02 Whole

University of Wollongong Thesis Collections
University of Wollongong Thesis Collection

University of Wollongong Year
Belt conveyer transfers : quantifying and

modelling mechanisms of particle flow
David Bryan Hastie
University of Wollongong
Hastie, David Bryan, Belt conveyer transfers : quantifying and modelling mechanisms
of particle flow, Doctor of Philosophy thesis, School of Mechanical, Materials and Mecha-
tronic Engineering, University of Wollongong, 2010. http://ro.uow.edu.au/theses/3094
This paper is posted at Research Online.

BELT CONVEYOR TRANSFERS: QUANTIFYING
AND MODELLING MECHANISMS OF
PARTICLE FLOW
A thesis submitted in fulfilment of the

requirements for the award of the degree
DOCTOR OF PHILOSOPHY
from
UNIVERSITY OF WOLLONGONG
by
DAVID BRYAN HASTIE, BE (Hons), ME (Hons)
SCHOOL OF MECHANICAL, MATERIALS AND

MECHATRONIC ENGINEERING
FACULTY OF ENGINEERING
2010
CERTIFICATION
I, David Bryan Hastie, declare that this thesis, submitted in fulfilment of the
requirements for the award of Doctor of Philosophy, in the School of Mechanical,
Materials and Mechatronic Engineering, Faculty of Engineering, University of
Wollongong, is wholly my own work unless otherwise referenced or acknowledged.
This document has not been submitted for qualifications at any other academic
institution.
_______________________________
(Signature)
David Bryan Hastie

May 2010
ii
iii
ABSTRACT
The purpose of this research was to determine if either analytical methods or numerical
discrete element modelling could be used with accuracy to design conveyor transfers.
This goal was achieved using two test materials, polyethylene pellets and corn, which
were selected for their different particle and bulk properties and also for a third product,
iron ore, but to a lesser extent due to test rig limitations.
The design of conveyor transfers has traditionally been based on either trial and error or
previous experience and seen as a black art rather than a science, as such very few
design guides are available.
The design of conveyor transfers can be based on experimental investigations, although

