Machine Monitoring Sensing and Robotic Control of a Mechanical Fragmentation Machine by Naji Aboufadel A Thesis submitted to the Department of Mining Engineering in Conformity with the requirements for the degree of Master of Science (Engineering) Queens University Kingston, Ontario, Canada July 1997 Copyright 0 Naji Aboufadel, 1997
133
Embed
Machine Monitoring and Robotic Control Aboufadel A Thesis submitted ... MACHINE MONITORING SENSiNG AND ROBOTIC CONTROL OF A MECILANICAL
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Machine Monitoring
Sensing and Robotic Control
of a Mechanical
Fragmentation Machine
by
Naji Aboufadel
A Thesis submitted to the Department o f Mining Engineering
in Conformity with the requirements for the
degree of Master of Science (Engineering)
Queens University
Kingston, Ontario, Canada
July 1997
Copyright 0 Naji Aboufadel, 1997
National Library BiMiotheque nationale du Canada
Acquisitions and Acquisitions et Bibliographie Services services bibliographiques
395 Weilington Street 395. rue Wellington OtïawaON K 1 A W Ottawa ON K1A ON4 Canada canada
The author has granted a non- exclusive licence allowing the National Library of Canada to reproduce, loan, distribute or sel1 copies of this thesis in microfom, paper or electronic formats.
The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
L'auteur a accordé une Licence non exclusive permettant à la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la forme de microfiche/film, de reproduction sur papier ou sur format électronique.
L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de ceîleci ne doivent être imprimés ou autrement reproduits sans son autorisation.
ABSTRACT
MACHINE MONITORING
SENSiNG AND ROBOTIC CONTROL
OF A MECILANICAL
FRAGMENTATION MACHINE
The monitoring and sensing of a Continuous Mining Machine (CMM) and its interaction
with the rnining environment is studied. Vanous aspects of the machine design and
material handling requirements are critically examined. Correlations between rock-mass
properties and machine variables are developed for automatic sensing of rock properties
based on Specific Energy principles. The tests and analysis show potentid benefits fiom
monitoring of machine variables for improving the cutting performance of the machine and
increasing its production rates.
1 would iike to express my sincere appreciation to Dr. L.K. Daneshmend of the Mining
Engineering Department at Queens University for his extreme patience, guidance and
support in wnting this research thesis.
Funding for this research was provided by the Federal Government under auspices of the
National Networks of Centres of Excellence Program: Specifically the Institute for
Robotics and Intelligent Systems, under prcjject ISDE - 1 .
HDRK Mining Research Ltd., Onaping, Ontario, provided access to technical information
and data on the CMM. 1 would like to thank Jeff Repski fomerly of HDRK, for his
assistance in this regard.
TABLE OF CONTENTS
PAGE -
ABSTRACT
ACKNOWLEDGMENT
LIST OF F I G U E S
LIST OF TABLES
LIST OF GRAPHS
Chapter 1
Introduction
Historical Overview on Mechaniration
of Rock Tunneling in Mining
Mechanical Excavation as a Substitute
for Drill and Blast
Mobile Excavators and Tunnel Development
in Minkg
New Mobile Excavator Technology For
Cutting Hard Rock
Scope of the Study
iii
Chapter 2
Literature Review
Other Mining Mechanical Excavation Technologies
2.1 . 1 Tunnel Boring Machines
2.1.2 Heavy-Duty Roadheaders
2.1.3 Raise Boring Machines
2.1.4 Mobile Excavators
Introduction To Machine Monitoring
Applications in Mining
Machine Sensing of Rock Properties
in a Dynamical Mining Environment
Machine Automation and Control
Cha~ter 3
Robotics and Control Issues of the
CMM Machine
3.1 An OveMew of the CMM as a Mining Robot
3.2 Machine Kinematics and Dynamic Behavior
3.2.1 Definition
3.2.2 Machine and Controller Inputs
3 .Z.3 Arrn Kinematics
PAGE -
3.2.4 Arm Dynamic Disturbances / Extemal
3.2.5 Arm Dynamic Disturbances / Intemal
3.3 Relationship Between the Control System
and Machine Behavior
General
Machine Controller Functions and
Design Configuration
Controller Limitations
System Responsiveness - Relation Between
the Controller and Time Delays
Ratio of (Peak PressurdAverage Pressure)
versus Controllability
3.4 Actual Machine Profile Generation
3.4.1 Effects of Disturbances on System Response
3 A.2 Machine Guidance and its Effects on Arm
Motion and Profile
PAGE -
Cha~ter 4
CMM Cutter Design, Fragmentation and Materials
Bandling Requirements
4.1 Introduction
4.2 Effects of Muck Removal on Mining
Excavation Machines
4.3 Muck Removal Modes for Tunnel Construction
4.4 Machine Supporting SeMces and Muck Transport
4.4.1 Auxiliary Equipment
4.4.2 Basic Muck Removal Techniques
4.4.3 CMMConveyor Systemversus
Ot her Haulage Systems
4.5 Muck Chip Size and Form Based on the CMM
Performance in Herdecke, Germany
4.5.1 RockConditionsat theHerdecke Site
4.6 Effects of Cutter Design on Machine Performance
CMM Cutter Design Based on TBM's
Cutter Performance
Effects of Increasing 1 Decreasing Disc Parameters
on Machine Design
The Theory Behind Cutter Wear
CMM Cutter Life Estimation
PAGE -
Cha~ter 5
Erperimental Data and Preliminary Analysis
5.1 Hardware and Software Instrumentation
5 .1 .1 CMM Signal Selection and Identification
5 . 1 .2 Hardware
5.1 .3 Software
5.1.4 Digitizing Procedure
5.2 Filtering
5.2.1 Noise
5.2.2 Filtering in Viewdac
5.2.3 Filtenng in Matlab
5.3 Processing of the Digitized Data - Initial Conclusions
5.3.1 Processing Methods
5.3.2 OveMew on Pattem Recognition
5.3.3 Pattern Recognition in Relation to
Sensing of Rock Properties
5.3.4 Specific Energy - Profile
5 .3 .5 Specific Energy Computations based on
the CMM Machine Design and Operation
PAGE -
vii
Chapter 6
Application of Specific Energy for Sensing of
Rock Properties
Relationship Between Specific Energy and Compressive
Strength of Rock
OverMew of Specific Energy Investigation
and Machine Sensing of Rock Properties
Specific Energy Computations
6.3.1 Quany Site, Herdecke, Germany
6.3.2 Creighton Site, Sudbury, Canada
Conclusion
Cha~ter 7
Conclusions and Future Work
7.1 Conclusions
7.2 Suggestions For Specific Future Work
REFERENCES
APPENDlX A
PAGE -
83
LIST OF FIGURES
FIGURES
I l .
DESCRIPTION
CMM Machine
C M . Machine Cutting Profile
TM60 Machine
Alpine-Tunnel-Miner (ATM 1 O 5 )
Robbins Mobile Miner (MM 1 30)
Details of a Design-Test loop
Building Blocks and Links Between the
Vanous Aspects of the Machine D u h g Testing
Elements of the Arms Feedback Control Loop
Arm velocity components while extending
Cornponents of Reaction Forces from the Rock
(Fr is the Resultant of Fx & Fz)
(a) Disc parameters and @) Contact area
with Rock
Result of a Typical Positioning Error on the Profile
Division of the Face Cut into Sub-areas
Effect of Hydraulic Delay on Aîm Position
PAGE -
FIGURES DESCRIPTION
Resulting Efféct of the Delay on the Arm Position
while Extending (Under Penetration)
Discl ring Panuneters
Actual Cutter Wear
Block Diagram for Recording and Digitizing Data
Slock Diagram for Pattern Recognition
and Classification
CMM Machine Dynarnics & Kinematics -
h s II Cutting Profile
Specific Energy Parameter Computations Based
on CMM Machine Dynamics and Kinematics
Face Map for Cut # 93072004
(Sandstone and Shale)
Face Map for Cut # 94 12080 1
(Nonte and Granite)
PAGE -
53
LIST OF TABLES
TABLE DESCRIPTION - PAGE
Arms II proportional gain settings 44
Ratio of time delays caused by a decrease 46
in the Kp and Ki gains of the controller
Arm 11 2 typical piston pressure ratio 47
peau average
Profile shifi due to hydraulic delay for
dlfferent RPM values
Quarry Sequences used for
Specific Energy Anaiysis
Creighton Sequences used for
Specific Energy Analysis
GRAPH
LIST OF GRAPHS
DESCRIPTION
Specific Energy versus Time
for Cut #9307 1903
Specific Energy versus Tirne
for Cut #93072202
Specific Energy versus Time
for Cut #93072004
Specific Energy versus Amis II
Radial Position for Cut #93072004
Specific Energy versus Amis II
Angular Position for Cut #93072004
Zone (1700 mm to 1800 mm)
Specific Energy versus Amis II
Angular Position for Cut #93072004
Zone (2500 mm to 2600 mm)
Specific Energy versus Time
for Cut #94 12070 1
PAGE -
xii
GRAPH
8.
9.
