies to see which combination provides the desired results.
The superior mixture should then be tested in a two-dimensional pilot-scale system and
eventually implemented in a full-scale field system.
|
Colorado School of Mines
|
74
DeBusk, W. Wastewater Treatment Wetlands: Contaminant Removal Processes This
document is SL155, a fact sheet of the Soil and Water Science Department,
Florida Cooperative Extension Service, Institute of Food and Agricultural
Sciences, University of Florida. Published: May 1999. Please visit the EDIS Web
site at http://edis.ifas.ufl.edu/BODY_SS293
Dietz, J., and R. Unz, 1988. Effects of Sphagnum Peat on the Quality of a Synthetic
Acidic Mine Drainage. In Proceedings of Mine Drainage and Surface Mine
Reclamation, Vol. I: Mine Water and Mine Waste, BuMines IC 9183, pp. 310 -
316.
Dvorak., Hedin, R., Edenborn, H., and P Mclntire, 1992. Treatment of Metal-
Contaminated Water Using Bacterial Sulfate Reduction: Results from Pilot-Scale
Reactors. Biotechnology and Bioengineering, Vol. 40, no. 5, pp. 609 - 616.
Clayton, L. A., J.L. Bolis, T.R. Wildeman, and D M. Updegraff 1998. A Case Study on
the Aerobic and Anaerobic Removal of Manganese by Wetland Processes.
Reviews of Economic Geology. Vol 6 B, Chapter 26. p. 647 - 655.
Cocos, I., Zagury, G., Clement, B., and R. Samson, 2002. Multiple Factor Design for
Reactive Mixture Selection for Use in Reactive Walls in Mine Drainage
Treatment. Water Research. Vol 32, pp 167 - 177.
Eger, P., and J. Wagner, 1995. Sulfate Reduction for the Treatment of Acid Mine
Drainage; Long Term Solution or Short Term Fix? In Proceedings of the
Conference on Mining and the Environment. Sudbury, Ontario, pp. 515 - 524
Elliott, P., Ragusa, S., and D. Catcheside, 1998. Growth of Sulfate-Reducing Bacteria
Under Acidic Conditions in an Upflow Anaerobic Bioreactor as a Treatment
System for Acid Mine Drainage. Water Research. Vol. 32, No. 12, pp 3724 -
3730.
Emerick, J., Huski, W., and D. Cooper, 1988. Treatment of Discharge from a High
Elevation Metal Mine in the Colorado Rockies Using an Existing Wetland. In
Proceedings of Mine Drainage and Surface Mine Reclamation, Vol. I: Mine
Water and Mine Waste, BuMines IC 9183, pp. 345 - 351.
Gerhardt, P., 1981. Manual of Methods for General Microbiology. Washington D C.:
ASM Publication
|
Colorado School of Mines
|
75
Gilbert, J.S., Wildeman, T.R., K.L. Ford, 1999. Laboratory Experiments Designed to Test
the Remediation Properties of Materials. In proceedings of the 15th Annual
Meeting of the American Society of Surface Mining and Reclamation. Scottsdale,
AZ, pp. 582 - 589.
Gottschalk, G., 1986 Bacterial Metabolism. Springer - Verlag.
Gusek, J., Wildeman, T., Miller, A., and J. Fricke, 1998. The Challenges of Designing,
Permitting and Building a 1,200 GPM Passive Bioreactor for Metal Mine
Drainage West Fork Mine, Missouri. In Proceedings of the 15th National Meeting
of the American Society for Surface Mining and Reclamation. St Louis, MO.
Gusek, J., 2001. Why do Some Passive Treatment Systems Fail While Others Work? In
Proceedings of the National Association of Abandoned Mine Land Programs.
Athens, OH.
Gusek, J. and T. Wildeman, 2002. A New Millennium of Passive Treatment of Acid
Rock Drainage: Advances in Design and Construction Since 1988. In Proceedings
of the National Meeting of the American Society for Mining and Reclamation.
Lexington, Kentucky.
Hall, M., 2000. Neutral detergent-soluble carbohydrates: nutritional relevance and
analysis (a laboratory manual). University of Florida Extension Bulletin 339.
Hammack, R., Edenborn, H., and D. Dvorak, 1994. Treatment of Water from an Open-Pit
Copper Mine Using Biogenic Sulfide and Limestone: A Feasibility Study. Water
Research. Vol 28, no. 11, pp 2321 - 2329.