this method can be costly to companies, taking vital resources away from the key goal
of continuous production. The analytical models have existed for some time and have
become widely accepted design tools; however, there is limited validation of these to
determine their overall performance (both advantages and disadvantages). The
analytical models are two dimensional in application and their accuracy with respect to
the three dimensional nature of transfer chutes is not clear. This is an area which needs
further investigation.
The design of transfer chutes has undergone an evolution since the advent of discrete
element modelling (DEM) as well as increases in computer processing power. The
potential to simulate and predict the behaviour of a transfer chute design before it is
constructed can be highly desirable with the prospect of saving substantial time and
money. This being said, there has been little validation published on the application of
DEM in industrial applications, although in recent years this has started to increase with
the realisation that companies need to be convinced this is a legitimate design tool.
Additional DEM validation is warranted with respect to conveyor transfers.
An experimental test program was undertaken following the design and commissioning
of a novel conveyor transfer research facility. This experimental work focussed on two
main areas; investigation of particle flow of material through a conveyor transfer hood
and spoon and the generation of conveyor trajectories. From these areas, real data was
obtained for a range of granular free-flowing products, using a combination of high-
speed video capture and still photography, for the purposes of validation. The effect of
belt speed, material feed rate and the positioning of the transfer hood and spoon were
considered in these investigations.
Additional to this experimental work was the testing and collection of a wide range of
particle and system characteristics for use in the analytical modelling and discrete
element modelling components of this research.
Two analytical models were then used to predict the particle flow of the test materials
through the conveyor transfer hood and spoon. Belt speed, material feed rate and
positioning of the transfer hood and spoon were all considered as part of this analysis to
provide direct comparisons with the data obtained from the experimental testing.
Prediction of the conveyor trajectories was performed using seven trajectory models
available in the literature. These comparisons investigated the effect of belt speed and
iv
mass flow rate on the trajectory profiles, again providing a direct link with the
experimental data.
The discrete element method was used to generate three dimensional simulations of the
material flow through the conveyor transfer hood and spoon and also conveyor
trajectories, based on 3D CAD models of the conveyor transfer research facility. These
simulation outputs were then compared to both the experimental data and data obtained
from the analytical models. Two software packages were used, Chute MavenTM and
EDEM. Chute MavenTM was used to produce the initial transfer chute and trajectory
simulations using spherical particles. High material feed rates corresponding to those
tested experimentally could not be simulated and so EDEM was employed to develop
further simulations. The fact that EDEM has the ability to model both spherical and
shaped (clustered) particles was utilised to investigate the effect of shape on simulation
output.
A critical aspect of any discrete element modelling is whether the outputs are realistic.
To minimise any potential issues, a wide range of bench-scale calibration experiments
and simulations were also completed to validate both DEM packages used.
It can be concluded that the analytical models for conveyor transfers provided close
approximations from a two dimensional perspective, however, there were some slight
over-predictions evident in some situations. For conveyor trajectories, the models
presented a substantial variation in prediction, however, one method stood out as being
accurate under all conditions for the materials tested experimentally.
Findings from the discrete element modelling showed the dynamic behaviour mimics
that of the experimental testing and there was a general agreement with both the
experimental investigations and analytical models for the conveyor transfer
comparisons. With respect to conveyor trajectories, the DEM results agreed with the
results seen experimentally and also predicted the same trajectory path as the one stand
out analytical trajectory method mentioned above.
The importance of DEM calibration and validation has also been documented and
shown to be an absolute necessity in the successful simulation of industrial applications.
v
ACKNOWLEDGEMENTS
I am extremely grateful to my supervisor, Associate Professor Peter Wypych, and co-
supervisor, Emeritus Professor Peter Arnold, for their supervision, guidance, continual
encouragement and invaluable advice over these past three years.
I also wish to acknowledge the financial support provided by the Australian Research
Council through their Linkage Projects (ARC-LP) funding scheme for the project
Quantification and Modelling of Particle Flow Mechanisms in Conveyor Transfers
which this PhD research has been directly associated with.
Thankyou also to the partner organisation associated with the ARC-LP, Rio Tinto
Technology and Innovation and Rio Tinto Iron Ore Expansion Projects for their
financial and in-kind contributions to the Linkage Project which allowed this research to
be pursued, with specific thanks to Dr. Ted Bearman, Dr. Thomas Fraser and Carl
Wilson.
I also wish to acknowledge the technical support from Leap Australia Pty Ltd and DEM
Solutions Ltd for the DEM code EDEMTM.
Thanks to the technical staff of the Bulk Materials Handling Laboratory, Mr Ian Frew,
Lab Manager, for his help through the construction stage of this research and numerous
discussions where ideas were bounced back and forth. Thanks also to Mr Ian McColm
and Mrs Wendy Halford for their assistance with the flow property testing that was
required.
Thanks must also go to Andrew Grima for aiding in the design, construction and
commissioning of the conveyor transfer research facility, the initial Matlab coding for
the DEM analysis, getting EDEM operational and also assisting with the experimental
testing.
Most importantly, thanks to my family for their support and understanding for the
duration of this thesis. An enormous thankyou must go to my wife, Justine, whose
patience was sometimes pushed to the limit while enduring endless nights and
weekends with me tucked away in the study analysing data and writing this thesis.
Finally, to my two adorable children, Alyssa and Byron, who didnt understand why I
wasnt able to play with them as much as they wanted me to, I promise that will change
now!
vi
TABLE OF CONTENTS
BELT CONVEYOR TRANSFERS: QUANTIFYING AND MODELLING
MECHANISMS OF PARTICLE FLOW
CERTIFICATION ii
ABSTRACT iv
ACKNOWLEDGEMENTS vi
TABLE OF CONTENTS vii
LIST OF FIGURES xv
LIST OF TABLES xxiii
NOMENCLATURE xxv
CHAPTER 1
INTRODUCTION 1
1.1 BACKGROUND 2
1.2 OBJECTIVES AND SCOPE OF THE RESEARCH 3
1.3 ORGANISATION OF CHAPTER CONTENT 5
CHAPTER 2
LITERATURE REVIEW 6
2.1 INTRODUCTION 7
2.2 ISSUES RELATING TO CONVEYOR TRANSFER DESIGN 8
2.2.1 Conveyor Belt Wear and Damage 8
2.2.2 Spillage of Material 9
2.2.3 Degradation of Material 9
2.2.4 Material Hang-ups 9
2.2.5 Blockage of the Transfer Chute 10
2.2.6 Noise Emissions 10
2.2.7 High Maintenance Costs 10
2.3 DUST CONTROL 10
2.4 CONVEYOR DISCHARGE AND TRAJECTORY 11
2.4.1 C.E.M.A. 12
2.4.2 M.H.E.A. 1977 13
2.4.3 M.H.E.A. 1986 14
2.4.4 Korzen 14
2.4.5 Booth 16
2.4.6 Golka 17
2.4.7 Dunlop 20
2.4.8 Goodyear 20
2.4.9 Roberts 21
2.4.10 Trajectory Discussion 23
2.5 UPPER TRANSFER CHUTE DESIGN OPTIONS 24
2.5.1 Hood Design 24
2.5.2 Impact Plates 27
2.5.2.1 Method of Lonie 27
2.5.2.2 Method of Korzen 29
vii
2.5.2.2.1 Initial Conditions 29
2.5.2.2.2 Non-cohesive Materials 30
2.5.2.2.3 Cohesive Materials 31
2.6 LOWER TRANSFER CHUTE DESIGN OPTIONS 33
2.6.1 Chute Angles 33
2.6.2 Rock Box 34
2.6.3 Spoon Design 34
2.6.3.1 Material Flow Through a Loading Chute 34
2.6.3.2 Material Flow in a Constant Radius Curved Chute 34
2.6.3.3 Material Flow in a Parabolic Curved Chute 35
2.7 MATERIAL FREEFALL 36
2.8 AIR SUPPORTED BELT CONVEYORS 41
2.9 DISCRETE ELEMENT MODELLING 43
2.10 APPLICATIONS OF DEM 43
2.11 CALIBRATION OF DEM 43
2.12 PARTICLE SHAPE IN DEM 46
2.12.1 Circular and Spherical Particles 47
2.12.2 Ellipsoid Particles 48
2.12.3 Multi-Sphere Particles 48
2.12.4 Spherocylinders 50
2.12.5 Superquadrics 50
2.12.6 Polygons 50
2.13 CONTACT FORCE MODELS 51
2.13.1 Linear Spring-Dashpot Model 51
2.13.2 Partially Latched Spring Model 53
2.13.3 Non-Linear Spring-Dashpot Model 54
2.13.4 Hertz-Mindlin Model 54
2.13.5 Improved Hertz-Mindlin Model 55
2.13.6 Hertz-Mindlin Model Without Slip 56
2.13.7 Hertz-Kuwabara-Kono Contact Model 56
2.13.8 Effect of Rolling Friction 56
2.14 DEM CODES AND LIMITATIONS 57
2.14.1 DEM Codes Developed by Research Groups 57
2.14.2 Commercially Available DEM Software Packages 58
2.14.2.1 PFC2D and PFC3D 58
2.14.2.2 Chute MavenTM 58
2.14.2.3 Chute Analyst 58
2.14.2.4 EDEM 59
2.14.2.5 Newton 59
2.14.3 Factors Influencing Simulation Time 59
2.15 COMPARISONS OF THE CONTACT FORCE MODELS 61
CHAPTER 3
THE CONVEYOR TRANSFER RESEARCH FACILITY 63
3.1 INTRODUCTION 64
3.2 KEY DESIGN CRITERIA 64
3.2.1 Conveyor Belt Speed 64
viii
3.2.2 Conveyor Belt Width 64
3.2.3 Conveyor Belt Surface 64
3.2.4 Belt Conveyor Selection 65
3.3 FACILITY LAYOUT 65
3.3.1 Conveyor Support Frames 66
3.3.2 Transfer Chutes 66
3.3.2.1 Transfer Chute 1 67
3.3.3 Feed Bin 68
3.4 INSTALLATION OF THE CONVEYOR TRANSFER RESEARCH
FACILITY 69
3.5 DRY COMMISSIONING 71
3.6 FINAL COMMISSIONING 71
3.7 FEED BIN AND CHUTE LINERS 72
3.8 DUST EXTRACTION 72
3.9 ADDITIONAL FRAMES 74
CHAPTER 4
CALIBRATION OF THE CONVEYOR TRANSFER RESEARCH
FACILITY 75
4.1 INTRODUCTION 76
4.2 DATA ACQUISITION 76
4.3 CALIBRATION OF THE CONVEYOR BELT SPEED 76
4.3.1 Additional Conveyor Belt Speed Calibration 77
4.4 CALIBRATION OF THE FEED BIN LOAD CELLS 78
4.5 CALIBRATION OF THE HOGANTM VALVE 79
CHAPTER 5
PARTICLE AND BULK CHARACTERISTICS 81
5.1 PARTICLE AND BULK CHARACTERISATION AND MEASUREMENT 82
5.1.1 Loose-Poured Bulk Density 82
5.1.2 Particle Density 82
5.1.3 Equivalent Volume Diameter 83
5.1.4 Particle Sphericity 83
5.1.5 Particle Size Distribution 84
5.1.6 Coefficient of Restitution 85
5.1.7 Wall Friction Angle and the Coefficient of Wall Friction 86
5.1.8 Internal Friction Angle 88
5.1.9 Static and Kinetic Wall Friction 88
5.1.10 Particle-Particle Friction 90
5.1.11 Angle of Repose 90
5.1.12 Surcharge Angle 90
5.1.13 Terminal Velocity of Particles 91
5.1.14 Poissons Ratio and Shear Modulus 91
5.2 TEST MATERIALS 92
5.2.1 Polyethylene Pellets 92
ix
5.2.2 Yandicoogina Iron Ore 93
5.2.3 Corn 94
5.2.4 Determining the Shear Modulus and Poissons Ratio 95
CHAPTER 6
DISCRETE ELEMENT MODELLING SOFTWARE 98
6.1 INTRODUCTION 99
6.2 CHUTE MAVENTM 99
6.2.1 Three Dimensional CAD Models 100
6.2.2 Model Parameters 102
6.2.2.1 Coefficient of Friction Between Particles 103
6.2.3 Simulation Data 104
6.2.4 Performing a Simulation 105
6.2.4.1 Optimisation of Simulation Time 105
6.2.5 Interpreting Results 106
6.2.6 Calibration of DEM at the Bench-Scale Level 107
6.2.6.1 Slump Model 107
6.2.6.2 Hopper Model 108
6.2.7 Sensitivity Analysis 109
6.2.7.1 Trajectory Geometry 110
6.2.7.2 Effect of Particle Size Distribution 113
6.3 EDEM 116
6.3.1 Computing Power 116
6.3.2 Global Model Parameters 117
6.3.3 Particle Definition 117
6.3.4 Defining the Geometry 117
6.3.5 Defining the Domain 118
6.3.6 Particle Factory 118
6.3.7 Running a Simulation 118
6.3.8 Simulation Analysis 119
6.3.9 Sensitivity Analysis 119
6.3.9.1 Sensitivity Analysis for Polyethylene Pellets 120
CHAPTER 7
CONVEYOR DISCHARGE ANGLES 127
7.1 INTRODUCTION 128
7.2 CRITICAL BELT SPEEDS 128
7.3 CONVEYOR DISCHARGE ANGLE MODEL COMPARISONS 129
7.3.1 Effect of Belt Inclination Angle on Critical Belt Speed 129
7.3.2 Effect of Belt Speed and Pulley Diameter on Conveyor
Discharge Angle 130
7.3.3 Effect of Static and Kinetic Friction 134
7.3.4 Effect of Adhesive Stress on Conveyor Discharge Angle 134
7.4 DETERMINING THE CONVEYOR DISCHARGE ANGLE 135
7.4.1 Experimental Determination 135
7.4.2 Determination by Analytical Method 136
7.4.3 Determination by Discrete Element Modelling 137
x
7.5 CONVEYOR DISCHARGE ANGLE FOR POLYETHYLENE PELLETS 137
7.5.1 Experimentally Determined Conveyor Discharge Angles 137
7.5.2 Analytically Determined Conveyor Discharge Angles 138
7.5.3 DEM Conveyor Discharge Angles 138
7.5.3.2 EDEM 139
7.5.4 Comparison of Conveyor Discharge Angles for Polyethylene
Pellets 140
7.6 CONVEYOR DISCHARGE ANGLE FOR IRON ORE 141
7.7 CONVEYOR DISCHARGE ANGLE FOR CORN 141
7.7.1 Experimental Conveyor Discharge Angles 141
7.7.2 Analytically Determined Conveyor Discharge Angles 142
7.7.3 DEM Conveyor Discharge Angles 142
7.7.3.2 EDEM 142
7.7.4 Comparison of Conveyor Discharge Angles for Corn 143
7.8 DISCUSSION 144
CHAPTER 8
CONVEYOR TRAJECTORIES 146
8.2 CONVEYOR TRAJECTORY MODEL COMPARISONS 148
8.2.1 Low-Speed Conveyor Trajectory Comparisons 149
8.2.2 High-Speed Conveyor Trajectory Comparisons 149
8.3 INFLUENCES ON CONVEYOR TRAJECTORY PROFILES 153
8.3.1 Effect of Belt Inclination Angle 153
8.3.2 Effect of Static and Kinetic Friction 153
8.3.3 Effect of Divergent Coefficients 155
8.3.4 Effect of Particle Shape and Size 156
8.3.5 Effect of Adhesive Stress 157
8.3.6 Effect of Bulk Density 157
8.4 CONVEYOR TRAJECTORIES OF POLYETHYLENE PELLETS 158
8.4.1 Experimental Conveyor Trajectories 158
8.4.1.1 Preliminary Setup 158
8.4.1.2 Laser Scanning 159
8.4.1.3 Final Setup 160
8.4.2 Analytically Determined Conveyor Trajectories 166
8.4.3 DEM Conveyor Trajectories 170
8.4.3.1 Scope of Simulations 171
8.4.3.2 Particle Geometry 171
8.4.3.3 Calibration of the Mass Flow Rate 171
8.4.3.4 Conveyor Geometry 173
8.4.3.5 Particle Parameters 174
8.4.3.6 Low Mass Flow Rate EDEM Trajectory Simulations 175
8.4.3.7 High Mass Flow Rate EDEM Trajectory Simulations 179
8.4.3.8 Further Investigation of Rolling Friction 180
8.4.3.9 Re-Visiting Low Mass Flow Rate EDEM Trajectory Simulations 183
xi
8.4.3.10 Re-Visiting High Mass Flow Rate EDEM Trajectory Simulations 183
8.4.3.11 Trajectory Simulation Comparison of 1% and 0.3 Coefficient
of Rolling Friction 184
8.4.4 Conveyor Trajectory Comparisons for Polyethylene Pellets 185
8.5 CONVEYOR TRAJECTORIES OF IRON ORE 189
8.6 CONVEYOR TRAJECTORIES OF CORN 189
8.6.1 Experimental Conveyor Trajectories 189
8.6.2 Analytically Determined Conveyor Trajectories 193
8.6.3 DEM Conveyor Trajectories 197
8.6.3.1 Particle Geometry 198
8.6.3.2 Calibration of the Mass flow Rate 199
8.6.3.3 Conveyor Geometry 200
8.6.3.4 Calibration of Rolling Friction 200
8.6.3.5 High Mass Flow Rate EDEM Trajectory Simulations 201
8.6.4 Conveyor Trajectory Comparisons for Corn 202
8.7 DISCUSSION 205
CHAPTER 9
CONVEYOR TRANSFER HOOD ANALYSIS 209
9.2 THE FLOW OF POLYETHYLENE PELLETS THROUGH A
TRANSFER HOOD 210
9.2.1 Experimental Particle Flow Investigation 210
9.2.2 Analytical Method Analysis 214
9.2.2.1 Analytical Method of Roberts 215
9.2.2.2 Analytical Method of Korzen 217
9.2.3 Discrete Element Modelling of Particle Flow 219
9.2.3.1 Chute MavenTM Simulations 219
9.2.3.2 EDEM Simulations 224
9.2.4 Method Comparisons for Polyethylene Pellets 225
9.3 FLOW OF IRON ORE THROUGH A CONVEYOR TRANSFER
HOOD 228
9.3.2 Analytical Method Analysis 231
9.3.2.1 Analytical Method of Roberts 232
9.3.2.2 Analytical Method of Korzen 232
9.3.4 Method Comparisons for Iron Ore 238
9.4 DISCUSSION 239
CHAPTER 10
PARTICLE FREEFALL 242
10.2 PARTICLE FREEFALL VELOCITY OF POLYETHYLENE PELLETS 243
10.2.1 Experimental Measurement of Freefall and Terminal Velocity 243
xii
10.2.2 Analytically Determining Terminal Velocity 246
10.2.3 Chute MavenTM DEM Simulation of Terminal Velocity 247
10.3 PARTICLE FREEFALL VELOCITY OF IRON ORE 250
10.4 PARTICLE FREEFALL VELOCITY OF CORN 252
10.5 DISCUSSION 254
CHAPTER 11
CONVEYOR TRANSFER SPOON ANALYSIS 255
11.2 FLOW OF POLYETHYLENE PELLETS THROUGH A TRANSFER
SPOON 256
11.2.2 Analytical Method Analysis of Roberts 260
11.2.4 Method Comparisons 268
SPOON 271
11.3.2 Analytical Method Analysis of Roberts 273
11.3.4 Method Comparisons 281
11.4 DISCUSSION 283
CHAPTER 12
BRING IT ALL TOGETHER 285
12.2 POLYETHYLENE PELLET COMPARISONS 286
12.3 IRON ORE COMPARISONS 292
12.4 DISCUSSION 293
CHAPTER 13
CONCLUSIONS AND FURTHER WORK 294
13.1 OVERVIEW 295
13.2 CONCLUSIONS 295
13.2.1 Conveyor Discharge Angle 295
13.2.2 Conveyor Trajectories 296
xiii
13.2.3 Conveyor Transfer Hoods 297
13.2.4 Particle Freefall 298
13.2.5 Conveyor Transfer Spoons 299
13.2.6 Discrete Element Modelling 299
13.3 FURTHER WORK 300
13.3.1 Experimental 300
13.3.1.1 Transfer Hood 300
13.3.1.2 Transfer Spoon 301
13.3.1.3 Conveyor Trajectories 302
13.3.1.4 Test Materials 302
13.3.1.5 Dust Generation 302
13.3.2 Analytical Modelling 303
13.3.3 Discrete Element Modelling 303
REFERENCES 304
BIBLIOGRAPHY 322
APPENDIX A
LIST OF PUBLICATIONS 329
A1 CONFERENCE PAPERS 330
A2 JOURNAL PAPERS 331
A3 OTHER PUBLICATIONS 331
A4 POSTER PRESENTATIONS 331
APPENDIX B
DETAILED CAD DRAWINGS OF THE TRANSFER HOOD AND
SPOON 332
APPENDIX C
MATLAB TRANSFER HOOD M-file 339
APPENDIX D
MATLAB TRANSFER SPOON M-file 343
xiv
LIST OF FIGURES
Figure 2.1 Transfer chute mechanisms 8
Figure 2.2 Element of material travelling around head pulley (Roberts, 2001) 21
Figure 2.3 Material discharge when belt and head pulley first come in
contact (Roberts, 2001) 22
Figure 2.4 Material discharge incorporating transition angle (Roberts et
al., 2004) 23
Figure 2.5 Hood design (McBride, 1997) 25
Figure 2.6 Inverted curved chute 25
Figure 2.7 (a) Rectangular cross-section, (b) circular cross-section and
(c) varying widths (Roberts, 2003) 26
Figure 2.8 Sliding on a straight chute 33
Figure 2.9 Sliding on a curved chute 33
Figure 2.10 Constant radius curved chute 35
Figure 2.11 The effect of particle shape on terminal velocity (Marcus et
al., 1990) 37
Figure 2.12 Sectional view of an AerobeltTM (Read, 1985) 41
Figure 2.13 Spherodisc representations of a tablet, (a) tablet shaped
particle, (b) 10 sphere representation and (c) 178 sphere
representation (Song et al., 2006b) 48
Figure 2.14 Examples of axisymmetrical particles 49
Figure 2.15 Bullet head nail composite particle (Nolan and Kavanagh, 1995) 49
Figure 2.16 (a) arbitrary particle shape, (b) represented by 1 sphere and
(c) represented by multiple spheres (Jensen et al., 1999) 49
Figure 2.17 Representation of a spherocylinder (Pournin et al., 2005) 50
Figure 2.18 Representation of the linear spring-dashpot model (Asmar et
al., 2002) 52
Figure 2.19 Representation of the partially latched spring model (Walton
and Braun, 1986) 53
Figure 3.1 Final conveyor transfer research facility design 66