DESCRIPTION
Specific Energy versus Time
for Cut #94 120702
Specific Energy versus Time
for Cut #94 12080 1
Specific Energy versus Arms II
Radial Position for Cut #94 12080 1
Specific Energy versus A r m s II
Angular Position for Cut #94 12080 1
Zone (2 100 mm to 2200 mm)
PAGE -
99
1 O3
xiii
Chanter 1
Introduction
1.1 Historieal Overview on Mechanization o f Rock Tunneline in Mining
Mechanization of rock tumehg has been progressing since 1846 when Henri-Joseph
Maus mounted a group of mechanical rock drills on a frame to speed the excavation of the
Mount Cenis tunnel between Italy and France. Drill and blast methods are progressing
today with high speed hydraulic powered drills and cornputer controlled jumbos, but
drilling a tunnel continuously without the cyclic blasting has obvious advantages. The
challenge has been to develop a machine and cutting tools capable of the job.
mobbins, 19951.
In 185 1 an Amencan engineer, Charles Wilson, developed a machine which was to
become the first successfùl continuous tunnel borer for rock. However, many problems
with disc cutter tools and other difficulties made it uncornpetitive with the developing
techniques of drill and blast tunneling.
Other attempts were made and abandoned both in the US and in Europe fiom the late
1850's to about 1920 when mechanical tumeling in rock seemed to be gradually
appearing. Practically no senous attempts were made for a period of 30 years until the
1952 developments by James Robbins which combined drag pick cutters with rolling discs.
These early attempts were also unsuccessful in hard rock until the building of a machine in
1956 which used only rolling disc cutters for cnishing the rock in a manner similar to
Wilson's design of 100 years earlier. During the next thirty years more than 200 rock
tunneling machines were used successfully in soft and medium hard rocks and more
recently in very hard rocks.
1.2 Mechanical Excavation as a Substitute for Drill and Blast
The mechanical excavation industiy is continually developing as machines are becoming
more versatile and capable of excavating through any type of ground conditions. The
economics of mechanical excavation compared to drill and blast techniques have been
steadily improving for both the civil underground construction and mining. Continua1
improvements in the understanding of rock fragmentation principles allowed for more
efficient machine design and operation which served to achieve further productivity
increases. An example on this is the introduction of new mechanical excavation
technologies such as Mobile Mining Machines. These machines attack rock a portion of
the tunnel at a given time whereas other mechanical excavation technologies such as
Tunnel Boring Machines (TBM's) attack rock full face. It is therefore believed that
current and future anticipated developments in mechanical excavation technology will
make it the "conventional" technique of the future.priant, 19951.
Mechanical excavation offers numerous advantages over drill and blast methods for al1
types of underground constmction, including tunnels, raises, shafts and chambers.
The first and most important issue in a mine is safety. By eliminating the use of blasting,
the personnel safety is greatly improved. There is no handling of blasting agents and no
generation of toxic fumes. Machines create practically no ground vibrations which anse as
an important factor in urban construction. In fact, drill and blast is not allowed in most
urban areas around the world simply due to the vibrations created from the blasting
operation which could potentially cause structure damage, both surface and underground.
Secondly, blasting imparts ground disturbance and damage which increases ground
support requirements to provide a safe and stable opening. With machine excavation,
ground disturbance is practically nonexistent which results in significantly reduced support
requirements and enhanced opening stability. In fact, the ground support savings provided
by the use of mechanical excavation techniques can alone justih their application. The
smooth excavation walls created by machine bonng also mean reduced ventilation
requirement S.
Unlike drill and blast, machines generate a unifonn muck size which allows for the
implementation of continuous material handling systems, such as conveyor belts with a
substantial improvement in machine utilization and the overall productivity of the
excavation system. Pei1 et al., 19941.
One of the more important advantages of mechanical excavation is that the machines are
conducive to remote control or full automation, which is a growing trend in the
underground construction and mining industry in order to remove workers fiom
hazardous environrnents and enhance safety. On the other hand, a drill and blast operation
is difficult, if not impossible, to automate due to its cyclical nature. [King, 19881.
The last significant advantage offered by mechanical excavators is their potential to
achieve high rates of advance at reduced excavation costs compared to drill and blast.
Despite being a relatively new technology, it is not uncommon for mechanical excavators
to attain advance rates several fold higher than what the drill and blast systems are capable
of achieving. Al1 of these advantages have thus contributed to a wider use of mechanical
excavation in rock. [Ozdemir, 19901.
1.3 Mobile Excavators and Tunnel Develo~meat in Mining
The circular profiles produced by tunnel bonng machines (TBM's) are well-suited for
most applications. However, there are cases where a non-circular opening with a Bat floor
is preferred, such as for highway tunnels and mine drifts. In the past, this has been
accomplished on several occasions by removing the boaom comers of a machine bored
tunnel with drill and blast techniques. Designs have also been developed for mounting
wheel type excavators fitted with disc cutters at the afl section of the TBM for cutting out
the comers to provide a more or less horse-shoe shaped opening. One major difficulty in
these applications has been effective dust control fiom the secondary cutting operation.
Passner and Sklansky, 19801.
Various types of hard rock mobile excavators such as the Continuous Mining Machine
(CMM) are cumntly under further development. In the underground mining industry,
there is a great need for mobile excavators which can efficiently excavate openings of
vatious shapes and sires in hard rock while incorporating sufficient mobility for easy
relocation within different working headings of the mine. Similar needs also exists for civil
underground constmction where short tunnels or chambers need to be excavated using a
mobile piece of equipment. Such applications are not suited for TBM's because of their
lack of mobility and the circular cross-section which they produce. To address these
needs, extensive efforts have been done worldwide for the development of mobile hard
rock excavators.
One of the more exciting new developments in the mining automation area is the
advancement towards automatic steering of mobile excavating machines. In addition, the
automatic optimization of the machine performance in response to changes in rock and
ground conditions is now being investigated. [Belanger, 19901.
1.4 New Mobile Excavator Technolon;v For C u t t i n ~ Bard Rock
The innovative approach of developing mechanical excavation equiprnent as a substitute
for the cyclical nature of drill and blast served as a need for the mining industry to develop
a powefil technology suitable for rnining hard rock. The result was a multi-phase
research project that was carried out under Hard Rock Mining Research Limited p R K )
for the development of the Continuous Mining Machine (CMM) project. See Figure [l].
CMM Cutting Pnnciple 2
Control Room 4
Subsequent Full Face Cuts
Pivot ( Rotor
(a) CMM - Side View
Disc Cutter
(b) CMM (Rotor & Arms) - Front View
Crawler
Cutter
(Provides
of Torque
(c) CMM Arms II (II 1, II 2 & II 3) - Side View
Figure [l]: CMM Machine
As a joint venture between INCO, Noranda, Falconbridge and Wirth (Germany), the
HDRK CMM project was organized and carried out under a clear mandate to establish an
efficient principle for cutting hard rock. In order to achieve this goal, the CMM was built
and tested in both Germany (Soft Rock) and Canada (Hard Rock) with the main objective
of ascertainhg the relative importance of a sound project foundation that would lead to
the successful development and performance of the machine. pepski, 19951.
The excavation process of the CMM machine is accomplished by the use of four
hydraulically activated cutting arms fitted with large diameter disc cutters. The arms are
hinge-mounted on a rotating support plate. As the plate rotates, the arms are swung out to
create spiral cuts to break the rock. The center am creates a "pilot" hole by slewing
towards the cenier of the bore. Once this Free face is created, the other three arms start
cutting spiral tracts outwards toward the edges. This provides an undercutting action
which is a highly efficient way of breaking rock. When the three arms reach the maximum
inner circular profile of the tunnel, they can be extended farther to cut the desired final
shape of the opening. See Figure [2]. This is al1 accomplished under cornputer control,
allowing for excavation of different site and shape openings.
Throughout the testing phase of the machine in both Gemany and Canada, a rigorous
approach to the engineering development process was incorporated, including a
comprehensive program of data collection, evaluation, etc. This ngorous approach served
in the advancement of the technology which was manifested in the cutting ability of the
CMM machine in both sofi and hard rock.
Front View of Cut
/ Edge OfCUt
Disc Cutter Maximum Inner Circular Profile Corner Cut
Arm II3 - Disc Cutter
---\ \j Disc Cutter
Ann II2 Disc Cuttei
\ 7 /tC,cut Portion of Rock Face (Center of Bore)
\ Start Positions v
Maximum Outer of Arrn 1 and Circular Profile of *' II Il2 and n3) Arms (11 1, 112 and 113)
However, the CMM machine is considered to be a proof of concept unit despite the fact
that HDRK has achieved its initial objective of designing and implementing a novel
machine design for hard rock cutting. Nevertheless, it still has lots of potential. It is
expected within the next five years to more than double its performance and have
reliability figures comparable to mature mechanical excavation technologies. Moreover,
many of the modifications and advancements in the CMM technology would serve to
improve the performance of the machine in areas such as continuous cutting, consistent
The CMM optimization greatly depended on trade-offs between the technical aspects of
the machine and the economics of it. See Figure [7].