Hedin, R., Hyman, D , and R Hammack, 1988. Implications of Sulfate-Reducing and
Pyrite Formation Processes for Water Quality in a Constructed Wetland:
Preliminary Observations. In Proceedings of Mine Drainage and Surface Mine
Reclamation, Vol. I: Mine Water and Mine Waste, BuMines IC 9183, pp. 382 -
388.
Herbert, R., Benner, S., and D. Blowes, 1998. Reactive Barrier Treatment of
Groundwater Contaminated by Acid Mine Drainage: Sulfur Accumulation and
Sulfide Formation. In Proceedings of the Groundwater Quality 1998 Conference,
Tübingen, Germany.
|
Colorado School of Mines
|
76
Huntsman, B.E., J.G. Solch, and M.D. Porter, 1978. Utilization of Sphagnum Species
Dominated Bog for Coal Acid Mine Drainage Abatement. GSA (91st Annual
Meeting) Abstracts, Toronto, Ontario.
Kemdorff, H. and M. Schnitzer, 1980. Sorption of Metals on Humic Acid. Geochimica et
Cosmochimica Acta. Vol 44, pp. 1701 - 1708.
Kepler, D., 1988. An Overview of the Role of Algae in the Treatment of Acid Mine
Drainage. In Proceedings of Mine Drainage and Surface Mine Reclamation, Vol.
I: Mine Water and Mine Waste, BuMines IC 9183, pp. 286 - 290.
Kinniburgh, D.G., 1986. General Purpose Adsorption Isotherms. Environmental Science
and Technology. Vol 20, pp. 895 - 904.
Kuyuck, N., and P. St-Germain, 1994. In Situ Treatment of Acid Mine Drainage by
Sulfate Reducing Bacteria in Open Pits: Scale-up Experiences. In Proceedings of
the International Land Reclamation and Mine Drainage Conference & 3rd
International Conference on the Abatement of Acidic Drainage. Pittsburgh, PA.
Vol 2, pp. 303-310.
Lang, Gerald, R.K. Wieder; A.E. Whitehouse, 1982. “Modification of Acid Mine
Drainage in Freshwater Wetland”, In Proceedings of the West Virginia Surface
Mine Drainage Task Force Symposium, Morgantown, WV. April.
Lapakko, K. and P. Eger, 1988. Trace Metal Removal From Stockpile Drainage by Peat.
In Proceedings of Mine Drainage and Surface Mine Reclamation, Vol. I: Mine
Water and Mine Waste, BuMines IC 9183, pp. 291 - 300.
Logan, M., Ahmann, D., Figueroa, L., and T. Wildeman, 2003. Assessment of Microbial
Activity in Anaerobic Columns Treating Synthetic Mine Drainage. In Proceedings
of the Joint Conference of the 9th Billings Land Reclamation Symposium and the
20th Annual Meeting of the American Society of Mining and Reclamation.
Billings, Montana.
Lyew D., Knowles R. and Sheppard J. (1994) The biological treatment of acid mine
drainage under continuous flow conditions in a reactor. Trans IchemE. Vol 72(B),
pp. 42-47.
Machemer, S., and T. Wildeman, 1992. Adsorption Compared with Sulfide Precipitation
as Metal Removal Processes from Acid Mine Drainage in a Constructed Wetland.
J. of Contaminant Hydrology. Vol 9, pp. 115-131.
|
Colorado School of Mines
|
78
Samuel, D., Sencindiver, J., and H. Ranch, 1988. Water and Soil Parameters Affecting
Growth of Cattails: Pilot Studies in West Virginia Mines. In Proceedings of Mine
Drainage and Surface Mine Reclamation, Vol. I: Mine Water and Mine Waste,
BuMines IC 9183, pp. 369 - 374.
Sencindiver, J.C. and D.K. Bhumbla, 1988 Effects of Cattails (Tvpha) on Metal Removal
from Mine Drainage. In Proceedings of Mine Drainage and Surface Mine
Reclamation, Vol. I: Mine Water and Mine Waste, BuMines IC 9183, pp. 359 -
366.
Seyler, J., Figueroa, L., Ahmann, D., Wildeman, T., and M. Robustelli, 2003. Effect of
Solid Phase Organic Substrate Characteristics on Sulfate Reducer Activity and
Metal Removal in Passive Mine Drainage Treatment Systems. In Proceedings of
the Joint Conference of the 9th Billings Land Reclamation Symposium and the
20th Annual Meeting of the American Society of Mining and Reclamation.