Figure 3.2 Hood and spoon geometry of the first design - (a) CAD
design, (b) conveyor transfer research facility 67
Figure 3.3 Square feed bin assembly 70
Figure 3.4 Conveyor in position 70
Figure 3.5 Splicing the cleated belt 70
Figure 3.6a Control panel front 70
Figure 3.6b Control panel internals 70
Figure 3.7 Impact of material inside feed bin 73
Figure 3.8 Donaldson shaker unit 73
Figure 3.9 Extraction after second transfer 73
Figure 3.10 Extraction at the transfer zone 73
Figure 3.11 Extraction at feed point 73
Figure 3.12 The conveyor transfer research facility installed at new location
set for trajectory investigations 74
xv
Figure 4.1 DT80 data acquisition unit 76
Figure 4.2 Millivolt reading vs. time 78
Figure 4.3 Calibrated mass vs. time 78
Figure 4.4 Angular scale for the HoganTM valve 80
Figure 4.5 HoganTM valve calibration graph for test materials 80
Figure 5.1 Stereopycnometer 82

Figure 5.2 Sample particle and circumscribing circle 84
Figure 5.3 (a) Sample of sieves, (b) sieves in mechanical sieve shaker 85
Figure 5.4 Jenike shear tester 87
Figure 5.5 (a) Melting of polyethylene pellets to create a test sheet,
(b) partially melted polyethylene pellets and (c) the final
polyethylene pellet wall sample 87
Figure 5.6 Jenike shear tester configuration for IYL test 88
Figure 5.7 Inclination tester 89
Figure 5.8 Preparation of test material for the friction test with (a)
polyethylene pellets, (b) iron ore and (c) corn 89
Figure 5.9a Angle of repose 91
Figure 5.9b Surcharge angle 91
Figure 5.10 (a) Polyethylene pellets, (b) corn, and (c) iron ore, size range
2.36 6.3 mm 94
Figure 6.1 Three dimensional CAD model of a conveyor transfer 101

Figure 6.2 Chute MavenTM model parameters 103
Figure 6.3 Chute MavenTM simulation data 104
Figure 6.4 The results of an experimental slump test (Kamaras, 2007) 107
Figure 6.5 Comparison of (a) experimental slump test and (b) DEM
slump test with restrain = 63% (Kamaras, 2007) 108
Figure 6.6 The results of an experimental hopper test (Kamaras, 2007) 108
Figure 6.7 Comparison of (a) experimental hopper test and (b) DEM
hopper test with restrain = 88% (Kamaras, 2007) 109
Figure 6.8 Set 4 trajectory curve for mass flow rate, ms = 0.5 t/h 112
Figure 6.9 Set 4 trajectory curve for mass flow rate, ms = 5 t/h 112
Figure 6.10 Set 8 trajectory curve with coefficient of friction between
particles, p = 0.222 112
Figure 6.11 Set 8 trajectory curve with coefficient of friction between
particles, p = 0.966 112
Figure 6.12 System geometry used for sensitivity analysis 120
Figure 6.13 Steady-state EDEM outputs from the 15 sensitivity analysis
tests for spherical particles 123
Figure 6.14 Steady-state EDEM outputs from the 15 sensitivity analysis
tests for shaped particles 125
Figure 7.1 Critical belt speed for (a) Dp = 0.5 m, (b) Dp = 1.0 m,
(c) Dp = 1.5 m 131
Figure 7.2 Variation in discharge angle based on belt speed for a pulley
diameter of Dp = 0.5 m 131
xvi
Figure 7.3 Variation in discharge angle based on pulley diameter for a
belt speed of Vb = 1.00 m/s 133
Figure 7.6 Effect of adhesive stress on discharge angle 134
Figure 7.7 Setup for determining material discharge angle and trajectory 136
Figure 7.8 Low-speed discharge of polyethylene pellets at Vb = 1.0 m/s 136
Figure 7.9 Example of the discharge of polyethylene pellets using EDEM
with spherical particles 139
Figure 7.10 Comparison of polyethylene pellet conveyor discharge angles
from experiments, trajectory models and DEM 140
Figure 7.11 Comparison of corn conveyor discharge angles from
experiments, trajectory models and DEM 144
Figure 8.1 Low-speed, horizontal conveyor, lower path, pulley diameter,