Positioning Accuracv of Arms
Production Rates -
Figure [7]: Building Blocks and Links Between the Various Aspects of the Machine
During Testing
-
-
Figure [7] shows the links between the main six building blocks that characterize the
CMM design successfully. These blocks include:
t
Cutting Variables
Depth of Cut Penetration
RPM
i
Motion Control
System Stifiess Position Error
Hydraulic Delays
t
Cont roller Movement Gains - of
Kp, Ki, Kv, Kd VPIVCS m o l
(1) The CMM user-input parameters such as depth of cut, penetration, head RPM
factors and individual controller gains
(2) The cutting am's kinematic parameters such as profile position, velocities and
accelerations
Resulting - Forces on Arms
-
+
Machine Variables
Ann Pressures Head Torque
Oil Flow
The CMM limitation parameters such as head power (or torque), arm hydraulic
pressure and oil tlow
The CMM main design characteristics such as the ami dimensions (lever effect
and hydraulic stifiess), and disk variables such as outer diameter, tip radius
and angle of attack
The CMM controller aspects such as its user input gains and the resulting arm
stifbess I responsiveness, and profile error reduction
The economic factors that will ultimately decide the optimization of the machine in its current set-up and what could be done to make it more efficient keeping in mind the following key parameters:
(i) Achievable profile accuracy
(ii) Production rates
(iii) Wear rates of cutting disks
(iv) Overall machine reliability
Machine Kinematics and Dvnamic Behavior
3.2.1 Definition
Kinematics is the science of motion which treats motion without taking into consideration
the forces that cause it. [Craig, 1 9861. Thus, the study of the kinematics of the CMM
machine refers to al1 the geometrical and time based properties of the motion such as
position, velocity, acceleration, and al1 higher order derivatives of the position variables.
On the other hand, the study of dynamics deals with the relationships between these
motions and the forces and torques that cause them, which in our case is essential for any
dynamic analysis that would be perfonned on the CMM machine.
3.2.2 Machine and Controller Innutg
The user input machine parameters such as RPM, penetration and depth directly affect
the resulting forces on the arms and thus influence the head torque and a m pressures. In
retum, these parameters will affect production rates achieved by the machine since lower
depth and penetration means less production per cut.
On the other hand, the resulting effect of the input controller gain settings is to infiuence
the positioning accuracy of the amis by reducing as much as possible positioning errors.
This is achieved by moving the valve spool dong its range. Any movement of the valve
results in a change in oil fiow and pressure to the arrn cylinder. The precision of the a m
response to the movements of the valve depends on the combined effects of interna1 and
extemal disturbances such as reaction forces fiom the rock as well as fiction and the
accuracy of the valve.
The gains selected by the operator also affect the rate at which the valve spool responds.
Controller gains must be carefully selected since oscillations resulting in over-penetration
of the am, or sluggish am response resulting in under penetration, can occur. Therefore
the exact response of the a m will depend on the combined effects of the am dynamics,
the valve dynamics, the controller transfer function and the above mentioned intemal and
extemal disturbances. The diagram shown in Figure[â] shows two major blocks; The
control elements which consist of the PID-F controller and the a m position feedback
device (a transducer that measures actual cylinder position), and the system under control
which consists of the servo-valve and the ann-cylinder assembly. As for the disturbance
variable, it refers to the m-valve-cylinder assernbly.
e = Position Error x = Actual Position r = Set Position ( 3D Forccs, Inertia, Other)
r
Feedback Element (Position Meter)
Controller System Under Control (Valve Cylinder)
Figure [8]: Elements of the Arms Feedback Control Loop
-
3.2.3 A m Kinematicg
Extending the arms to the drift wail decreases the RPM. Thus, in order to achieve
constant rolling speed of the disk, the RPM fùnctions were selected as hyperboias. These
functions were equai to k/D in the circular portion of the cut, and to e/D in the corner
portion of the cut. The " D term is the outermost diameter of the tunnel, and the " k and
"e" terms are the fiinction factors.
Figure [9] show the components of the arm velocity resulting fiom the rotating movement
of the head and the extending/retracting movement of the cylinder. The first velocity
component is referred to as the perpendicular velocity and is denoted by Vp. It is a
fiinction of the head RPM and thus follows the am's radial position. The second velocity
component is normal to Vp and is referred to as V,,, which signifies the
extendinghetracting velocity, and equals the arm's rate of change of actual position.
The resultant of these two components is the path velocity, Vpath. This component of
velocity is tangent to the path followed by the tip of the disk. The magnitude of the Vp
component is at least 7 times larger than the Vdr component which implies that the
V(path) velocity is mostly af5ected by Vp.
P I + Constant Extending I
Disc Ve/r p. 1 Cutter
ingular Position /& Vr- Pivot
I
Figure [9]: Arm Velocity Components while Extending
3.2.4 Arm Dvnamic Disturbances / Erternal
The arm-valve-cyiinder assembly is subjected to interna1 and extemal disturbances. The
main component of the extemal disturbances are the 3-dimensional forces exerted by the
rock on the arms and are labeled Fx, Fy and Fz as shown in Figure[ 1 O].
Figure[lO]: Components of Reaction Forces fiom the Rock ( Fr is the Resultant of Fx & Fz)
Force Fx is the force exerted on the disk in a direction normal to its axis. Fy is the force
exerted on the disk in a direction normal to the disk axis and perpendicular to Fx. Fz is the
force exerted on the disk in a direction parallel to its mis. These force components are
measured o n a m I (inner single a m ) and am II 2 (one of three outer ms) using strain
gauges. Each component is vital in interpreting the behavior of the machine and its
interaction with the geology of the face. This is illustrated as follows:
(1) Force Fx is resolved along the am cylinder axis affecting cylinder pressure. Its
amplitude depends on the hardness of the rock and its homogeneity as weil as
the user-input depth of cut. Its fiequency depends on the head RPM and the
extendingt retracting velocity of the arm; it also depends on the a m stifhess
which, in tum, is a function of the controller gain parameters and hydraulic
stifiess. Force Fx also resolves in a direction parallel to the face Le. tangent to
the disk vertical contact area with the rock. This component of Fx affects cutter
Wear.
(2) Force Fy is parallel to the am pivot, therefore it does not affect cylinder
pressure. However, this force causes the rotating motion of the disk and, by
summing-up the four a m combined effect of this component (provided al1 arms
were equipped with strain gauges), this force would provide an indication of the
actual toque resisting the rotating head. Force Fy is thus an indicator of the power
eficiency of the head. The fiequency and amplitude of Fy depend on rock geology,
head RPM as well as hydraulic stifhess of the head .
(3) Force Fz acts somewhat similarly to force Fx in a sense that it affects cutter
Wear and cylinder pressure; its amplitude and frequency also depend on rock
geology, head RPM, controller gains and hydraulic stiffiess of the arm loop.
Usually, Fx is higher in the beginning of the cut. Then it goes through a minimum towards
the middle where its amplitude increases once again near the end of the cut. This behavior
is mainly due to a combination of effects such as the changing hydraulic stifkess in the
system, the decreasing head RPM and the changing disk contact area with the rock as the
arm extends towards the drift edges.
The change in hydraulic stifiess is a function of the bulk modulus of the oil and the
combined oil volume in the cylinder and the hydraulic lines. [Spotts, 1984). The disk
contact area is a function of the a m radial position, the disk outer diameter, the disk edge
radius, the penetration as well as the edge angle. These parameters are shown in
Figure[ 1 1 1.
(a)
Edge Radius
Dise Radius
Disc Centerline -
The area decreases as the am extends into the cut because of the increase in the tunnel
radius. The outcome is a decrease in forces required to cut the rock which on the other
hand has a positive effect on disc Wear.
Moreover, force Fz displays a behaviour similar to that of Fx but has a lower amplitude.
The resolving of a component of Fx and Fz dong the cylinder axis causes disturbances in
the cylinder pressures. The amplitude of these disturbances aside fkom the local variations
in the geology is a function of the arm geometry. As the arm extends into the drift edges,
the lever effect of the am first increases rapidly, goes through a maximum then starts
decreasing again.
On the other hand, peak values were found to be much lower in the circular cut than they
are in the corners. This is in part due to a sliding action on the back of the disk and the
increased confinement in the rock as the a m progresses into the cut. Other suspected
factors are protruding rocks left from under-penetration in the previous path as well as
fiequent direction changes in the corner when going from the extending portion to the
constant radius portion and vice-versa.
3.2.5 Arm Dvnamic Disturbanced Interna1
As we have seen in the previous section, 3-dimensional forces can cause variations in the
cylinder and head pressures. But extemal forces are not the only type offorces that affect
the dynamics of the system. Intemal forces have also been show to affect am pressures.
These forces are mainly due to gravity, rotating centripetal effects due to head rotation
and arm rotation around its pivot, as well as fnctional forces at the arm and cylinder
pivots.
These disturbances occur at different levels in the arm-valve-cylinder assembly. This is
illustrated as follows:
At the hydraulic circuit level:
(1) The oil compresses under load i.e. the arnount of compression is proportional to
the forces on the cylinder. The compression yields positioning errors which are
corrected by the controller. The oil inside the cylinders, valves, pumps and
hydraulic lines act like a series of springs with different stifiess coefficient.