Billings, Montana.
Spratt, A.K. and R.K. Wieder, 1988. Growth Responses and Iron Uptake in Sphagnum
Plants and Their Relation to Acid Mine Drainage Treatment. In Proceedings of
Mine Drainage and Surface Mine Reclamation, Vol. I: Mine Water and Mine
Waste, BuMines IC 9183, pp. 279 - 285.
Stumm, Werner and James J. Morgan, Aquatic Chemistry. Wiley-Interscience, New
York, 1970; 2nd edition, 1981, 3rd edition, 1996.
Thombre, M., Thomson, B., and L. Barton, 1996. Microbial Reduction of Uranium Using
Cellulosic Substrates. In Proceedings of the 1996 HSRC and WERC Joint
Conference on the Environment. Albuquerque, NM
Tuttle, J., Dugan, P., and C. Randles, 1969. Microbial Sulfate Reduction and Its Potential
Utility as an Acid Mine Water Pollution Abatement Procedure. Applied
Microbiology. February pp. 297 - 302.
UDSA. 1993. Acid drainage from mines on the National Forest: A management
challenge. U.S. Forest Publication, 1505, 1-12.
Waybrant, K., Blowes, and C. Ptacek, 1998. Selection of Reactive Mixtures for Use in
Permeable Reactive Walls for Treatment of Mine Drainage. Environmental
Science and Technology. Vol 32, pp 1972 - 1979.
|
Colorado School of Mines
|
81
A.1 Preliminary Column Studies
A. 1.1 Introduction
The initial stages of this research focused a review of the current literature, the
development of a conceptual microbial model, and a column study. The literature review
illustrated that the primary focus of the research to date was on establishment of a highly
active SRB community. Organic materials were selected based upon their initial
metal/sulfate removal rates rather than on performance longevity. There were no
mentions of the other microbial populations, or the important roles they play in PTS.
There also lacked a systematic approach for selecting a substrate mixture. The success of
these PTS to reduce sulfur and precipitate metals were mixed.
A conceptual microbial model was developed that tried to incorporate the
microbial communities and carbohydrate fractions found within PTS. With this microbial
model, we were able to hypothesize that a PTS could successfully reduce sulfate over a
long period of time if the proper carbohydrate fractions were selected. SRB rapidly
consume simple carbohydrates. As the source of simple carbohydrates diminishes within
a PTS, SRB become reliant on other microbial communities to degrade the more complex
carbohydrates. The slower degradation rates of the complex carbohydrates will ensure a
low but long-term energy source for SRB.
Two column studies were performed during the first year of research. The first
study wanted to reproduce the methods and results based on previously performed studies
and literature. The focus was establishing an active SRB community, achieve similar
SRR to those found in the literature, and remove metals through sulfide precipitation.
This study also provided practical experience for assessing procedures of packing,
sampling, and monitoring columns, as well as identifying any potential problems. The
second column study investigated the sulfate reducing capacity of several individual
|
Colorado School of Mines
|
82
substrates over a long period of time. Columns were packed with a single substrate and
an inoculum source and differences in initial and long-term SRR, as well as metal
removal capabilities, were examined.
A. 1.2 Method
The following section will discuss the methods for both the initial and mini
column studies.
A. 1.2.1 Initial Column Study
The three preliminary columns were made out of a 30 cm long by 5 cm diameter
glass column having 4 side sample ports from which effluent or intermediate sample
could be taken. The sample ports were sealed with a rubber gasket and a screw-on cap.
Plastic caps fit over the ends of the columns and had rubber o-rings to make sure the
covers airtight. A metal screen was placed on the inside of each end-cap to prevent the
flow of bulk material exiting the column. A threaded (female) lurer-lock fitting was
screwed into each end-cap. The male lurer-lock/ tube fitting then connects the column
to the influent and effluent tubes. In our experiment we used a 12 channel Ismatec IPC
peristaltic pump and 0.89 mm Ismatec pump tubes. With this size tubing and the pump
set at 9, a constant flow of 165 ml/day was pumped through the column.
The preliminary columns were loosely packed with the mixture described in
Table Al. The total mass of reactive mixture added and the reactive volume of our
column setup are presented in Table Al. The concentrations and properties of the influent
mine water collected at the Dinero tunnel (Leadville, CO) are presented in Table A3. The
influent water was pumped from the bottom to the top of the column to prevent short-
|
Colorado School of Mines
|
83
circuiting of influent. Effluent was collected in iced sample bottles and was stored in a 4°
C refrigerator prior to analysis.