Dp = 0.5 m, belt velocity, Vb = 1.25 m/s 150
Figure 8.2 Low-speed, horizontal conveyor, upper path, pulley diameter,
Dp = 0.5 m, belt velocity, Vb = 1.25 m/s 150
Figure 8.3 Low-speed, horizontal conveyor, pulley diameter, Dp = 1.0
m, belt velocity, Vb = 1.25 m/s 150
Figure 8.4 Low-speed, inclined conveyor, belt inclination angle, b = 10,
pulley diameter, Dp = 1.0 m, belt velocity, Vb = 1.25 m/s 150
Figure 8.5 High-speed, horizontal conveyor, pulley diameter, Dp = 1.0 m,
belt velocity, Vb = 3.00 m/s 151
Figure 8.6 High-speed, inclined conveyor, belt inclination angle, b = 10,
Figure 8.8 Booth Vb = 1.5 m/s 154
Figure 8.9 Booth Vb = 3.0 m/s 154
Figure 8.10 Korzen Vb = 1.5 m/s 154
Figure 8.11 Korzen Vb = 3.0 m/s 154
Figure 8.12 Variation in trajectories based on different divergent coefficients 156
Figure 8.13 Effect of particle size distribution on Korzen (1989) method 157
Figure 8.14 Effect of bulk density on trajectory profile 158
Figure 8.15 Trajectory, Vb = 1.5 m/s, ms = 24 tph 159
Figure 8.16 Laser scanned upper trajectory profile (Andrews, 2008) 160
Figure 8.17 Final conveyor trajectory test arrangement 161
Figure 8.18 Example grid referencing, Vb = 2 m/s, ms = 2.6 tph 162
Figure 8.19 Flat underside of the trajectory stream at the point of
discharge for a belt speed of Vb = 4 m/s and mass flow rate of
ms = 37.8 tph 163
Figure 8.20 Trajectory wings for a belt speed of Vb = 4 m/s and mass
flow rate of ms = 37.8 tph 163
xvii
Figure 8.21 Experimental polyethylene pellet trajectories for low mass
flow rates 164
Figure 8.22 Experimental polyethylene pellet trajectories for high mass
flow rates 164
Figure 8.23 Comparison of belt speed to material discharge velocity (the
yellow line represents the distinction between the lower and
upper halves of the particle stream) 165
Figure 8.24 Comparison of belt speed to material discharge velocity 166
Figure 8.25a Analytically determined conveyor trajectories for Vb = 1 m/s 168
Figure 8.25b Analytically determined conveyor trajectories for Vb = 2 m/s 168
Figure 8.25c Analytically determined conveyor trajectories for Vb = 3 m/s 169
Figure 8.25d Analytically determined conveyor trajectories for Vb = 4 m/s 169
Figure 8.25e Analytically determined conveyor trajectories for Vb = 5 m/s 170
Figure 8.26 Particle representations of polyethylene pellets used in EDEM 171
Figure 8.27 Mass flow rate calibration with EDEM simulating 100,000
particles 172
Figure 8.28 Calibration curves for mass flow rate of polyethylene pellets 173
Figure 8.29 Conveyor geometry imported into EDEM 174
Figure 8.30 Bins used for data extraction for (a) low mass flow rate
simulations and (b) high mass flow rate simulations 175
Figure 8.31 Graph of discharge velocities versus width of belt for
spherical particles 177
Figure 8.32 Graph of discharge velocities versus width of belt for shaped
particles 177
Figure 8.33 Graph of discharge velocities versus width of belt for
spherical particles using a 0.3 coefficient of rolling friction 182
Figure 8.34 Graph of discharge velocities versus width of belt for shaped
particles using a 0.3 coefficient of rolling friction 182
Figure 8.35 Low mass flow rate EDEM simulations for spherical and
shaped particles using 0.3 coefficient of rolling friction 183
Figure 8.36 High mass flow rate EDEM simulations for spherical and
shaped particles using 0.3 coefficient of rolling friction 184
Figure 8.37 Comparison of the low mass flow rate EDEM simulations of
spherical particles using 1% coefficient of rolling friction and
0.30 coefficient of rolling friction 185
Figure 8.38 Comparison of the high mass flow rate EDEM simulations of
spherical particles using 1% coefficient of rolling friction and
0.30 coefficient of rolling friction 185
Figure 8.39 Upper trajectory boundary for the high mass flow rates for
the experimental data and trajectory models 187
Figure 8.40 Low experimental trajectories super-imposed over the low
mass flow rate EDEM trajectories for spherical and shaped
particles with 0.3 coefficient of rolling friction 188
Figure 8.41 High experimental trajectories super-imposed over the high
particles with 0.3 coefficient of rolling friction 188
Figure 8.42 High mass flow rate trajectory streams for the trajectory
models and EDEM simulations 190
xviii
Figure 8.43 Experimental corn trajectories for high mass flow rates 191
Figure 8.44 (a) Vertical positioning of the Redlake X3 MotionPro high-
speed digital video camera for analysis of the particle
discharge velocity and (b) an example of corn for Vb = 1 m/s 192
Figure 8.45 Comparison of the particle speed of corn and the belt speed at
the discharge point of the conveyor 192
Figure 8.46a Analytically determined conveyor trajectories for Vb = 1 m/s 195
Figure 8.46b Analytically determined conveyor trajectories for Vb = 2 m/s 195
Figure 8.46c Analytically determined conveyor trajectories for Vb = 3 m/s 196
Figure 8.46d Analytically determined conveyor trajectories for Vb = 4 m/s 196
Figure 8.46e Analytically determined conveyor trajectories for Vb = 5 m/s 197
Figure 8.47 Particle representations of corn used in EDEM 198
Figure 8.48 High mass flow rate EDEM simulations for spherical and
shaped particles 202
Figure 8.49 Upper trajectory boundary for the high mass flow rates for
the experimental data and trajectory models 204
particles 205
Figure 8.51 High mass flow rate trajectory streams for the trajectory
models and EDEM simulations 206
Figure 9.1 Detail of the conveyor transfer hood 210

Figure 9.2 Material flow through the conveyor hood (a) Vb = 2 m/s and
ms = 2 tph, (b) Vb = 2 m/s and ms = 31 tph, (c) Vb = 3 m/s Pos
A ms = 2 tph, (d) Vb = 3 m/s Pos A ms = 38 tph, (e) Vb = 3
m/s Pos B ms = 10 tph, (f) Vb = 3 m/s Pos B ms = 38 tph 212
Figure 9.3 Particle tracking using Image Pro Plus 214
Figure 9.4 Average particle velocity at each angular position around
transfer hood 214
Figure 9.5a Material stream height through the hood 215
Figure 9.5b Material stream width through the hood 215
Figure 9.6 Force diagram for the inverted curved chute 215
Figure 9.7 Predicted stream velocity of polyethylene pellets through the
transfer hood using the Roberts method 217
Figure 9.8 Flow representation for analysis by Korzen 217
transfer hood using the Korzen method 219
Figure 9.10 Example output from a Chute MavenTM simulation 222
Figure 9.11 Extracted Chute MavenTM simulation outputs of the various
product feed rates for the different belt speeds and transfer
hood positions 222
Figure 9.12 Chute MavenTM DEM simulation results for all mass flow
rates 223
Figure 9.13 Conveyor transfer geometry for Vb = 2 m/s imported into EDEM 224
Figure 9.14 EDEM simulation results for transfer hood geometries 225
Figure 9.15 Comparison of methods for a belt speed of 2 m/s 226
xix
Figure 9.16 Comparison of methods for a belt speed of 3 m/s with hood
position A 226
Figure 9.17 Comparison of methods for a belt speed of 3 m/s with hood
position B 227
Figure 9.18 Iron ore dust build upon the transfer hood wings 230
Figure 9.19 Iron ore particle flow through the transfer hood 230
Figure 9.20 Transfer hood setup for Vb = 3 m/s showing excessive dust 231
Figure 9.21 Average particle velocity in the transfer hood 231
Figure 9.22 Predicted stream velocity of iron ore through the transfer
hood using the Roberts method 232
Figure 9.23 Predicted stream velocity of iron ore through the transfer
hood using the Korzen method 233
Figure 9.24 Chute MavenTM DEM simulation results for all mass flow
rates 234
Figure 9.25 Shaped representations of iron ore particles 235
Figure 9.26 EDEM simulation results for low and high mass flow rates 237
Figure 9.27 Comparison of EDEM simulation outputs for varying
coefficient of rolling friction and shear modulus 237
Figure 9.28 Comparison of methods for a belt speed of 2 m/s 239
Figure 10.1 Experimental particle freefall setup 244

Figure 10.2 Experimental and theoretical freefall results for polyethylene
pellets, (a) Vb = 2 m/s, ms = 1 tph, (b) Vb = 2 m/s, ms = 9 tph,
(c) Vb = 3 m/s, ms = 1 tph, (d) Vb = 3 m/s, ms = 9 tph,
(e) Vb = 4 m/s, ms = 9 tph 245
Figure 10.3 Comparison of experimental freefall velocity results 246
Figure 10.4 Particle freefall velocity obtained from Chute MavenTM
simulations 248
Figure 10.5 Particle freefall velocity obtained from Chute MavenTM
simulations 248
Figure 10.6 Chute MavenTM DEM freefall data from conveyor transfer
simulations 249
Figure 10.7 Experimental setup to measure the freefall velocity of iron ore 251
Figure 10.8 Experimental and theoretical freefall results for iron ore 251
Figure 10.9 Experimental and theoretical freefall results for corn 253
Figure 11.1 Detail of the transfer spoon 256

Figure 11.2 Material flow through the conveyor spoon (a) Vb = 1 m/s and
ms = 2 tph, (b) Vb = 2 m/s and ms = 2 tph, (c) Vb = 3 m/s
position A ms = 2 tph, (d) Vb = 3 m/s position B ms = 10 tph,
(e) Vb = 3 m/s position C ms = 10 tph 259
Figure 11.3 (a) Average experimental particle velocities for low mass
flow rates 260
Figure 11.3 (b) Average experimental particle velocities for high mass
flow rates 260
Figure 11.4 Force diagram for the curved chute 261
Figure 11.5 (a) Predicted average stream velocity through the spoon by
Roberts method for low mass flow rates 262
xx
Figure 11.5 (b) Predicted average stream velocity through the spoon by
Roberts method for high mass flow rates 263
Figure 11.6 Two examples of the DEM simulation outputs showing both
good and bad flow trends 264
Figure 11.7 Simulation results for all experimental belt speeds and spoon
geometries from the Chute MavenTM transfer hood and spoon
simulations 266
Figure 11.8 Simulation results for all experimental belt speeds and spoon
geometries from the Chute MavenTM transfer spoon simulations 266
Figure 11.9 EDEM simulation results for transfer spoon geometries 267
Figure 11.10 Comparison of the three analysis methods for each belt speed
and spoon geometry 270
Figure 11.11 Experimental testing of iron ore showing dust generation 272
Figure 11.12 Average particle velocity in the transfer spoon 272
Figure 11.13 Predicted stream velocity through the spoon by Roberts method 274
Figure 11.14 Simulation results from the Chute MavenTM transfer hood
and spoon simulations and the spoon only simulations 274
Figure 11.15 EDEM simulation results for low and high mass flow rates 275
Figure 11.16 Ten EDEM simulations for the high mass flow rate looking at
variations of coefficient of rolling friction and shear modulus
for spherical and shaped particles 276
Figure 11.17 The angular velocity (rpm) of particles with respect to the
vertical displacement of the particles through the transfer spoon 278
Figure 11.18 The horizontal displacement of particles across the transfer
spoon with respect to the angular velocity (rpm) of particles
through the transfer spoon 279
Figure 11.19 Front view of the particle velocity through the spoon for all
ten simulations used in these comparisons 280
Figure 11.20 Adjusted EDEM simulation results for low and high mass
flow rates 282
Figure 11.21 Average particle velocities through the transfer spoon for all
methods 282
Figure 12.1 Representation of how to estimate the impact velocity on the

spoon 287
Figure 12.2 Polyethylene pellets, Vb = 1 m/s, ms = 2 tph 288
Figure 12.6 Polyethylene pellets, Vb = 3 m/s, Position A, ms = 2 tph 290
Figure 12.7 Polyethylene pellets, Vb = 3 m/s, Position A, ms = 37.8 tph 290
Figure 12.8 Polyethylene pellets, Vb = 3 m/s, Position B, ms = 10 tph 290
Figure 12.9 Polyethylene pellets, Vb = 3 m/s, Position B, ms = 37.8 tph 291
Figure 12.10 Polyethylene pellets, Vb = 3 m/s, Position C, ms = 10 tph 291
Figure 12.11 Polyethylene pellets, Vb = 3 m/s, Position C, ms = 37.8 tph 291
Figure 12.12 Iron ore, Vb = 2 m/s, ms = 15.3 tph 293
Figure 12.13 Iron ore, Vb = 2 m/s, ms = 63.8 tph 293
xxi
Figure B1 Hood and spoon elevation view 333
Figure B2 Hood and spoon isometric and sectional view 334
Figure B3 Transfer hood 335
Figure B4 Transfer hood Polystone Ultra liner 336
Figure B5 Transfer spoon 337
Figure B6 Transfer spoon Polystone Ultra liner 338
xxii
LIST OF TABLES
Table 2.1 Divergent coefficients (Golka et al., 2007) 18
Table 2.2 Summary of some of the applications DEM has been used to
simulate, including corresponding authors 44
Table 2.3 Relative time to simulation various shapes (Potapov and
Campbell, 1998) 47
Table 4.1 Belt speed calibration chart 77