Leaks at the valve and other locations of the hydraulic circuit add a friction
(energy dissipating damping effect) component to the spring effect.
(2) The hydraulic tubing in the circuit expands as pressure buiids-up (when the
amis are loaded). Thus, the tubing acts like a spring. Intemal fnction also
dissipates energy which again translates to a dashpot eflect. One direct
consequence to the expansion and contraction of the hydraulic lines is the
variable resistance to oil flow causing oscillations in the oil flow rate.
At the valve level:
(1) The oil stifiess compression effect mentioned before is influenced by the
constriction due to the valve opening. When the valve is completely closed,
one can assume that only the oil inside the valve and cylinder will undergo
compression. As the valve opens, the effect of the oil insida the lines gets added
to that of the oil in the cylinder yielding higher error (the amount of compression
being a function of the tube length and cross-sectional area).
( 2 ) Leaks in the valve prevent production of a precise flow rate to the cylinder as
comrnanded by the controller. Although these leaks are usually accounted for by
the manufacturer, the amount of leakage will increase as the valve wears out.
(3) The valve's intemal dynarnics and response time, accuracy and repeatability
may also have an effect on how predictable the valve response is.
At the arms level:
(1) The effect of gravity. As the arm rotates around its pivot with the head, the
amplitude and direction of the resulting force along the cylinder axis changes.
(2) The effect of centnpetal forces due to the head rotation. As the head rotates, a
centripetal force which tends to pull the arms outwards is generated.
(3) The effect of the a m rotation around its pivot. Every time the a m penetration
increases, the a m inertia causes the a m to resist that movement.
At the overall ann-cvlinder assemblv level:
Frictional forces which ultimately affect cylinder pressure exist at the following contact
points:
(1) Cylinder pivot
(2) Rod-am pivot
(3) Arm cutter head pivot
(4) Rod-piston intemal contact surface
(5) Rod-cylinder cap contact area
Finally, the amplitude and direction of the resulting fiction force depends on the
direction and amplitude of the resultant of al1 3-dimensional forces (Fx, Fy, Fz).
3.3 Reiationshi~ between the Control Svstem and Machine Behavior
3.3.1 Ceneral
The problem of optimization of a control system may be formulated if the following
information is given: [Spotts, 19841.
1. System equations
2. Class of allowable control vectors
3. Constraints on the problem
4. Performance index
5. System parameters
Therefore, the solution of an optimal control problem is to determine the optimal
control vector within the class of allowable control vectors. Mainly, this vector depends
on:
1. The nature of the performance index
2. The nature of the constraints
3. The initial state or initial output
4. The desired state or desired output
Consequently, and for the exception of special cases, the optimal control problem may
be so complicated for an analytical solution that a computational solution has to be
obtained.
3.3.2 Machine Controller Funct ions and Dmigp Confimi ration
In order to achieve optimum performance in the cutting time as well as reduce possible
interruptions due to the errors in the profile, the a n s position must be accurately
controlled. The comrnand flow used is the following:
(1) At any instant in time, the position feedback device measures and sends the arm
cylinder actual position to the controller.
(2) Using this information along with a set position sent by the PC and the profile
generating software, the controller gets a rneasure of the instantaneous
positioning error of the am.
(3) A valve position command is then sent by the controller; The valve reacts to
this command by moving its spool and thus causing an increase or decrease in
oil flow and pressure to the cylinder.
The closed-loop position control system allows, through the setting of the user-input P D -
F gains, influencing of the dynamic behavior of the cylinder which is mainly its stifiess
and response speed. The stiffhess of the system (when it is stationary) is proportional to
the loop gain which, in mm, is proportional to the individual amplification gain of al1 the
elernents of the loop. This is shown in equation (1). The multiplication of al1 gains of the
closed-loop is called the " closed-loop gain ".
As shown in the equation, the stiffhess is also equal to the ratio of the disturbance force on
the system or the reaction forces fiom the rock on the arms over the displacement. This
implies that in order to make the am stiffer Le. more resistant to a reaction force from the
rock, the closed-loop gain must be increased by increasing one or more of the individual
controller gains which in retum affects the valve position.
In spite of the fact that the overall response of the electro-hydraulic valve depends on the
closed-loop gain, the individual gains have different contributions to the response.
Equation (2) shows the control algorithm of the Parker controller [Parker, 19921. While
the servo-valve conunand value "c" ultimately translates to a position of the spool, it is the
result of the contribution of al1 the elements on the right hand side of the equation (2).
As defined by the [Parker, 19921, we have:
KP = User input proportional gain coefficient
Ki = Internai integral gain coefficient
Kv = Intemal velocity feedfonvard gain coefficient
Kd = Interna1 derivative gain coefficient
e = Position error = r-x
r = Commanded or set position
x = Position feedback actual position
s = Laplace transform of the derivative d d t
As the feedback positioning transducer sends the actual position of the am "x" to the
controller, the latter compares it to the commanded position or the advanced set point
position "r". The difference (r-x) is equal to the positioning error "e". These position-
related variables are multiplied by one or more of the controller gains as shown in
equation (2). The proportional gain K, affects the amplitude of the error. Thus, as the
error increases or decreases, the system will respond by opening or closing the valve by an
arnount proportional to the error amplitude. The proportional gain is the most vital of the
controller gains since it provides adequate response to transient variations in the system
caused by the random forces at the arms. If the value of the proportional gain is set too
low, the arm will not react strongly and fast enough to the disturbing force fiom the rock.
This would result in the am position undershooting its set position. On the other hand, if
the proportional gain is set too high, the am would then correct any enor too fast and will
ultimately overshoot the set position.
The integral gain " (K, * Ki* e) / s " in equation (2) above is mainly used to correct static
inaccuracy of the system. The integral gain multiplies the integral of the error with respect
to time. This means that its response builds up with time. Thus, the area under the curve
becomes larger. This characteristic allows the integral terni to deal adequately with static
error. Overshooting and cycling will result if the integral gain is set too high. In other
words, the controller output will eventually become faster than the arm system can
respond to, and the integral time becomes higher than the am dead tirne.
Conceming the velocity feedforward tenn, Kv, this term is very important when one wants
to build anticipation into the system. The velocity feedforward gain terni [(Kv*s)*r] acts
on the rate of change or "slope" of the profile set point function. In other words, the
steeper the set point function, the higher the effect of the feedforward gain. Therefore, by
knowing that the set points require faster position increments h m the arms, increasing
the velocity feedforward gain reduces the am following error when faster variations of the
set point occur. See Figure(l21.
Actual Position 1
4
I
O Undtr Penetration I
* I
I
* 1
4
D Transition Point m I
m I
I
I
I
m I
I
m I
l * I
l I b # Over Penetration m
l a ,-*.-..-*...-.-.-.--.--......-.-*.--*..-..*..*.-.-.-......., 1
Figure [12]: Result of a Typical Positioning Error on the Profile
Therefore, increasing the Kv gain will eventually cause the system to respond too abniptly
to the rate of change of the set point which would result in an overshoot of the system's
set position. In other words, the actual position of the arm will cross the set point curve.
As a consequence, an error of opposite sign will occur.
Finally, since the gain terms associated with Kp, Ki and Kv are al1 added together and are
al1 positive, their sum represents the first portion of the valve response to the disturbance
forces. But the tenn containing the derivative gain [(Kd*s)*x] is subtracted fiom the
previous three terms. Keeping this in mind, and knowing that the derivative gain acts on
the rate of change or "slope" of the actual position of the am, one could deduce that this
terrn has a very similar behaviour to that of the velocity feedforward t e n .
3.3.3 Controller Limitations
The type of controller installed on the CMM machine is the Parker controller. This
controller has a major limitation. It only allows the user to input two sets of controller
gains for a whole cut perfoned by the machine. These sets are referred to as P(extend)
and P(retract). We would like to note that the "extend" and "retract" terms do not
necessarily mean that the first one applies to the extending cylinder and the latter to the
rettacting cylinder; they are merely designations used by [Parker, 19921, and any one of the
two sets can be used whenever judged necessary.The limitation of having only two sets of
gains resulted in not being able to fine-tune the control loop. This is because of the
changing kinematic and dynamic behavior of the arrns under vatiable loading. The solution
to this problem would be to change gains in a manner that follows the behavior (response)
of the ann through the positioning error and therefote control the a m movements
accuratel y.
Changing the gain between the first and second half of the cutting process in a corner cut
is necessary to compensate for the slower reaction time when the cylinder retracts; this is
due to the fact that for identical system pressure on the piston and the rod side, the piston
area is twice the rod area. Thus, half of the force is available on the rod side. Faster
reaction is therefore necessary to compensate (partially) for that inherent limitation of the
hydraulic cylinder. In order to recti& the problem and achieve faster reactions, higher gain
settings should be given on the cylinder retraction side.