A. 1.2.2 Mini Column Single Organic Material Studies
The second study used 40 ml round bottom centrifuge vials (Table Al) to assess
column sorption. A 1/8” hole was drilled in the screw cap and the bottom of each column.
A tubing elbow was then epoxied into each whole. A plastic mesh screen was placed
at the bottom and top of each column to prevent exiting of the bulk material. Duplicate
columns were packed using the compositions found in Table A2. Due to different bulk
densities of the organic materials, the amount of sand varied for each column. Again, the
influent was pumped in an up-flow direction using 0.19 mm Ismatec pump tubing with
the peristaltic pump set at six. The flow rate and the post-twenty-day synthetic influent
characterization can be found in Table A3. Dinero tunnel water was used as a primer
influent, which allowed the SRB a chance to start reducing sulfate prior to receiving the
higher concentration influent. Effluent was collected in iced sample bottles and was
stored in a 4° C refrigerator prior to analysis. Table A4 lists the chemical analyses
performed and the machines used.
Table Al Column characteristics
Mini-Column
Preliminary Column
Single Organic material
Column Volume (ml) 589 40
Reactive Volume (ml) 589 40
Diameter (cm) 5 1.25
Length (cm) 30 8.5
Reactive Mass (g) 185 17
|
Colorado School of Mines
|
87
A. 1.3.2 Mini Column Single Organic Material Studies
After the successful design and operation of the preliminary columns, a new
experiment examining the effects of different substrates on SRR and metal removal was
initiated (Figure A3 & A4). Columns showed little reduction in sulfate concentrations
compared to similar column studies with a more diverse mixture (M. Logan et al., 2003).
The only signs of reduction were when the hydraulic residence time was inadvertently
longer. This suggests that the retention times were too short, and did not allow sufficient
time for effective sulfate removal. Due to this problem, the differences between the
organic material’s SRR could not examined. The problems encountered using the mini
column design led a return to the larger glass column setup and additional design
modifications.
With little to no sulfate reduction occurring, sorption is a viable reason why some
metals were removed. All substrates were effective at removing zinc from the influent
mine water, however the removal of iron and manganese varied. The ability of corncobs
and walnut hulls to remove manganese was especially interesting. Manganese sulfides are
typically not removed in PTS because they remain soluble up to a very high pH. The
differences in metal manganese sorption abilities seen in this experiment led to the
investigation of all organic material’s sorption capacity.
|
Colorado School of Mines
|
91
o o ■M" O o N N CO CO ■M"
o o O) CD O) O) CO CD O) ÇO
g> $ 8 8 8 8 6 8
d O O >
C
CO CO
s 8 S
o> 2 2 CO CO
8 8 8 CO O o
§ g 8 8 8 8 K 8
$ 9
3 S 8 8 8 8 s 8
5 o d
N- CM
8 R ? ? 9! 9 Si
2 d CO
I - S 8 9 9 d d
CO in co
8 K
9
o d d
CMCM<9 to O) o> CO CO CO It CO
CMCMCO CO 8 8 CM ” O)
!