Table 4.2 Belt speed calibration check 78
Table 5.1 Particle and bulk characteristics of polyethylene pellets 93

Table 5.2 Particle and bulk characteristics of iron ore 94
Table 5.3 Particle and bulk characteristics of corn 95
Table 5.4 Shear modulus values of materials and products 96
Table 5.5 Poissons ratio values of materials and products 97
Table 6.1 Sensitivity analysis based on 100% particle restrain 110

Table 6.2 Sensitivity analysis based on 50% particle restrain 111
Table 6.3 Sensitivity analysis based on comparisons between 50% and
100% particle restrain 113
Table 6.4 Sensitivity analysis based on comparisons between 97.1%
and 85.37% particle size distributions 114
Table 6.5 Variables used to investigate the sensitivity of the Rayleigh
time step 120
Table 6.6 Sensitivity analysis settings for polyethylene pellets 122
Table 7.1 Critical belt speeds for the various methods 128
Table 7.2 Parameters used for comparisons 129
Table 7.3 Experimentally determined discharge angles for polyethylene
pellets 138
Table 7.4 Discharge angles for polyethylene pellets determined from
the trajectory models 138
Table 7.5 Discharge angles determined from the Chute MavenTM DEM
simulations 139
Table 7.6 Discharge angles determined from the EDEM simulations for
polyethylene pellets 140
Table 7.7 Experimentally determined discharge angles for corn 142
Table 7.8 Discharge angles for corn determined from the trajectory
models 142
Table 7.9 Discharge angles determined from the EDEM simulations for
corn 143
Table 8.1 Conveyor parameters used for comparisons 148

Table 8.2 Discharge velocities versus pulley diameter for a belt speed
of Vb = 3.0 m/s 151
xxiii
Table 8.3 Combinations of coefficient of static and kinetic friction used
for the Korzen method (1989) 155
Table 8.4 Range of trajectory profiles for low-speed conditions by
Booth (1934) and Korzen (1989) 155
Table 8.5 Selected equivalent spherical particle diameters for
comparison 156
Table 8.6 Experimental trajectory test setups 161
Table 8.7 Mass flow rate calibration simulation of polyethylene pellets
in EDEM 172
Table 8.8 Required number of polyethylene pellets to achieve the
experimental mass flow rates 173
Table 8.9 Belt speed settings for all EDEM simulations 175
Table 8.10 Belt speeds used to generate the correct particle discharge
velocities for spherical particle simulations for the low mass
flow rate 178
velocities for shaped particle simulations for the low mass
flow rate 178
velocities for spherical particle simulations for the high mass
flow rates 180
Table 8.13 Results of rolling friction sensitivity simulations for
Table 8.14 Additional rolling friction sensitivity simulations for
Table 8.15 Experimental trajectory test setups 189
Table 8.16 Mass flow rate calibration of corn in EDEM 199
Table 8.17 Required number of corn grains to achieve the experimental
mass flow rates 199
Table 8.18 Results of rolling friction sensitivity simulations for corn 200
Table 8.19 Additional rolling friction sensitivity simulations for corn 201
Table 9.1 Product feed rates used in experimental tests 211

Table 9.2 Chute MavenTM DEM simulation parameters 221
Table 9.3 Average stream velocity from particle friction calibration 221
Table 9.4 Experimental geometries simulated with EDEM 224
Table 9.5 Results of rolling friction sensitivity simulations 236
Table 10.1 Experimental freefall tests 243

Table 10.2 Estimates of the experimental terminal velocity of
Table 10.3 Predicted terminal velocity of polyethylene pellets 247
Table 10.4 Range of Chute MavenTM DEM simulations performed 247
Table 10.5 Predicted terminal velocity of iron ore 252
Table 10.6 Predicted terminal velocity of corn 253
Table 11.1 Product feed rates used in experimental tests 257

Table 11.2 Product feed rates used in EDEM simulations 267
xxiv
NOMENCLATURE
ALL CHAPTERS (EXCEPT CHAPTERS 2.9 2.15)
a acceleration or deceleration of material on a straight chute m/s2

a1 height to material centroid m
Ar Archimedes number -
A1 initial cross-sectional area of material m2
A2 exit cross-sectional area of material m2
Aa cross-sectional area of material outgoing from impact plate m2
Ap cross-sectional area of material inflow to impact plate m2
AP projected area of particle m2
b belt thickness m
bd width of the discharged material stream m
bw width of material at discharge point m
B width of chute m
B0 initial width of chute m
c cohesion kN/m2
C constant of integration -
C1 chute constant -
CD drag coefficient -
d* dimensionless particle diameter -
dc diameter of a circle m
dk equivalent spherical grain diameter m
dm elementary mass of material stream kg
Dsv equivalent volume diameter of a particle m
E elastic (Youngs) modulus Pa
F frictional force N
FA adhesive force N
FD drag force N
Flateral lateral force due to angled impact plate N
FN normal force N
FS shear force N
Fx horizontal component of force acting on impact plate N
Fy vertical component of force acting on impact plate N
g gravitational acceleration m/s2
G shear modulus Pa
h material depth m
h1 initial height of material m
h2 exit height of material m
hd material depth at discharge m
hp material stream depth at the moment of impact with impact plate m
H height of material in chute m
H0 height of material in chute at a particular location m
K constant of integration -
Kv pressure ratio -
Lt length of conveyor transition m
m mass flow rate kg/s
N normal force N
P1 initial material pressure Pa
xxv
P2 final material pressure Pa
Q mass flow rate tph
rp particle radius m
R constant radius of curvature of chute m
Rb radius to outer belt surface m
Rc radius of material centroid/centre m
Rd radius of discharge m
Re Reynolds Number -
Rh radius to outer depth of material surface m
Rp head pulley radius m
Rt radius of curvature of the trajectory m
s distance from head pulley axis to impact plate m
s distance around chute (equation 2.47) m
s0 distance from head pulley axis to centre of material element m
t increment time for trajectory path s
Tk kinetic frictional resistance on the belt surface N
TR Rayleigh time step s
Ts static frictional resistance on the belt surface N
U* dimensionless terminal velocity -
v velocity of mass element m/s
v acceleration of mass element m/s2
v(x) resultant velocity of inclined freefall m/s
v terminal velocity m/s
va material outgoing velocity from impact plate m/s
vb belt velocity m/s
vd discharge velocity m/s
vip velocity of material inflow to impact plate m/s
v0 discharge velocity of upper boundary, = V2 m/s
v0l discharge velocity of lower boundary, = V1 m/s
v0y initial velocity in the y direction m/s
vp vertical component of vip m/s
V terminal velocity m/s
V terminal velocity adjusted for shape m/s
V1 discharge velocity of lower boundary m/s
V2 discharge velocity of upper boundary m/s
Vb belt velocity m/s
Vcr critical velocity m/s
Vd velocity of material at discharge point m/s
Vfinal vertical component of material velocity discharging from feeder m/s
Vinitial velocity at drop height h at point of impact with chute m/s
Vp1 initial material velocity m/s
Vp2 exit material velocity m/s
VP particle volume m3
Vs tangential velocity of material at discharge point m/s
wb belt width m
X distance travelled along tangent line of belt and pulley mm
x horizontal distance at which y(x), (x) and v(x) are calculated m
x0 initial x co-ordinate for start of upper trajectory m
x1 x co-ordinates of trajectory for lower boundary m
x2 x co-ordinates of trajectory for upper boundary m
xxvi
y(x) y component of trajectory of particle freefall m
y y co-ordinate of conveyor trajectory m
y0 initial y co-ordinate for start of upper trajectory m
y1 y co-ordinates of trajectory for lower boundary m
y2 y co-ordinates of trajectory for upper boundary m
Y distance material falls below line of discharge mm
z error approximation -
z depth of conveyor transition m
GREEK
initial material discharge angle measured from the vertical
b conveyor belt inclination angle
c angle of chute at tangent point
d material discharge angle measured from the vertical
d1 material discharge angle measured from the vertical for lower
trajectory
d2 material discharge angle measured from the vertical for upper
trajectory
i angle of exiting flow to the vertical
ip angle of material inflow to impact plate
r angle at which particle slip begins to occur
angle of impact chute
specific gravity -
A contact area m2
m mass of element kg
r change in radius m
1 divergent coefficient -
2 divergent coefficient -
b divergence or dispersion coefficient of bulk stream width -
transition angle, measured from the horizontal
coefficient of restitution -
f air viscosity Ns/m2
angular coordinate of the mass element at flow-round zone of
impact plate
angle to vertical when normal force becomes zero, discharge
angle
0 angle at which material leaves belt
angle of varying width chute
coefficient of friction -
e coefficient of equivalent friction -
f absolute viscosity of air Ns/m2
i coefficient of internal friction -
k coefficient of kinetic friction -
p coefficient of external friction on impact plate -
s coefficient of static friction -
Poissons Ratio -
(x) trajectory direction angle
b loose-poured bulk density of material kg/m3
xxvii
f air density kg/m3
s particle density kg/m3
a adhesive stress kPa
kinematic angle of sliding friction
wrap angle around discharge pulley
sphericity -
A particle shape coefficient -
xxviii
CHAPTER 2 Section 2.9 to Section 2.15
CN normal damping coefficient -