Table[l] clearly illustrates this concept where a cornparison between the combined gain
settings P(extend) and P(retract) is performed. Also the ratio of the gains is stated. In
both tests 1 and 2, the ratio of the Retract / Extend is > 1, which means that the
combined retracting gain is higher than the combined extending gain. -
ARMS II COMBINED GAIN SETTMGS
RATIO (RETEXT)
0.943
2.3
O. 94
1.335
Table[l] : Anns iI proportional gain settings
RETRACT(%)
O. 15
0.635' IO-'
O. 15
0.447 12* 1 O-)
EXTEND (%)
0.159
0.276* IO+
O. 159
0.3348.
CYCLE
1
2
ARM TYPE
1 ,
II
i
II
Finallv, despite the limitations of the Parker controller [Parker, 19921, the evaluation
of the changes in the values of the gain settings was still successful. This was because
these settings were in pan done by calculating the following:
(1) The time delay between the extending side and the retracting side of the profile
(2) The ratios of peak to average values for key machine parameters.
3.3.4 Svstem Resmnsiveness - Relation Between the Controller and Time Deiavs
Different tests were performed where the Kp gain was decreased by a factor of 2.5, while
the Ki gain was decreased from 0.0 1 1 to O % (integral gain switch tumed-off). This
simultaneous decrease of gains has caused a decrease in the am responsiveness, that is the
arms became slower to react. An indicator to this response variation is the time delay
between the set point position curve and the actual radius position curve.
While such delays are normal Cor hydraulic systerns behavior, they cm be influenced by
changing the controller settings. The ratios of time delays when arms (111,112 and 113) are
extending into the corner and retracting out f'rom the corner are shown in TabIe[Z]. These
ratios show that the delay is 2.5 to 3.5 higher when Kp and Ki are decreased as pointed
out earlier .
TIME DELAY (Seconds)
Table[2]: Ratio of time delays caused by a decrease in the Kp and Ki gains of the controller
Extending into the comer
I st cut
2nd cut F
Ratio of delays
3.3.5 Ratio of (Peak Pressure/ Average Pressure) vcrsus Controllability
Retracting fkom the comer
When calculating peak.1 average ratios, panicular care is taken for evaluating average
pressures or forces. Since rock failure seems to occur after only a few millimeters of disk
penetration, peak pressures occur right before failure. Once failure has occurred, the amis
basically reach the set penetration level after which their function would be basically
lirnited to clearing the remaining rock chips. Consequently, the arms would require much
lower piston pressures than they normally use.
0.0205
0.0717
3 .508
Furthemore, after the larger chunks of rock are cleared, the arms might be in a position to
cut in air until the next rock is encountered. This means that in order to correctly evaluate
0.053
O. 135
2.5 5
the tme ratios of [peau average] piston pressures, these ratios must be calculated when
the am is in rock and not in air. This procedure was done for three typical test cuts where
the depths of cut was 100rnrn for the first test, followed by 125rnm and 150mm for the
second and third test consecutively. The penetrations were kept constant at a value of
12mm for the three tests. The results are shown in Table[3]. Looking at these results, it is
obvious that the ratios of the [peakl average] pressures are consistent and range tiom 1.5
to slightly beyond 1.6, therefore, demonstrating good levels of controllability of the amis.
ARA4 II 2 TYPICAL PISTON PRESSURE RATIO PEAK/ AVERAGE
CUT RATIO P E N
AVERAGE
Table[3]: Ann II 2 typical piston pressure ratio peaW average
3.4 Actual Machine Profile Gentration
3.4.1 Effects of Disturbances on Svsîem Resmnse
The resulting effect of disturbances on the system is manifested in the way the systern
responds. If one envisions the am-cylinder assembly as having a certain spring stifiess
(mechanical + oil stifiess) and a certain amount of darnping (due to fnctional forces),
then the am-cylinder assembly can be modeled as a mass attached to a spring and a
dashpot in parallel. The assembly would then be subjected to forced vibrations by
components Fx, Fy and Fz. When a piece of rock fails, the am is suddenly released into
the air, and provided that it does not touch the rock for a short period of time, it oscillates
at its damped natural fiequency which may Vary between 0.5 to 2 Hz depending on the
am positions. Since the a m is in air, positioning errors do not matter and thus the effect
of the oscillations is very Iimited. If on the other hand, the forces excite the a m at a
fiequency close to its natural fiequency, large oscillations will result fiom operation near,
or at, resonance, a potentially harmful situation. Fortunately, this only occurs rarely (most
forcing frequencies being well below system natural frequency). Furihermore, when this
happens the forcing frequency changes very rapidly and the arm gets quickly out of the
resonance condition.
Since each arm can be modeled as a mass-spring-dashpot assembly, it becomes clear that
the natural fiequency of the damped oscillations of the arm changes during a particular
cut. This is basically due to the changing stiffhess of the arm as well as smaller changes in
fiction. Therefore, the natural fiequency of arm 1 decreases as it converges towards the
drift center-line, and the natural fiequency of arrn II decreases as it expands towards the
edges of the drift.
3.4.2 Macbine Guidance and i ts Effats on A m Motion and Profile
In this section, focus is on the algorîthm followed to guide and supervise the motion of the
arms as well as to calculate their profiles.
As we have seen earlier, al1 electro-hydraulic systems have inherent delays due to several
factors:
( 1 ) Inertia of the mechanical componeiits
(2) Friction and non-linearity in the system
(3) Processing delays due to lirnited computational capabilities and the complexity
of the mathematics and their computer algorithms.
(4) Others
Mainly, al1 systems have a finite response speed. For the case of the arms of the CMM
machine, the delays in the system were evaluated at 120 to 1 50 msec. A value of
145 msec was used for compensation purposes.
In the CMM machine, the position of the anns determines the geometncal shape of the
profile. See Figure [ 131.
A m II Corner Cut End of Circle Cut
Figure [13]: Division of the Face Cut into Sub-areas
Assuming perfect controllability of both the arms and the cutter head, and provided there
are no delays in the system, no oscillation of the head aiid the piston has unlimited force to
limit the effect of the inertia to close to zero. Then, the am should be reaching
instantaneously its set position with an accuracy within the limits of its measuring
instruments. This is a hypothetical situation, and in order to understand the dificulties in
controlling positioning enors, each one of these issues would be presented clearly.
Leaving aside the control aspects of both the head and arms, and focusing on the arm
positioning algorithm, the first step would be to calculate the shift in profile due to delays
in the system. The hydraulic delay implies that at the instant a set point is given to an am,
it takes 145 msec for the system to react. Meanwhiie the head is turning. This implies that
by the time the am starts responding to the set point, the head has rotated by an angle
which is a function of the angular velocity (RPM). At RPM = 10 (equivalent to 60
degfsec) in 145 msec, the head would have rotated by 8.7 degrees. Some values for
different RPMI delay combinations are provided in Table[<(]. It is to be noted that the
upper limit of 10 rpm used in the table occurs at the end of the circular cut which is
basically the beginning of the corner cut.
RPM
10
Table[4]: Profile shifi due to hydraulic delay for dflerent RPM values
ANGüLAR SHIFT DüE TO DELAYS RADIUS (mm)
(3 VALUES S H O W AS SAMPLE) DELAY (msec) SHLFT@eg)
The position shift resulting fiom the delay at different radial positions as show in
Figure[l4], is calculated as follows:
Equivrlent .. . Distance Equrl to . .
.*.. ' .
I I 342 mm rt Rad --.
2260 mm Head Angular "..-.-.. ... Rotation in 145 msec ""-....+
1
Figure [ 141 : Effect of Hydraulic Delay on Arm Position
By the time the a m starts to react (&er 145 msec), the angular position of the a m would
have moved from point A to point C, the distance AC being equal to 342 mm. At point 4
the arm is supposed to start extending. But because of the delays, it only starts extending
at point C. Since the radius AWBC, the am lands too short at point D as shown in
Figure [15 ] on the following page. From triangle ABC, one can compute the following:
If AB=2260 mm and the RPM= 1 O, then the radius of the arm afier 145 msec should
be equal to:
CB = 226O/(cos(8.7)) = 2286 mm
Therefore: The position error = 2260 - 2286 = -26 mm (under-penetration)
which resuits in a 26 mm under-penetration position error. This error, if constantly
repeated for al1 points dong the extending profile, will cause an angular shift in the
profile.
This is clearly shawn in Figure [15] where the angle of shifting is calculated using
triangle ADC knowing that AB=BD. The shift is equal to 8.7 degrees and is calculated
by using equation (3):
Angular shifl= arccos
Resulting A m Positioning Error (Under Penetmtion)
- - . .. . , -.*.._ -.- .... .-.* .. . . Head Angular Rotation i
in 145 msec
Figure [15]: Resulting EfTect of the Delay on the Arm Position while Extending (Under Penetration)
Finally, knowing the relationship between the RPM and the resulting shift at different
rpm's, one is able to anticipate the system's behaviour by generating a profile based on
an advanced set point. As a result, this profile generates the set point 145 msec earlier
in order for the ann to reach the correct position.
CMM Cutter Desimi. Fragmentation
and Materials
Handline Reauirements
4.1 Introduction
This chapter provides an overview of materials handling systems that are in use today in
the tunnel construction industry. The chapter also stresses the importance of chip size
formation as well as cutter design and their effects on machine performance.