CO
$ 8 <D to 8 8 q
CO CO CO CO 1 q 1
i 8 8 o o o
co ? CMC? M s 1 8
d o d
o>
I 1 1
s
o o d o
I
to
1
s a s 8 s q 8
o d o d o o o d
I
1
S 8 C CO O CO N CT> C CO O 00 C COO \
O) CO CMco q to to
o o O o d d
i r d
to N to o> 05 CO o o
CO CMCO to 00 CMO) O to
q q CO CO
d d o o o d
ca
t—CM ■5 TOa
X oI X oI CM8 9
CM CM O O II | JC £ E a £
CO CO o o
Q_ o o o o <y y
ETAR
NOITCUDER
ETAFLUS
F O
SISYLANA
EVITARAPMOC
EHT
M
ORF
ATAD
W
AR
EOS
52.2
3.05
52.2
05
99S
05
66.5
66S0
8.24
387.0
54
1.32
336.0
54
1.32
207.0
|
Colorado School of Mines
|
92
1 a
a vq «n ts co mvi o\o>t^-ONVommcNO\CT\oooooooovoinrtTi-
« 3 rn rn rj cj <d ' ( S i n i n m m t N c s c s c s m m ^ ' - i m m
S s
L
«nrn--cKvd^gOvO4 ) <mON N<N NO *n P NO NO NO O
CS CN NO r- CN CN
p o o
m oo on oo • oo p CO (S (N ON o o NO P ON ON 00 OO
6 Oo N Om O O'd O* cq 2 m CK O o 2 2o 2o O ON N § g Oo( n g ON NO ONO n OO NN OO NN
NO ON m NO m ^
1 H3 00 O5 O O OO n 0^ o 0 r O< - OS ^ ( Oo no n o Ornn N r«—n i tn <od s rO~n ON>7 C K OO n N o OO d n O OoO d n o Od N O ONn N OO N 0 O 00 0N C OKO n O OO N
a
vO *3 <N ON CN r~ 00 OO NO NO NO NO ON P
^ .2F od od 2 2 ^ <n o
ts
C CNS C CN
N
r5 N CNO CN NO On o
CN
nJ
Bi B l a 0“0 KO p On r o~ N TO f VO ÔN o NO p OO 1 r— 5, 0 00 0 C oO d N OO n p ON Tf N 'fO ' Nt O NO 00 O o NO ON o
CN CO CO CO CO «n V) <N <N NO NO m •o NO NO NO NO 't NO NO
: :
4 o NO NO NO NO CO OO <N r- CO t*. CO CO CN NO 't
t-5 r5 CN o o NO NO m rn NO 'T CO CO o O
rt Z CO NO NO CN CN CN CN
■ ^ ^ 5 3 e e ^ e = ! < s o o ^ < 3 o i - 020 300 ^1—H CO NO O
4 CO Tf CN p5 NO
CN CN
-h' r~- c P CO ON
Ii C CN O C'd O* Jjjr NO 'o d- Co Od r-" CO o o ' o z 2 « > o ' 2 2 S S r- no
£
1 JI3 i mn m (N m -I V o) N 0O 0 oOO O oO O o o •c n O•c sr • ~ ON • o •- < t ^N o •oo •N 1-n •H ' •4- O•N O • '« • tr •^- -4 r •F~ —I
I
\
i i O. •<g ^On oO VfdN N«>O o^ t^- <
m•
-s —c
••
s
:^
'Q oo
•
or
^•
^
^
'<
^
^r -•^ tO •s' •oi t
o•
o o•o oo
•
oo
•
' ^t
•
c« •sn
^ fS CN —'
a «5 CN O 4) — O| k C CN L, ■— Mi ^ **H —1 CN 1 1 1
« «5
sa 273 2cd l j l l l i l l i l l j j j j
e* < < I l i i a
CO
SESYLANA
ARNCSDN
EHT
M
ORF
ATAD
W
AR
|
Colorado School of Mines
|
98
SORPTION DATA
sample mg/L mg/L sample mg/L mg/L
alfalfaO 1.18 0 mwO 0.01126 0
a25 14.51 25 mw25 16.3682 25
a50 29.83 50 mw50 37.9071 50
a75 46.66 75 mw75 60.3713 75
a125 78.31 125 mw125 103.994 125
a175 122.03 175 mw175 147.82 175
AFO 0.25149 0 mwO 0.01859 0
AF25 15.2054 25 mw25 31.0398 50
AF50 32.1442 50 mw50 56.4172 75
AF75 50.9551 75 mw75 95.7102 125
AF125 92.3441 125 mw125 151.685 175
AF175 139.843 175 mw175 214.723 250
beerwasteO 0.22 0 oakO 0.02583 0
bw25 15.81 25 ow25 18.3125 25
bw50 36.34 50 ow50 39.4145 50
bw75 57.59 75 ow75 62.0704 75
bw125 106.027 125 owl 25 106.627 125
bw175 153.834 175 ow175 148.616 175
bwO 0.3333 0 oakO 0.01948 0
bw25 17.5194 25 oak25 33.5973 50
bw50 34.7785 50 oak50 61.4151 75
bw75 65.1372 75 oak75 95.4064 125
bw125 101.053 125 oak125 158.566 175
bw175 165.631 175 oak175 215.031 250
sugarbeetO 0.15009 0 pineO 0.07634 0
sb25 10.0758 25 pi25 16.4004 25
sb50 23.0974 50 pi50 36.119 50
sb75 37.371 75 pi75 57.9171 75
sb125 68.8453 125 pi 125 102.123 125
sbO 0.16296 0 pi 175 140.41 175
sb25 10.3317 25 piO 0.05954 0
sb50 22.5134 50 pi10 5.06154 10
sb75 39.4116 75 pi10 5.03738 10
sb125 73.6317 125 pi25 16.4459 25
sb175 107.839 175 pi50 38.1631 50
pi75 57.7331 75
pi 125 114.416 125
8 8 8 8
|
Colorado School of Mines
|
99
ce co ce co
sample mg/L mg/L sample mg/L mg/L
corncobO 0.04964 0 poplarwO 0.0656 0