CT tangential damping coefficient -
E* equivalent Youngs modulus GPa
FC cohesion force N
FN normal contact force N
FT tangential contact force N
FT* magnitude of tangential force at start of current slip plane N
G shear modulus GPa
G* shear modulus GPa
K0 initial tangential stiffness N/m
K1 spring constant for loading N/m
K2 spring constant for unloading N/m
KC cohesion constant N/m
KN stiffness of the spring in the normal direction N/m
KT tangential stiffness coefficient N/m
mi mass of particle i kg
mj mass of particle j kg
mij mass of particles i and j kg
R radius of particle m
R radius for Hertz-Mindlin model m
GREEK
relative approach (overlap) after initial contact m
0 the value of where the unloading curve goes to zero m
r empirical constant related to the coefficient of restitution -
fixed parameter -
coefficient of critical damping -
C cohesion displacement m
N displacement of particles in the normal direction m
R constant based on Poissons ratio and coefficient of friction -
T displacement of particle in the tangential direction m
Tmax maximum displacement in the tangential direction m
coefficient of restitution -
coefficient of friction -
r coefficient of rolling friction -
Poissons Ratio -
N normal component of relative velocity between particles m/s
slip slip velocity m/s
T tangential component of relative velocity between particles m/s
SUBSCRIPTS
i particle i
j particle j
xxix
CHAPTER 1 - INTRODUCTION
Chapter
1
INTRODUCTION
1
1.1 BACKGROUND
Conveyor transfers are required to direct a bulk material from a point of discharge to a
point of receival, most often via conveyor belts. Their main objective is to control the
material flow throughout the process. For many years, the design of conveyor transfers
has been seen as a black art, relying on trial and error or experience, rather than
science, to achieve the desired outcome.
The velocity of the material flow through a conveyor transfer should be controlled in
such a way that the exit velocity of the material closely matches the belt speed of the
receiving conveyor. Consequences of the flow velocity being too high includes; particle
attrition, dust generation, chute wear and excessive noise. Conversely, if the flow
velocity is too low, stagnation zones could develop, resulting in issues such as spillage
or chute blockage. Conveyor transfers generally take the form of either hood and
spoon type chutes, rock boxes or impact plates.
An unreliably performing transfer chute can be costly, especially in areas such as the
mining sector where large bulk quantities are handled and continuous production is of
utmost importance. A good transfer design from the outset can negate the costs
associated when the chute and/or conveyor belt needs to be replaced or repaired on a
regular basis due to bad design principles. Any site with belt conveyors installed
generally has multiple conveyor transfers and it is imperative that all are functioning to
maximum efficiency. The construction of a full-sized or even small-scale test rig for
validation purposes is an expense most companies would prefer to do without, with the
time and monetary requirements taking focus away from the all important material
production.
Another area within a conveyor transfer which needs attention is the conveyor
trajectory. The path of a material trajectory will dictate the design and location of a
subsequent transfer chute, whether it is a hood and spoon style, rock box or impact
plate. Trajectories are also of importance in the formation of stockpiles of bulk
materials. Numerous trajectory models have been published in the literature but little
has been published with respect to the accuracy of their predictions
2
Continuum-based analytical models have been developed to predict the flow behaviour
of bulk materials through conveyor transfer chutes (Korzen, 1989; Roberts, 1999;
Roberts, 2003). Generally analytical methods do not take into account all relevant
information, especially regarding particle parameters. This can lead to an incomplete
picture of the particle flow.
More recently, discrete element modelling (DEM) has become increasingly popular in
the analysis and visualisation of material flow through conveyor transfers to the point
where expensive test chutes may no longer need to be constructed to test designs, with
the design process occurring solely on computer workstations.
DEM has the ability to model individual interactions between particles, determining
information such as particle displacement, velocity, acceleration, rotation and force at
each minute time-step, allowing the bulk solid behaviour of a material stream to be
investigated at an unprecedented level of detail. The parameters used in DEM can be
determined from the properties of materials, such as static and kinetic friction, shear
modulus, Poissons ratio and the coefficient of restitution.
At present there is still some reluctance to rely on DEM alone due to the noticeable lack
of validation. Although the validation of DEM is somewhat lacking, there have been
several investigations of conveyor transfers, including that of Grger and Katterfeld
(2007), who have previously simulated material flow at transfer stations and verified the
results experimentally. Their main focus was the forces generated on an impact plate
and the mass flow rates through the transfer station. However, DEM validation of the
experimental particle velocity through a conveyor transfer is novel. Ilic et al. (2007)
have presented comparisons between the analytical method of Roberts (1999; 2003) and
DEM focussing on a slewing stacker transfer chute, however, there was no comparison
made to experimental results. Even though there was some agreement between the
analytical method and DEM, there can be no conclusions drawn as to whether these
methods accurately predict that which occurs in reality.
1.2 OBJECTIVES AND SCOPE OF THE RESEARCH

The scope of this research was to investigate and validate the particle flow mechanisms
within a conveyor transfer, predominantly from a particle velocity perspective. The
3
ability for either analytical models or discrete element models to produce accurate
predictions of the actual particle flow through a conveyor transfer has not been
documented in depth previously. Although various analytical models exist for conveyor
trajectories and chute flow, there is a lack of published work in the public domain
showing the use of them in industrial applications. In all likelihood these analytical
models are being used by industry but highly probably that their use and results are
being kept in-house. This confirms the need for further research into these analytical
models and their validation.
Focus was placed on granular non-cohesive free-flowing materials to allow the best
opportunity of comparison with the analytical models and also discrete element
modelling software.
The first and most important part of this research was to design and commission a
suitable experimental test facility to allow measurement of the particle flows through a
conveyor transfer and also allow for conveyor trajectories to be investigated. This test
facility needed to allow for varying belt speed and mass flow rate along with having the
ability for complete adjustment of the conveyor transfer. Sections of the conveyor
transfer needed to be constructed of acrylic to allow visualisation of the particle flow via
high-speed video capture and subsequent analysis.
The various experimental test programs were then repeated via analytical models for
conveyor trajectories and chute flows. Specific particle properties were determined to
allow for accurate predictions and comparisons.
Discrete element modelling was also investigated as a design tool, initially using the
Chute MavenTM DEM software and later with the EDEM software to account for some
limitations found with Chute MavenTM. Primarily, EDEM was to be used only in a
limited capacity, generating simulations which were not achievable with Chute
MavenTM, however, it was found that investigations of particle shape influence and the
effect of various particle parameters meant more emphasis was placed on the EDEM
simulations and resulted in a substantial contribution to the overall research contained in
this thesis.
4
1.3 ORGANISATION OF CHAPTER CONTENT

This thesis consists of 13 chapters and is broken into two main sections. The first
section builds the basis and background to the thesis. After the introduction, Chapter 2
contains the literature review, detailing many of the aspects critical to conveyor
transfers and providing detail of the analytical models for trajectories and chute flow.
Detailed information on the discrete element method is also discussed in this chapter.
Chapter 3 contains explanation of the conveyor transfer research facility and the process
undertaken to arrive at the final design. Chapter 4 follows with the calibration of the
conveyor transfer research facility, including the load cells on the feed bin and the
Hogan valve to regulate the mass flow rate. Chapter 5 explains the methods used to
determine the particle characteristics needed for the analytical modelling and DEM
simulations with summaries of the particle properties for each material included at the
end of the chapter. In Chapter 6, the discrete element modelling programs are described
and sensitivity analyses are included to evaluate the influence of several parameters.
The second part of this thesis contains the findings from the experimental work, coupled
with the results of the analytical and discrete element modelling. In Chapters 7 and 8,
the conveyor discharge angles and conveyor trajectories are investigated. Chapter 9
continues with the investigation of the material flow through the conveyor transfer
hood, with comparisons of all methods presented. Chapter 10 deals with particle freefall
and Chapter 11 follows with investigation of the conveyor transfer spoon. The results
obtained in Chapters 7 through 11 have then been combined and presented in Chapter
12 to provide an overall picture of the material flow through the conveyor transfer.
Chapter 13 completes the thesis by presenting the key conclusions and also detailed
suggestions where further work is still required.
5
CHAPTER 2 LITERATURE REVIEW
Chapter
2
LITERATURE REVIEW
6
2.1 INTRODUCTION
In years gone by there was a misconception that transfer chutes were little more than
pieces of equipment required to get product from conveyor A to conveyor B and that it
was a waste of time to engineer a transfer chute, only to regret that decision when it was
too late (Colijn and Conners, 1972). Similarly, the design of the transfer chute may have
been constrained by existing conveyors and no material flow analysis was performed to
understand the characteristics of the material in question (Winkelman, n.d.). Reviews of
transfer chute designs gave Rozentals (1991) the sense that the designer was relying on
luck to get the material to travel in the right direction or the result of the various impacts
would eventually have the material heading where it was intended. In reality, a
conveyor transfer could never deliver 100% of the material to the next step in the
process. This could be as a result of any number of issues, including; air entrainment,
material adhesion, spillage and overflow of material (Rozentals, 1991).
The design objective of an ideal transfer chute should be for the final velocity and
direction of the material passing through it to be the same as the velocity and direction
of the receiving belt (Colijn and Conners, 1972) and the normal velocity component
should be minimised to limit wear on the receiving belt and also to minimise spillage
(Roberts, 2001). Additionally, the worst case scenario for a given material should be
designed for. Often the material properties and material dynamics were not given
enough attention during the design phase (Roberts, 2001) but should have be
determined from the beginning, dictating how a specific transfer chute could be
designed successfully (Pitcher, 1981). Material properties which should be considered
include the angle of repose, bulk density, abrasiveness, flowability, particle size
distribution, moisture content, particle shape and surcharge angle.
At times chutes are produced with square or rectangular cross-sections for reasons such
as ease of fabrication and lining, although they can cause problems if the material being
conveyed has a high moisture content and/or is adhesive. When this is the case, curved
chutes are the preferred option which will aid in centralising the stream and the
momentum of the flowing stream will aid in keeping the chute clean (Stuart-Dick and
Royal, 1991).
7
There are a number of key mechanisms which occur within a transfer chute. As shown
in the simplified conveyor transfer of Figure 2.1, discharge, trajectory, impact, freefall
and chute flow models are all present, each influencing the others to some degree. These
flow mechanisms are the main focus of this research and the theory behind each will be
addressed in detail in the remainder of this chapter.
Figure 2.1 Transfer chute mechanisms
2.2 ISSUES RELATING TO CONVEYOR TRANSFER DESIGN

There are many design factors which need to be addressed when considering the
application of a given conveyor transfer. Transfer chutes are often regarded as the area
requiring the most maintenance (Alspaugh, 2004) and without careful planning and/or
lack of knowledge of conveyor transfers, some or all of the following issues can arise.
2.2.1 Conveyor Belt Wear and Damage

Wear on the surface of a receiving conveyor belt can be high due to abrasive friction if
material does not flow onto the conveyor belt cleanly. This damage can occur from
large rocks if they impact heavily in an uncontrolled fashion on the receiving belt
conveyor (Benjamin, 1999).
8
2.2.2 Spillage of Material