Dunng tunnel construction, the materials handling systems must be designed and used in a
manner that would include provisions to transport everything in the way of materials or
personnel in and out of the tunnel while the work is being perfonned. As a result, the
following items must be taken into consideration:
1. Muck removal
2. Personnel transportation in and out
3. Air for ventilation
4. High pressure air for air-operated tools or equipment
S. Material for temporary support system
6. Water supply or removal
7. Grouting materials and equipment
8. Permanent lining system
Out of these items, the decision on a muck removal system that would be able to
provide a smooth and efficient tumeling operation is of major concem. Nowadays, a
time anaiysis of an overall tunneling operation is expected to show that muck removal
is basically one of the controlling elements goveniing the progress of the work.
[Mathews, 1 9801.
Furthemore, handling and transportation of the other items necessary for the
construction of the t u ~ e l can be usually accommodated by either the muck removal
system or by a suppiemental compatible system.
4.2 Effects of Muck removal on Minin? Excavation Machines
It is clear that any improvements in the excavation efficiency of Tunnel Boring Machines,
as indicated by current advertisement of a continuous advance TBM, requires that
advances be made in muck removal systems. Without such advances, the added costs of a
similar machine would be a wasted investment. [Janzon, 19953.
The significance of muck removal in the case of the Continuous Mining Machine (CMM)
was also of profound importance. This was noticed while the CMM mac!iine was
operating in the Herdecke mine site in Germany. Muck generated during cutting was
removed by conveyors located at the CMM discharge. The conveyors were of smaller
capacity than the CMM conveyor. This caused occasional interruptions to the mns,
particularly when overtlowing conditions occurred at the headsnd of the conveyors.
Chips were then selected fiom the muck pile as well as from the conveyor discharge for
supplemental study. Results of chip analysis and its effects on muck removal strategies are
bnefly discussed in later sections of this chapter.
4.3 Muck Removal Modes for Tunnel Construction
The following are the basic materials handling systerns currently available for tunnel
muck removing: [Yu, 1 9721.
1. Rail
a. Conventional
b. Monorail
3. Belt conveyor
4. Pipeline: Hydraulic
Pneumatic
Most likely, no one system will handle al1 of the materials handling requirements for any
given tunnel task, therefore some combination of systems or modes will be required. A
discussion of these modes and their advantages and disadvantages for an overall materials
handling system is presented in the next section.
4.4 Machine S u ~ ~ o r t i n n Services and Muck Transport
4.4.1 Aurilirriv Eauinment
The availability of a suitable excavation mining machine alone will not get a tunnel
through. The machine has to be served by many supporting seMces such as the
following: [Holden, 19801.
1. Electric Power
Though some machines are equipped with hydraulic motors for cutterhead rotation and al1
machines have a hydraulic system for thrust, gripping and many subordinate functions, the
primary power is always supplied in the fom of electricity.
2. Water
Water is definitely needed for cooling purposes and dust suppression. It is basically piped
in from outside the tunnel wall and connected to the water system on the back-up Ma a
hose to allow continuous machine operation.
3. Com~ressed air
Because of the increasing use of hydraulically powered rock drills for probing or dnlling
rock bolt holes, compressed air requirements nowadays are limited to powenng maybe a
few hand tools or for additional flushing of probing holes.
4. Ventilation
Fresh air must be supplied to the tunnel face for the crew working there. It is also used to
dilute and remove exhaust fumes from diesel engines
5 . Tracks
T U M ~ tracks are installed for railbound transpon of people and materials to and from the
face and, in most cases, of muck out of the tunnel. This is achieved by using a single track
because a double track would require partly refilling the invert to obtain a wide enough
road bed.
4.4.2 Basic Muck Removd Techniaues
The muck can be taken out of the tunnel by a variety of means or systerns, the most
common methods are listed below: [Holden, 19801.
1. Trucks
Trucks are basically used in large diarneter tunnels that are of short length. The large
diameter is required to permit two trucks to pass each other in the tunnel and to be able to
tum behind the excavating machine. The disadvantage of this method is that it requires a
large fleet of trucks. This is because the loaded muck will be mn out by the same truck to
be unloaded out side the tunnel.
2. Muck trains
Trains of muck cars moved by a diesel engine or battery powered locomotives are loaded
from the back up conveyor belt located behind the excavating machine. The train is then
mn out to a tip arranged at some short distance outside the tunnel where the muck
emptied.
3. Convevor svstems
Conveyor systems are in most cases used for muck transport in long tunnels of such a
small cross section that a muck train system would be hard to keep up with the excavation
rate of the machine.
4. Fluid ~umping systems
The method of moving solids suspended in water or a slurry is by pumping them through a
pipe line. This method is comrnon in ore treatment plants and lends itself for muck
transport under certain conditions where:
(a) The availability of a sufficient quantity of water exists
(b) The excavated material is not too abrasive, and
(c) A high percentage of fines in the excavated material and a specified
maximum particle size can be regulated
From the above, it wili be clear that such a pumping lends itself admirably for use with a
slurry shield but would be hard to visualize for use with hard rock.
S. Pneumatic svstems
Such systems have the following basic elements:
An air source, such as a blower, discharges into an air lock injector for the matenal to be
transported and forces the material into a pipe line which in tum directs it to its
destination. It has to be noted that sufficient air pressure is required to maintain a high
enough particle transport velocity throughout the system.
As is the case of hydraulic systems, lirniting particle size is necessary. Few systems have
ever worked satisfactorily with larger than 50 to 75 mm particles, somewhat depending on
the specific density of the material. They are aiso lirnited in transport length and require
the installation of booster blowers at comparatively short distances dong the line. If the
material is abrasive, the cost of pipe lines maintenance becornes high. Thus, pneumatic
systems will oniy be applicable for muck transport under very specific, project dictated
circumstances.
4.4.5 CMM Convevor Svstem venus Other Eaulaee Svstems
The continuous conveyor haulage system used in the Herdecke mine to support muck
removal fiom the CMM machine has been shown to provide important benefits over the
other muck removal systems discussed earlier.
The CMM conveyor haulage system is continuous and always available. In other words,
there is no interruption in the boring cycle and no downtime waiting for muck cars. It is
also safer since most locomotive diesel fumes are eliminated and tunnel traffic is reduced.
Moreover, this system has been show to greatly reduce rnanpower requirements where
very little maintenance was needed in spite of the faster and more productive muck
removal performance.
The advantages of the conveyor haulage system used for the CMM machine over other
systems as show above is basically due to the following:
1. The conveyor system, could be directly attached to the CMM machine trailing
gear, thus the systems advancing tailpiece permits the advance rate to be totally
dictated by the CMM machine.
2. The installation of an advancing tailpiece window ailows carrying idlers to be
installed without stopping the system.
3. A belt storagekake-up unit enables tu~e l ing operations to advance as much as
300 meters before additional belting is required.
4. Tripper-booster drives make possible conveyor belt runs 10 km long.
Thus, the result of using this tumeling conveyor system has been shown to be
very efficient in terms of productivity and profitability.
4.5 Muck chi^ Size and Form Based on the CMM Performance in Aerdecke. Germany
4.5. t Rock conditioiis ai the Herdeckc Site
The rock conditions at the Herdecke site are classified as medium hard with medium
abrasivity . The Unconfined Compressive Strengt h (UCS) ranges between 10 1 to
15 1 MPa for sandstone and 25 to 55 MPa for shale. [Repski, 19951.
4.6 Effects of Cutter Design on Machine Performance
4 . 6 CMM Cutter Desien Based on TBM's Cutter Performance
Based on previous experience from TBM's, it is well known that increasing the cutter
diameter reduces the penetration rate for the same force level. This is primarily because of
a larger contact area under the cutter. Thus, a thinner edge results in lower forces, smaller
chips and slightly different chip angles. This indicates that one should use a cutter edge
that is as thin as possible.
Also, it was shown that for TBM's the cutter diameter in hard rock conditions should be
at least 43 cm, to get sufficient penetration rate and cutter life in
sm3/ cutter (depth of cut * volume per cutter) .
Assuming that the CMM will need average cutter loads ofapproximately 220 - 250 kN to
cut efficiently in hard rock, one should not start testing with less than 43 cm diameter
cutter rings because the strength of the steel is not good enough to apply the force level
needed to utilize the CMM fully.
Furthemore, if chipping of steel fiom the rings becomes apparent, the solution would be
to increase the cutter edge width or use smaller cutters with 46 to 48 cm radius.
Moreover, an optimum diameter rnay be found when tiirther testing of the machine reveals
more data on Wear, forces and penetration.
4.6.2 Effccts of Increasind Dccreasine Disc Parameten on Machine Desirn
The cutting discs are the interface between the machine and the drift. They convert a
certain portion of the power generated by the machine into a cutting action on the rock.
They are therefore one of the most critical components of any excavating machine.The
design of the disc directly affects the following:
a) Thetechnicalaspectsofthemachine.Thatis,itscuttingabilityin
different types of rock, and the efficiency with which machine power is
converted to actual rock cutting.
b) The economics of the machine. The Wear rate of the discs will dictate
how efficient the machine is in terrns of maintenance.