cc25 6.75906 25 pow25 15.8556 25
cc50 8.4015 50 pow50 37.1022 50
cc75 9.9946 75 pow75 59.1627 75
cc125 19.3042 125 pow125 102.708 125
cc250 111.395 250 pow175 151.338 175
ccO 0.15875 0 powO 0.06876 0
cc25 4.58293 25 pow50 26.7933 50
cc75 2.492 75 pow75 50.177 75
cc125 3.97091 125 pow125 87.0872 125
cc175 2.6224 175 pow175 142.766 175
cc175 125.337 175 pow250 200.641 250
wetlandinocO 0.01274 0 walnuthullO 0.01386 0
I25 1.50203 25 wh25 1.13406 25
I50 8.44349 50 wh50 3.5092 50
I75 26.5312 75 wh75 11.789 75
1125 54.3545 125 wh125 45.0359 125
1175 92.3861 175 wh250 146.258 250
iO 0.02281 0 whO 0.00944 0
I25 8.08146 25 wh25 1.38329 25
ISO 6.85953 50 wh50 4.32762 50
I75 14.4907 75 wh75 12.6231 75
1125 39.1644 125 wh125 41.4531 125
1175 76.3525 175 wh175 82.8785 175
wh175 79.8778 175
manureO 0 0 walnutwoodO 0.00595 0
m25 1.99999 25 ww25 16.6897 25
m5O 7.0097 50 ww50 38.2598 50
m75 5.27752 75 ww75 59.1026 75
m125 19.677 125 ww125 108.28 125
m175 41.2525 175 ww175 161.49 175
mO 0.20462 0 wwO 0.00831 0
m25 8.63639 25 ww10 5.38441 10
m50 16.7542 50 wwlO 5.42597 10
m75 26.3294 75 ww25 16.4585 25
m125 39.8605 125 ww50 37.6662 50
m175 48.1014 175 ww75 60.7105 75
m250 88.5387 250 ww125 107.325 125
m250 91.0355 250
|
Colorado School of Mines
|
ABSTRACT
The final slope design of the West Wall has become the primary importance for
the economy of Chuquicamata Open Pit Mine. It determines the safety of the operation
and, consequently, the economic viability of the mine. Although considerable progress
has been made in the field of rock mechanics applied to rock slope stability, the estima-
tion of rock mass strength poses difficulties. Realistic estimation rock mass strength be-
comes even more critical when joint sets have a dominant influence on the behavior of
the rock mass.
A back analysis of slope failure cases using a statistical approach for estimating
strength parameters of discontinuities in the West Wall is described. The statistical tech-
nique known as maximum likelihood estimator method is used for the analyses. The main
advantage of this method is that it allows incorporation of both failed and unfailed cases
into a back-analysis, thus increasing the accuracy of parameter estimation. The approach
requires an assumed statistical distribution for the safety factor (FS) and it is assumed to
be lognormal distribution. The estimated values of cohesion and friction angle for plane
failure are 20.7 kN/m2 and 35.3º, respectively. The estimated values of cohesion and fric-
tion angle for wedge failures are 15.22 kN/m2 and 35.1º, respectively. The mean and
standard deviation for the factors of safety are 1.005 and 0.099 in plane failures, respec-
tively. The mean and standard deviation for the factors of safety are 1.016 and 0.179 in
wedge failures, respectively.
In order to estimate the rock mass strength and deformability, the approach sug-
gested by Hoek-Brown (1997) is used. The Hoek-Brown failure envelope is translated to
a linear Mohr-Coulomb envelope to provide input to the numerical models. Parameters m
and s (m and s are constants which depend upon the geological characteristics of the rock
mass) in the Hoek-Brown criterion were calculated assuming disturbed rock mass condi-
iii
|
Colorado School of Mines
|
tions. Cohesion and friction angle values for the rock mass compare well with the ob-
served behavior on in-situ rock mass indicating that the method is satisfactory for
strength estimation.