Uncontrolled flow through a transfer chute could result in spillage of material as a result
of non-central loading of material onto the receiving belt (Lonie, 1989;
M&J.Engineering, 1996; Winkelman, n.d.). This could cause the belt to drift, exposing
material to idlers or gaps between the belt and the skirts (if installed). The belt could
also drift to the point that it rubbed on external framework causing damage to the belt.
Skirts are commonly installed after an exit chute to contain the material from spilling
over and to locate it onto the belt as it is conveyed. Pitcher (1981) recommended that
skirt should diverge from a width of approximately one third belt width at the load point
and increase to approximately two thirds belt width. The length of the skirts should be
long enough to allow the material to become stationary relative to the movement of the
belt. Additionally, for each 1 m/s of belt speed, the skirt boards should increase in
length by 1 m.
Another possible cause of spillage is at start-up or shutdown. In these cases, material is

not travelling at full speed and depending on the design of the transfer; material may fall
short of being captured. To remedy this problem, dribble chutes are often installed
which catch this slower moving material and direct it onto the receiving conveyor
independently of the main transfer.
2.2.3 Degradation of Material

Dust generation can result from the degradation of material as it travels through a
transfer chute in a turbulent nature (Lonie, 1989) and with ever increasing laws in place
to address environmental impact, is an important issue. Material that is formed as a
result of degradation can at times become worthless within a given process and becomes
a waste material. This will certainly lead to an increase in the cost of a material to the
end user and also an extra expense for the operator as they likely need to incorporate
extra equipment to their process to separate this degraded material from the bulk (Snow,
1991).
2.2.4 Material Hang-ups

If a material has an adhesive nature, it can stick to inclined and/or vertical surfaces
within a transfer chute and over time the build up can result in blockages (Benjamin,
1999) and can even lead to failures of the structure (Lonie, 1989).
9
2.2.5 Blockage of the Transfer Chute

If the velocity of material travelling through a conveyor transfer drops enough, then a
combination of adhesion, cohesion and material size distribution can cause a chute to
block as it is unable to clear fast enough (Benjamin, 1999). This can lead to major
problems when the system is restarted with the receiving conveyor being flood fed. For
a period of time the material will be confined between the skirts (if installed) but once
the end of the skirts is reached, in all likelihood the material will spill over the sides
(Rozentals, 1991).
2.2.6 Noise Emissions

Material impacting on surfaces within a transfer chute will always generate a degree of
noise, how much noise depends on the design. If the angle of incidence is too high then
considerable impact could occur and as a result, noise. Turbulent flow through a transfer
chute can also result in excessive noise generation (Lonie, 1989).
2.2.7 High Maintenance Costs

This can result from excess wear within the transfer chute and even go as far as plant
downtime due to failures resulting from a poorly designed chute, such as a belt failure
(M&J.Engineering, 1996; Winkelman, n.d.) and as a result of having to address any
number of the issues raised above.
2.3 DUST CONTROL

As a material stream enters a transfer chute, it will entrain air which will add to the
generation of dust. Any turbulent flow through the transfer chute will aggravate this
further which can result in substantial amounts of dust being expelled to atmosphere
unless dust extraction is installed.
Dust generated from a conveyor transfer can have an adverse affect on both personal
health and plant safety. Spontaneous combustion of accumulated dust can also occur
and fugitive dust can settle on beams, pipes or other equipment and act as a fuel source
for a possible explosion (Douberly, 2003).
10
It is quite common for the cause of dust generation not to be addressed, rather the
symptoms, by installing such equipment as enclosed skirting systems and dust
collectors (Walde, 2003). The dust can be extracted from an enclosed transfer by
various means, including extraction fans and dust filters but this creates another issue,
how should the captured dust be handled? If the dust is returned to the conveyor belt
downstream from the transfer point then in all likelihood the dust issue will return at the
next transfer point. Another method of controlling dust generation is to use liquid dust
suppression chemicals which are sprayed directly onto the material, however, water
alone can be used as a sufficient suppressant for short term applications (Jones, 1998).
The fact is, simply having a controlled flow of material through a transfer chute will
greatly minimise and possibly remove dust generation entirely.
2.4 CONVEYOR DISCHARGE AND TRAJECTORY

The first two flow mechanisms present in a conveyor transfer are the conveyor
discharge and the trajectory as material travels over the head pulley. There are
numerous methods available in the literature focusing on the modelling of material
discharge and trajectory, including; C.E.M.A. (1966; 1979; 1994; 1997; 2005),
M.H.E.A. (1977; 1986) , Korzen (1989) and Booth (1934), with numerous others such
as; Golka (1992; 1993), Dunlop (1982), Goodyear (1975) and Roberts (2001). These
trajectory methods can be used for the positioning of stockpiles if that is the intended
use of a particular conveyor belt or for product travelling through a conveyor transfer.
Of course there will be interaction with the transfer chute at some point, at which time
other mechanisms come into play and need to be addressed.
There are two main trajectory types, low-speed and high-speed. Low-speed trajectories
imply that at discharge from the head pulley the material wraps around the head pulley
to some degree before trajectory. High-speed trajectories result in discharge being at the
point of tangency between the belt and the head pulley and no material wraps around
the head pulley. Each of the methods presented has formulae to determine whether low-
speed or high-speed conditions are present. In all trajectory cases the larger size fraction
of material will permeate to the outer surface of the trajectory while leaving the finer
fraction to define the lower trajectory surface (Snow, 1991).
11
All the methods listed above determine the discharge conditions of the material stream
as well as determining the trajectories for horizontal/inclined conveyor geometries.
C.E.M.A., M.H.E.A. and Goodyear are the only methods which allow calculation of the
trajectory for declined conveyor geometries, which have been excluded from this review
as the current research does not address declined conveyors.
2.4.1 C.E.M.A.
The Conveyor Equipment Manufacturers Association have released six editions of the
C.E.M.A. guide, Belt Conveyors for Bulk Materials since 1966. Each edition has seen
slight adjustments to some values in reference tables. Two reference tables are used to
determine the load height and centre of gravity for a range of surcharge and trough
angles for belt widths between 450 mm and 2400 mm and the fall distances which are
used to plot the vertical component of the trajectory path.
The tangential velocity, Vs, of the material is calculated, as shown in equation 2.1 In
equation 2.2, if the speed condition is 1, then high-speed conditions apply. If the
speed condition is < 1, then low-speed conditions apply. There is an additional case for
an inclined belt where if the speed condition = 1, the material will leave the belt at the
upper most point of the pulley. For low-speed conditions, the discharge angle, d, is
found by solving equation 2.3.
2 Rc ( RPM of end pulley )

Vs = 2.1
60
where
Rc = a1 + b + R p
2
Vs
Speed condition = 2.2
g Rc
Vs 2
d = cos 1 2.3
g Rc
To plot the trajectory, a line is drawn tangent from the discharge point at radius, Rc and
equally spaced along this line are intervals which determine the position of the vertical
fall distances. The method of calculating these intervals has been amended in the current
12
C.E.M.A. guide (2005), for high-speed conditions, the interval now equates to 1/20th of
the belt speed and for low-speed conditions the interval is calculated as 50 mm for each
metre per second of the tangential velocity. The tabulated fall distances are then
projected vertically down from each point and a smooth curve drawn through these
points to produce the centroidal trajectory curve. The lower and upper trajectories can
also be plotted, using the distance from the centroid to the belt for the lower trajectory
and the distance from the centroid to the upper material surface for the upper trajectory.
As is clearly evident from this procedure, a constant width trajectory path results,
however, the C.E.M.A. guide does make the point that for light fluffy materials, a high
belt speed will alter the upper and lower limits with both vertical and lateral spread
resulting from air resistance for such materials.
2.4.2 M.H.E.A. 1977

The Mechanical Handling Engineers Association guide, Recommended Practice for
Troughed Belt Conveyors (M.H.E.A., 1977) addresses both low-speed and high-speed
belts via the principal of centripetal acceleration. For the low-speed belt condition, the
left hand term of equation 2.4 must be 1 and the material will then wrap around the
head pulley to the discharge angle, d. For a high-speed belt the left hand term of
equation 2.4 must be > 1 and as a result, material will start its trajectory at the point of
tangent between the belt and the head pulley. Of note is the pulley radius, Rp, is used in
these calculations and excluded the thickness of the belt.
2
Vb
= cos d 2.4
g Rp
The calculations result in a plot of the lower trajectory boundary. Intervals are marked
off along the tangent line from the discharge point at divisions of 50 mm for each metre
per second of belt speed. A table supplies the vertical fall distances for each of these
divisions, at which point a smooth curve can be drawn. An approximation for the upper
trajectory can also be made by first determining the angle, d2, at which the upper
surface of the material starts its trajectory, based on equation 2.5. Following the above
method, a smooth curve can then be drawn to define the upper trajectory.
13
Vb 2 Rh
= cos d 2 2.5
g Rp 2
2.4.3 M.H.E.A. 1986

The Mechanical Handling Engineers Association guide, Recommended Practice for
Troughed Belt Conveyors (M.H.E.A., 1986) updates the method of determining the
conveyor trajectory and on close inspection is identical to the C.E.M.A. method, up to
and including the 5th edition. There are some very minor differences to the values listed
in the fall distance table and the M.H.E.A. method (1986) uses metric rather than
imperial units, resulting in some small rounding differences in the tables of data,
ultimately causing a minor variation in the final trajectory calculations.
2.4.4 Korzen
Of the methods reviewed, Korzens (1989) is the most complex in its approach. It can
accommodate adhesive materials and includes inertia, slip, static friction, s, and kinetic
friction, k into its calculations in the determination of the discharge velocity, Vd, and
discharge angle, d. For high-speed belts d = b and Vd = Vb . For low-speed
horizontal and inclined belts, the angle at which material begins to slip, r, is
determined from equation 2.6.
Vb 2
r = tan 1 s sin 1 sin ( tan 1 s ) 2 a 2.6
Rc g h
where,
m
h=
b Vb wb
h
Rc = R p +
2
The discharge angle, d, is found from equation 2.7 by first substituting the initial
conditions V ( ) = Vb and = r to determine the constant of integration. By then
substituting V 2 ( ) = Rc g cos and = d into equation 2.7 the discharge angle, d,
is found. The discharge velocity, Vd, is then found using equation 2.8.
14
2 Rc g
V 2 ( ) = Ce 4 k +