The disc parameters that most influence these technical and economic aspects are:
a) The actual dimensions of the discs; this includes the outer diameter, edge radius and edge angle. See Figure [16].
b) The material of the disc Le., physical properties such as strength and ductility.
&
Dise Outer Diameter
w Edge Radius
Figure [ 1 61 : Disc / Ring Paramet ers
As we have seen in the previous section, larger contact area with the rock implies that
more pressure must be supplied at the cylinder level to compensate for the fact that the
tip of the disc is spread over a larger surface.
On the other hand, increasing disc size dso implies slower Wear rates since equal amounts
of energy are spread over a larger area. Finally, the deeper the disc penetrates rock, the
larger the contact area; therefore, increasing penetration will increase contact area. In view
of these facts, it is apparent that optimizing the disc sizes will imply making a compromise
between forces required to perfonn a cut and Wear rate of the disc.
4.6.3 The Theorv Behind Cutter Wear
To mathematically quanti@ the amount of wear on the discs is very difficult due to the
intncacy and the number of factors that affect such wear. However, this does not mean
that good predictions on cutter life cannot be achieved.
For the C M . machine, Wear of the discs is linked to several major areas of concem:
a) The cutting pnnciple involved i.e., how the rock fails under the disc force.
b) The mechanical properties of the materials involved. This includes such factors
as hardness of the disk as well as drillability and abrasivity of the rock.
c) The effect of geological parameters such as the degree of fracturing on the face.
It is well established that mechanical Wear results fiom the fiction that occurs when the
discs move relative to the rock face. This Wear is due to the sliding action of the discs. The
Wear rate is a direct function of the amplitude of the forces on the discs and the fiequency
of occurrence. Therefore, one can conclude that if both the amplitude and frequency of
the forces exerted on the discs by the rock are reduced, this will reduce Wear on the discs.
From basic mechanics, it is clear that to properly estimate the amplitude of the fiction
force due to the forces at the disc, one must have precise values of both the static and the
dynamic fiction coefficients. Furthenore, a distinction between the different types of
Wear has to be divided according to the following categories:
a) Cuttinn wear:
It is noted to [Spotts, 1 9841 that cutting Wear results " when a hard rough surface is
rubbed over a sofier one with or without lubricants". Thus, onless the materials have
widely different hardness, cutting Wear soon ceases as the parts become wom in. This
statement is essentially correct for machinery Wear - such as a drilling operation - since the
bit and the piece of material remain in contact for a relatively long period of time i.e. ,
enough to even out the material surface at the drilled spot. For the CMM, and due to
continuous rotation of the cutting head, the disc remains in contact with any instantaneous
rock location for a short period of time; This implies that the disc is constantly cutting new
rock surface i.e., the rock surface never evens out. Therefore, cutting Wear is always
present throughout the cut , but is limited.
b) Abrasive wear:
It is also noted to [Askeland, 19941 that "Abrasive Wear occurs when the material is
removed from the surface by contact with hard particles, which may either be present at
the surface of a second material or may be present as loose pa~icles between two
surfaces". Abrasive Wear is the most common type of wear for machinery. Typically,
materials with high hardness and high hot strength (ratio of the strength at the operating
temperature to the strength at room temperature is close to 1 ) will be more resistant to
the abrasive Wear resulting fiom cutting. It seems that Wear on the CMM discs is strongly
related to abrasive Wear; as a matter of fact, preliminary findings suggest that when a
metal disc is used to cut rock, a boundary layer forms between the disc and the rock
surface. The disc surface and the rock surface are then linked through this boundary layer
which is typically made of small rock fragments which cause abrasive Wear of both the
disc and the rock surface. These findings are supponed by the calculations and cornparison
of the fnction coefficients of rock to rock and rock to metal contact. The rock to metal
fnction coefficient came out to be equal to that of the rock to rock friction coefficient
which confirms the boundary Iayer assumption. [Askeland, 1 9941.
Wear is also affected by adhesion of the surfaces rubbing together as well as increases in
surface temperature. As the disc rubs the rock surface, fnction causes a temperature
increase at the disc ring. Temperature increase causes a change in the mechanical
properties of the ring steel, particularly, its ductility, tensile strength and yield strength.
However, for the change to be substantial, a noticeable increase in temperature must be
measured (in the 100 to 300 OC range).
4.6.4 CMM Cutter Life Estimation
Cutter Wear on the CMM machine was measured by estimating the Wear on the discs
attached to arms II and arm 1. (See Figure [17]). This Wear was only due to contact
between the cutter and the new face because at the tip of the ring the Wear was found to
be to negligible.
I Figure [17]: Actual Cutter Wear
The CMM Cutter Life was estimated to be equal to the following:
(Volume Bored by CMM) I (Total # of Cutters Used)
where:
Total # of Cutters Used =( # of Cutters Used on Arm 1) + ( # of Cutters Used on Arms II)
and:
# of Cutters Used on A m 1 =
(Wear on Disc * Wear Rate)/ (Maximum Recommended Wear on Discs)
# of Cutters Used on Arms II =
[(Wear on Disc * Wear Rate)/ (Maximum Recommended Wear on Discs)] * 3
Exoerimental Data and Preliminaw Analvsis
5.1 Hardware and Software Instrumentation
S. 1.1 CMM Simal Selection and Identification
Monitoring and control of the CMM machine was performed by the use of sensors and
transducers which generated signals that were sent to a patch panel. These signals were
then taken fiom the patch panel and recorded on a Teac Cassette Data Recorder. A
strip chart recorder was also used to check for dead signals resulting from defective
transducers. See Figure [ 1 81.
Control Sensors , O Panel
Viewdac 1
Teac Recorder Rnw Data
Excel L
Analogue
i
k
r I
Figure [ 181: Block Diagram for Recording and Digitizing Data 68
PC - A/D- 16 Channel
0
Viewdac ASCII Binary
Out of the 100 sensors connected to the CMM machine and sent to the patch panel to
be recorded by the Teac Cassette, only the following 16 signals were digitized using the
Viewdac data acquisition system:
Arm 1 radius actual
Arm II 2 radius actual
Arm 1 rod pressure
Arm 1 piston pressure
Arm II 2 rod pressure
Arm 11 2 piston pressure
Arm I Fx ( force exerted on the disc in a direction normal to its axis)
Arm 1 Fy ( force exerted on the disc in a direction normal to the disc axis and perpendicular to Fx)
Arm I Fz ( force exerted on the disc in a direction parallel to its axis)
Arrn II 2 Fx
Arm 11 2 Fy
Arm II 2 Fz
Rotating pressure left (head feed pressure)
Rotating pressure right (head retum pressure)
Rotor RPM
Rotor angular position
S. 1.2 Hardware
Teac Cassette Data Recorder
The Teac recorder is a magnetic tape recording unit that can record and play 2 1 tracks on
a VHS tape simultaneously. Each track can record data through either Frequency
Modulation (FM), Pulse Code Modulation (PCM) or Direct Recording. In this study, and
for the purposes of digitizing the data, the Pulse Code Modulation (PCM) was used.
S t r i ~ Chatt Recorder
Strip chart recorders provide simultaneous data recording and displaying (through a
limited memory buffer). One stnp chart recorder was used during the digitizing process
which provided continuous monitoring of four signals. Because of the limited ability of
stnp charts due to paper plotting space, the recorder used only provided a quick look at
selected monitored signals. More detailed analysis was perfonned at a software level using
Viewdac, Excel and MatLab.
5.1. 3 Software
Viewdac
Viewdac is an integrated package for data acquisition, control, analysis and graphies.
Viewdac allows for performing mathematical manipulations of data such as moving
averages, filtering and Fast Fourier Transfomis (FFT). For our case of digitizing the data,
Viewdac was programmed to read 16 signais simultaneously and plot these signals on-
screen. In order to allow data importing to Excel or MatLab, the software provided the
capability to Save the raw unfiltered or filtered signals in ASCn files. This was performed
because of Viewdac's limited graphics capabilities and its lack of flexibility in data
manipulation.
Matlab
Matlab is mathematics software that allows the manipulation of tabular data loaded
from Viewdac ASCII files in a matrix form. By the aid of a special program written in
Matlab language, we were able to filter the raw ASCII data imported from Viewdac.
Excel
Excel is a spreadsheet software that could also handle large amounts of data and present
this data in tabular or graphical form. This software was used in parallel with Matlab in
the post processing stage of analyzing the data.
Digitking the data was performed by playing previously recorded tapes on the Teac recorder and copying them to the cornputer using a data acquisition system. These tapes contained previous recordings of cuts performed by the CMM machine in both Canada and Gemany.
5.2 Filtering
5.2.1 N o i e
The signals recorded from the CMM machine contained two types of noise:
(a) 50 Hz EMF-ind~ced
(b) Digital noise
The 50 Hz noise was due to the proximity of signal lines to power lines and various
electrical equipment. In an attempt to block this 50 Hz interference, al1 cables were
shielded, but this did not completely block the interference. Therefore, the amplitude of
the 50 Hz noise was added to the amplitude of the recorded signal. Please note that these
noisy signals can be removed by using either hardware or software filters as shown in the
next section.