The performance of the current and alternative slope geometries is evaluated us-
ing Finite Element Method and Discrete Element Method numerical models. The com-
puter input parameters are based on the geotechnical data obtained from field measure-
ments at the mine. To evaluate the factor of safety, the shear reduction technique is used.
Since the factor of safety is defined as a shear strength reduction factor, this is computed
with a numerical code reducing the rock mass shear strength until collapse occurs. The
resulting factor of safety is the ratio of the rock’s actual shear strength to the reduced
shear strength at failure. The factor of safety and probability failure for a variety slope
geometry options are calculated by SLOPE1 (Smith and Griffiths, 1998) and UDEC
(Itasca, 1999) computer codes.
A new slope geometry is proposed in order to improve the stability of the West
Wall after the final pushback. This new slope geometry is based on the concept of mini-
mizing the load acting over the shear zone. The study shows that 250-m wide platform
over the shear zone and increased the interramp angle to 44º allows an overall slope angle
of 33º in the West Wall with a factor of safety of 1.45. With these changes, it may be
possible to continue mining down to depths of 1000-m.
iv
|
Colorado School of Mines
|
Figure 4.3 Factor Safety Formula for a Wedge Failure …………………………. 66
Figure 4.4 Linear Regression of Laboratory Test ……………………………….. 68
Figure 4.5 Values of Cohesion Back-calculated from Failed Cases versus Fric-
69
tion Angle Estimated from Laboratory Tests …………………………
Figure 4.6 Efficiency of the Statistical Approach of Estimating Factors of Safety
for Different Standard Deviation ………………………… 71
Figure 4.7 Results Obtained Using Plane Failures and Limited Maximum Like-
lihood ………………………………………………………………… 75
Figure 4.8 Shear Strength and Shear Stress Acting on the Plane Relationship for
the Plane Failure Cases ………………………………………………. 76
Figure 4.9 Results Obtained Using both failed and Unfailed Cases for Plane
Failure Formula and Utilizing Full Maximum Likelihood …………... 76
Figure 4.10 Shear Strength and Shear Stress Acting on the Plane Relationship
Failed (+) and Unfalied (o) Cases ……………………………………. 77
Figure 4.11 Results Obtained Using Wedge Failures and Limited Maximum
Likelihood ……………………………………………………………. 78
Figure 4.12 Shear Strength and Shear Stress Acting on the Planes Relationship
for Wedge Failure Cases ……………………………………………... 78
Figure 4.13 Results Obtained Using both failed and Unfailed Cases for the
Wedge Failure formula and Utilizing Full Maximum Likelihood …… 79
Figure 4.14 Shear Strength and Shear Stress Acting on the Planes Relationship
for Wedge Failure Cases, Failed (+) and Unfalied (o)………………... 79
Figure 5.1 Cross-Section Along the Coordinate 4200-N ………………………... 87
Figure 5.2 Failure Mechanism Controlling the Interramp Slope Design of the
West Wall …………………………………………………………….. 89
Figure 5.3 Relationship between Slope Angles and Slope Heights for the Inter-
ramp Slope of the West Wall ………………………………………… 91
ix
|
Colorado School of Mines
|
Figure 5.4 Proposed Interramp Slope Design for the West Wall ………………... 91
Figure 5.5 Slope Geometry Option 1, and Continuous Slope …………………… 93
Figure 5.6 Slope Geometry Option 2, Platform 200-m Wide on the Shear Zone .. 94
Figure 5.7 Slope Geometry Option 3, Platform 250-m Wide on the Shear Zone .. 95
Figure 5.8 Deformed Mesh Plot and Factor of Safety Corresponding to the Op-
tion 1 Solution with FS = 1.35 ……………………………………….. 98
Figure 5.9 Deformed Block-mesh Plot and Factor of Safety Corresponding to
the Option 2 Solution with FS = 1.60 ………………………………... 99
Figure 5.10 Deformed Mesh Plot and Factor of Safety Corresponding to the Op-
tion 3 Solution with FS = 1.60 ……………………………………….. 100
Figure 5.11 Deformed Block-mesh Plot and Factor of Safety Corresponding to
the Option 3 Solution with FS = 1.45 ………………………………... 104
Figure 5.12 Deformed Block-mesh Plot and Factor of Safety Corresponding to
the Option 3 Solution with FS = 1.40 ………………………………... 105
Figure 5.13 Deformed Block-mesh Plot and Factor of Safety Corresponding to
the Option 3 Solution with FS = 1.45 ………………………………... 106
x
|
Colorado School of Mines
|
AGKNOWLEDGEMENTS
I wish to express my gratitude to Dr. M. Ugur Ozbay, for his guidance and
support given during the preparation of my thesis. I wish also the acknowledgment the
help and guidance provided by the remaining members of my thesis committee: Dr. Tibor