As for the digital noise, it was a very high fiequency noise which could have been
either generated within the electronic circuits or was caused by the Viewdac data
acquisition system itself Whatever the reason, it was characterized by a very short
blip that had a very high amplitude and short nse and fdl times that seemed to occur at
random. In general, this had no effect on the test results that were generated.
5.2.2 Filtcrinp in Viewdac
The data acquisition and monitoring software developed in Viewdac provided lowpass
filtenng capabilities that could be performed on certain data by filtering each column
separately. The user also has the choice of either saving the filtered or unfiltered data in
ASCII files. Having this choice as an option, the opponunity of understanding important
information regarding those signals that were aflected, or not, by the 50 Hz noise was
moaly provided.
The CMM machine data was digitized and filtered in Viewdac using a lowpass filter of
approximately 48 Hz. The sampling frequency was kept the same at 256 Hz.
5.2.3 Filterinr in Matlab
Mer digitizing the data using the Vieadac data acquisition system, it was saved as
ASCII files. In order to open these files in Matlab and filter them, a program had to be
written using programming in the Matlab language. See Annex [A].
5.3 Processinn of the Dinitized Data - Initial Conclusions
5.3.1 Pmcessin~ Mcthods
In order to make use of the digitized data, a decision on the issues or methods of how to
analyze this data had to be taken into consideration. In Our case and for the purposes of
this thesis, two methods of analysis are discussed:
(1) Pattem Recognition
(2) Specific Energy
Pattern recognition is only discussed from the theoretical point of view, whereas specific
energy is used to perfonn the actual analysis.
5.3.2 Oveiview on Patîem Recognition
Pattem recognition refers to the analysis of the complex processes involved in recognizing
patterns. The senses and the brain perform these tasks in the human being. [Duda and
Hart, 19731. The field of pattern recognition is involved especially with the manufacture of
artificial systems that achieve similar ends, whether using the same methods or not. Pattern
recognition is used in the recognition of sensory input, but also applies to information
already stored. This entails looking, for exampie, at a set of events and trying to detect a
recumng pattern or sorne other feature that makes prediction possible. A statistical
technique that provides for such an analysis is called the Stochastic Process.
Statistical pattern recognition is based on a well-founded mathematical theory which has
proven to be usefbl in numerous applications in the rnining industry, and especially in the
automatic recognition of different rock types and geologic conditions. pollitt and Peck,
19911. ~oreover, this method is well suited for coping with noisy and distorted patterns.
For the case of classification problems, the objective is to find the pattem class that a
given input is most likely to belong to. Thus, the input is described in tetms of "features"
where the set of features defines a "feature space" of al1 possible measurernents. Thus, for
a given input, features fiom the raw data are measured to form a feature vector. In other
words, if we have n features, this implies that the feature space is n-dimensional.
In general, a good feature extractor makes the job of designing a classifier trivial,
whereas a reliable classifier could use raw data as features. See Figure [19].
Figure [19]: Block Diagram for Pattern Recognition and Classification
Subsequently, if the features are defined appropriately, then al1 feature vectors derived
from patterns belonging to the same class fonn a cluster in the feature space. In order to
classify, we must know or estimate the probability distributions of the features for each
pattern class, or we could use other altemate techniques such as the " K-Nearest
Real World
, Classifier , Receptor I , Featun Extractor
Neighbor" classification, where the rule classifies a set of n samples by assigning them one
by one to the label rnost fiequently represented among the k nearest samples.
[Tou and Gonzalez, 19741.
5.3.3 Pattern Recomition in Relation to Sensinn of Rock Proaertics
The automatic identification of different rock types is based on monitoring of machine
variables by extracting correlations between machine parameters and rockmass properties.
Thus, in order to achieve these correlations, we should nomalize for al1 system variables
so as to clearly identiQ the effects of changes in the geology.
For the CMM machine, the machine parameters are the 3 dimensional forces on the cutter
arms along with the piston pressures of the a m actuators. As for the çeological
parameters, the compressive strength (UCS) is used for the purposes of the analysis.
Based on testing of the CMM machine in both Germany and Sudbury, it was shown that
the variables that mostly affected the machine parameters are divided into the following
two categories:
(1) User-Selected:
(a) Penetration of the arms
(b) Depth of cut
(CI
(d) Disc size (edge radius, disc diameter)
(e) Arm kinematics-geometry
(2) Other:
(a) Disc Wear
(b) Geology
(c) Leaks in the system
(d) Wear in the system
(e) Calibration changes
Therefore, reducing the effect s of these variables on machine sensing of rock properties
reduces the performance error of the CMM machine in discriminating between different
types of geological pattems while operating in a mine.
Pattern recognition concepts and techniques are well suited for the analysis of relatively
unstructured data where they require a training session of the data set. This training
data set or Labeling is representative of the various different pattems corresponding to
the different conditions or classes that are to be discriminated. [Bieniawski, 19741.
In terms of the CMM machine data analysis, pattern recognition techniques could be easily
used to divide the data into separate classes. Thus, one could divide the separate classes to
Shale and Sandstone under the sofi rock category, and to granite and norite onder the
hard rock category. In addition, more classes could be found by further looking into rock
structure, faults and rock strength that are representative of the various locations of
different edges and geometries in the rock.
The most important step in using pattern recognition for the data analysis of the CMM
machine would be in the gathering of the data. This is usually done under controlled
conditions in which the class is reliably established by direct observation such as by
looking at core logs. Once this is established, the labeling of the data is set and different
rock types are separated under different classes.
Following the labeling of the data, an inspection of the vanous variables is performed in
order to rank which of the variables are to be used for decision making. Not only that, but
a knowledge of the machine/rock interaction is a step forward towards fomulating
intermediate variables that can have a higher correlation with the classes than the raw data.
An example of this is the calculation of the rock strength based on variables such as thrust,
torque, speed and advance.
Once the relationship between the machine parameters and the rock mass properties has
been established, various statistical tests and approaches could be performed in order to
detect the degree and ranking of the various correlations. Thus, the resulting analysis
forms the basis for choosing between the various pattern recognition techniques and their
different algonthms.
As far as testing for pattern recognition techniques and algorithms, it is essentially a
programming methology which is not covered in this thesis.
5.3.4 S~ecific Energy - Profile
Specific Energy (SE) is the energy required to remove a unit volume of rock. Other
definitions of SE are also specified in terms of the new surface area created. Moreover, SE
can be closely approximated to be equal to the compressive strength of the rock.
[Bieniawski, 19741.
During a typical cutting process, the hardness of the rock varies randomly. Thus, adjusting
the cutting parameters according to these changes can significantly improve the cutting
performance of the machine if an on-line measurement of the rock hardness is possible.
However, calculation of this property dunng the cutting process is quite hard and
troublesome. Altematively, the Specific Energy can be calculated on line
during a cut, and because it is considered to be one of the quantities that gives an
approximate estimate of the rock hardness for any particular cut perfonned by the
machine, it is therefore usefùl to study its effects on the performance of such machines.
[Brady and Brown, 19851.
5.3.5 S m i f i c Enernv Com~utations Bsiscd on the CMM Machine Desien and Oaeration
In this study, two different methods of computation for the Specific Energy per unit
volume are suggested:
(1) Computation of Specific Energy based on manufacturer
recommended values of operation specific to the CMM machine .
(2) Computation of the Specific Energy by on-Iine caiculation of the CMM
machine dynamics. (forces, torques and RPM etc) .
In the fint case " Computation of Specific Energy using CMM Machine Manufacturer
Guidelines ", the following is provided:
- - - -
Radius of Cut (mm) *
O -1278
Where RI1 000= (am radius in mm) / (1000 mm 1 m)
1278-2270 (end of circle) 2270-2640 (end of cut)
and A p = (head feed pressure - head return pressure) in bars
-- -
~ o G u e (KN.m)
0.917 Ap
But each bar = 10' Pa. Using this conversion factor would give the pressure in Pa.
-- - -
Specific Energy (MJ / m3)
0.7 18*(R11000) * h p 1.66* Ap
From the above, the total Specific Energy would be:
filtcr = [time;FIRad-A';F112WA'; F1Rod-P'; FIPis-P'; FI12Rod-P'; FII2Pis-P'; Fffx': FFy': FEZ'; FII2Fx'; FII2Fy'; FII2Fz'; FRP-Lefi'; FRP-Right'; FRot-RPM': FRot-Apos]: fil-save = input ('Please enter the file name for the filtercd data using the correct path: Y s ' ) :
fid-write = fopcn (fil-savc, 'w'); if fid-writc ==-1
fprind('\nh Sorq p u have entercd an invalid ID file namc; \nui'); clse fpmtf ('\n\n The file has been successfully opened io be savcd \n\n');
end fprintf (fid_\Mite,?4&.2f, %6.2f, %6.2f, %.2f, %6.2f. %.2f. Yd.2f. %.2f, %6.2f. Y&.2f, %6.2f. Yi .2f . %6.2f, %.2f, %6.2f, %6.2f\n7. fillcr); fclose(fid-wri te); fprintf ('Mn The file saving operation is completed\n\n'); plot (time, IRad-A, 'r', time, FiRad-A, 'g'); grid; zoom; % End of Program