G. Rozgonyi, Dr. D. Vaughan Griffiths and Dr. Kadri Dagdelen.
Special appreciation is expressed to Juan Rojas and Germán Flores of the
Chuquicamata Mine of CODELCO-CHILE for their support and encouragement during
my studies. Also I would like to thank my friends Milko Diaz, Manuel Contreras,
Alessandro Tapia, and Ricardo Torres for their help and cooperation throughout this
study.
Mahesh Vidyasagar and Mike Brewer deserve special thanks for their help
reviewing this thesis, support and friendship. Special thanks to many fine fellow graduate
students have given advice and help, but names will not be mentioned at the risk of
excluding anyone.
Finally, special gratitude to my wife Roxana and my daughters Daniela and
Alexandra for their support, assistance, patience, sacrifice and love, and most of all,
thanks to God for the strength, which allowed me to achieve this important goal in my
life.
xii
|
Colorado School of Mines
|
1
CHAPTER I
INTRODUCTION
The primary objective of this work is to design the ultimate pushback on the west
wall of the Chuquicamata Open Pit Mine. To achieve this objective the following tasks
are set:
1. Rock mass strength parameters are estimated using back-analysis and statisti-
cal approaches.
2. Different pushback designs are evaluated using numerical modeling based on
estimated strength parameters.
The West Wall is the largest and flattest of all the walls in the pit, where each de-
gree of variation in the final pit-slope angle represents approximately 200 million tons of
waste rock. The relatively flat slope angle is required due to the large-scale instability
that is concentrated on the west wall. This instability became a serious problem for the
mining operation in 1979. Since that time, the West Wall has shown deformations reach-
ing several meters per year.
The failure mechanism involves a quasi-stable toppling within the upper 100-m of
the slope. It is believed that this failure mechanism is caused by the weak shear zone
present near the current bottom of the wall. This shear zone is compressed in response to
load applied from the upper part of the slope, and is repeatedly squeezed upwards into the
pit bottom. Two conspicuous joint sets in the west wall appear to be important in
controlling large-scale slope behavior. The first set dips approximately 70º west and
strikes in a north-south direction. This set of joints control a toppling mode of failure in
|
Colorado School of Mines
|
2
the wall. The second set dips 35º to 45º into the pit, also striking north-south direction.
This set of joints tends to affect the depth of the rock mass movement.
The failure mechanism described above has resulted from years of geotechnical
investigation at the mine. Using the large amount of field data collected by the
geotechnical department at the mine was able to identify important details that have
helped in the development of a conceptual model of the failure mechanism. The field data
have also been used for validating and calibrating the numerical models used for slope
stability analysis at the mine.
The geotechnical constraints become even more relevant as the pit expands and
deepens. The 1999 Life of Mine Plan presents a scenario in which the pit’s depth will
exceed 1000-m over 15 years. With this scenario, the major concern is the stability of the
West Wall, taking into account the present instability affecting this slope. Currently, the
west wall is 750-m deep and displacement measurements taken so far show that the wall
moves at rate of 4 to 5-m per year. During the 1980’s and early 1990’s, the recommended
design for the west wall was to reduce the slope angle while deepening the pit. However,
this classical judgment in slope design has not worked on the west wall. The slope angle
has been reduced a number of times, yet the movement remained persistent.
Therefore, significant demands are placed on geotechnical analysis. The primary
requirement is to establish an optimal slope design for the ultimate pushback on the West
Wall, in which safety and economic concerns are satisfactory. In order to have an optimal
slope, a good set of rock mass strength properties are needed. To accomplish these goals,
it was decided that using back analysis as a tool for verifying rock strength would provide
reliable information which could then be used in numerical analysis.
In this study, representative failure cases have been collected and have been used
for estimating the strength parameters of the discontinuities. This analysis was done by
means of back-analysis, combined with a statistical analysis which is discussed in
Chapter 4. The maximum likelihood statistical technique was used to estimate cohesion
and friction values of a dominant discontinuity set in the West Wall. Further, statistical
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.