[( )
4 k 2 1 cos 5 k sin
] 2.7
1 + 16 k
2

Vd = Rc g cos d 2.8
The discharge angle and discharge velocity calculated above are for the centre height of
the material stream. The upper and lower trajectories can also be determined based on
ratios of the central discharge velocity, see equation 2.9 and 2.10.
Rp
Vd ,lower = Vd 2.9
R p + 0.5 hd

R p + hd
Vd ,upper = Vd 2.10
R p + 0.5 hd

where
m
hd =
b Vd wb
As previously mentioned, Korzen (1989) incorporates air drag into the calculations by
way of an adjusted air drag coefficient to account for shape variations. The detailed
analytical analysis developed is explained by a series of successive iterative
approximations. By providing an x coordinate, the trajectory, y(x), trajectory angle,
(x), and resultant velocity, v(x), can be determined for the freefall of a particle by
equation 2.11, equation 2.12 and equation 2.13.
g aw g
y ( x ) = x tan d x2 x3 2.11
2 Vd cos d 3 dm Vd cos d
2 2 2 2
g aw g
( x ) = tan 1 tan d x x 2

2.12
Vd cos 2 d dm Vd cos 2 d
2 2

aw 2
x g aw g
v ( x ) = Vd e dm
cos d 1 + tan d 2 x x 2

2.13
V d cos 2
d dm V d
2
cos 2
d
where,

dm = (d k ) 3 b
6
15
The first iteration determines y(x), (x) and v(x) for a free falling particle in a vacuum
(air drag effects are not considered). Successive iterations are performed until the
relative error between successive steps is no greater than 1% or 2%. Once the analysis
has been completed for a range of x values, the x and y values are plotted to produce the
material trajectory.
2.4.5 Booth
Booth (1934) found that while using the available theory to determine conveyor
trajectories, a large discrepancy was present between the theory and that in reality. After
careful investigation and confirmation of these errors, it was concluded that the existing
theory was incomplete, not taking into account the effects of the material slip. An
analytical analysis was developed to produce a more representative theory, however, is
based on a single particle only.
To determine whether low-speed or high-speed conditions apply, equation 2.2 is again

used, however, the belt radius is used instead of material centroid radius. The initial
estimate of the discharge angle for low-speed conditions is determined by again using
equation 2.3 but using the belt radius instead of the material centroid radius. If high-
speed conditions exist, the material will discharge at the point of tangency between the
belt and head pulley. If low-speed conditions exist, the material slip angle, r, is found
by solving equation 2.14.
2
V 1
cos r b = sin r 2.14
g Rb
The actual discharge angle and discharge velocity can be found from equation 2.15
using the same method as for equation 2.7.
V 2 ( )
=
(2 2
)
1 cos 3 sin
+ C e 2 2.15
2 g Rb (4 2
)
+1
Due to its complexity, Booth also reproduced this method graphically as a quicker
substitute which still showed reasonable accuracy. As there is no product height
16
defined, only a single trajectory is determined, that of the lower particle stream. There is
also no mention as to whether the material stream expands, contracts or remains
constant through the trajectory path.
2.4.6 Golka
Golkas method (1992; 1993) for determining material trajectory is based on the
Cartesian co-ordinate system. He also states from the outset that this is for materials
without cohesion or adhesion. For low-speed conditions, the discharge angle of the
lower trajectory stream, d1, is determined from equation 2.16. A unique discharge
angle, d2, is also determined for the upper trajectory stream from equation 2.17 by
determining the discharge velocity, V2, of the upper surface and the adjusted material
height, h2.
2
V1
= cos d 1 2.16
g Rp
V2 2
= cos d 2 2.17
g ( R p + hd )
where,
0.5
2h
V2 = V1 1 +
R
p
2h 0.5
hd = R p 1 + 1
R p

A unique parameter is introduced in this method, that of the divergent coefficient. Two
divergent coefficients are used, 1 for the lower and 2 for the upper trajectory streams,
which take into account variables such as air resistance, size distribution, permeability
and particle segregation. The divergent coefficients are quantified by using the
information provided in Table 2.1 which has been adapted from Golka et al. (2007).
17
Table 2.1 Divergent coefficients (Golka et al., 2007)

Flow description Material Maximum 1, 2, b*
characteristics particle
diameter
Normal weather conditions, Non-dusty > 40 mm 0 0.05
Low-speed belt < 40 mm 0.05 0.08
Very dusty < 3 mm 0.08 0.1
< 0.42 mm 0.1 0.12
Windy environment, Non-dusty > 40 mm 0.1 0.15
High-speed belt < 40 mm 0.15 0.18
Very dusty < 3 mm 0.2
* Note: the divergent coefficients can be negative or positive
The divergent coefficients, 1 and 2, are used in the determination of the trajectories
and the divergent coefficient, b, is used in determining the divergence of the trajectory
stream width, bd, as it falls away from the head pulley, see equation 2.18, where bw is
the width of material on the conveyor belt and L is the horizontal distance from the
centre of the head pulley.
bd = bw + L b 2.18
Three trajectory cases are presented, that of totally low-speed conditions (case 1,
equations 2.20 to 2.23), that of totally high-speed conditions (case 2, equations 2.24 to
2.27) and the case where the lower trajectory has low-speed conditions while the upper
trajectory is under high-speed conditions (case 3, equations 2.28 to 2.31). Each case is
based on comparisons of the upper and lower tangential velocity to that of the critical
velocity, Vcr, the point where the transition from low-speed to high-speed conveying
occurs, see equation 2.19. Once the calculations are complete, the x and y coordinates
are plotted to produce the lower and upper trajectory streams.
Vcr = g R p 2.19
CASE 1: V1 < Vcr and V2 < Vcr
x1 = V1 t cos d 1 2.20

y1 = x1 tan d 1 x1 g
2 (1 + 1 )
2 2 2.21
2 V1 cos d 1
18
x 2 = V2 t cos d 2 + x 0 2.22

y 2 = ( x 2 x 0 ) tan d 2 ( x 2 x 0 ) g
2 (1 + 2 )
+ y0 2.23
2 2
2 V2 cos d 2
where,
x0 = ( R p + hd ) sin d 2 R p sin d 1
y0 = ( R p + hd ) cos d 2 R p cos d 1
CASE 2: V1 > Vcr
x1 = Vb t cos b 2.24
(1 + 1 )
y1 = x1 tan b x1 g
2
2.25
2 2
2 Vb cos b
x 2 = Vb t cos b + x0 2.26
(1 + 2 )
y 2 = ( x 2 x 0 ) tan b ( x 2 x0 ) g + y0
2
2 2 2.27
2 Vb cos b
where,
x0 = h sin b
y 0 = h cos b
CASE 3: V1 < Vcr and V2 > Vcr
x1 = V1 t cos d 1 2.28
(1 + 1 )
y1 = x1 tan d 1 x1 g
2
2 2 2.29
2 V1 cos d 1
x 2 = V2 t + x 0 2.30
y 2 = ( x 2 x 0 ) g
2 (1 + 2 ) + y 2.31
2 0
2 V2
where,
x0 = R p sin d 1
y0 = ( R p + h2 ) - R p cos d 1
19
2.4.7 Dunlop
For high-speed conditions, the Dunlop Conveyor Manual (1982) determines the X
interval along the tangent line from equation 2.32 and the fall distance, Y, from
equation 2.33. The series of points produced is then joined to produce the trajectory
curve for the lower stream.
X = Vb t 2.32
g t2
Y= 2.33
2
A graphical method is used to determine the trajectory for low-speed conditions. The
graph consists of two charts side by side and by knowing the belt speed and pulley
diameter, it is a simple exercise to determine the discharge angle and the X interval. The
vertical fall distance, Y, is again calculated using equation 2.33. If it is found that the
desired belt speed does not intersect with the required pulley diameter then the method
for high-speed belts should be used. The graphical method for low-speed belts is also
limited to pulley diameters between 312 mm and 1600 mm.
There is reference to the material depth in the worked examples (Dunlop, 1982), but no
associated formula to calculate the upper boundary exist. There is also no specific
mention as to whether the material stream expands, contracts or remains constant
through the trajectory. However, when referring to the worked examples, the stream
appears to be converging.
2.4.8 Goodyear
The Handbook of Conveyor and Elevator Belting (Goodyear, 1975) uses the same two
equations as the Dunlop method to determine the X interval and the vertical fall
distances. For low-speed conditions, the discharge angle is determined from equation
2.34, where Rc is the radius to the centre height of the material.
Vb 2
= cos d 2.34
g Rc
20
The Goodyear Handbook defines a low-speed and high-speed condition for a horizontal
conveyor geometry and also for inclined conveyors, in a similar way to the C.E.M.A.
guide. However, the Goodyear method does not include the special inclined case put
forward by C.E.M.A. Once X, Y and d are known, it is a straight forward matter to plot
these coordinates. The trajectory plotted is for the centre height of the material stream
but there is no mention of how either the upper or lower trajectory streams are
determined. Goodyear (1975) states that the actual trajectory may be different to that of
the one calculated due to other forces acting on the particle stream which havent been
used in these calculations.
2.4.9 Roberts
Figure 2.2 depicts an element travelling around the head pulley before discharge. When
the normal force, N, becomes zero, discharge will commence and to determine the angle
at which discharge occurs, equation 2.35 is used (Roberts, 2001). The mass of the
element, m, being analysed is determined from equation 2.36 and the adhesive force,
FA, from equation 2.37.
Figure 2.2 Element of material travelling around head pulley (Roberts, 2001)
v2 F
= g cos + A 2.35
Rp m
m = b A ( h r ) 2.36
FA = a A 2.37
21
If the belt is travelling at high-speed then material will discharge from the belt at the
point where the belt and head pulley first come into contact, see Figure 2.3 (Roberts,
2001). To determine the minimum belt speed for this to still be the case, equation 2.38
is used, where b represents the belt inclination angle.

Vb = gR p cos b + a 2.38
b gh
please see print copy for image
Figure 2.3 Material discharge when belt and head pulley

first come in contact (Roberts, 2001)
Roberts (2001) assumes that air drag is negligible when determining the trajectory of
material and as such uses simplified equations of motion to determine the trajectory.
The equation of the path is given by equation 2.39, which is in the same form as for
both the Korzen (1989) and Golka (1992; 1993) methods.
gx 2
y = x tan + 2 2.39
2v cos 2
Roberts et al. (2004) presents an altered equation for the trajectory based on the fact that
as the belt transitions from the trough formed by the idlers to the head pulley, the
transition angle will influence the resulting trajectory. The transition angle is given by
equation 2.40, see Figure 2.4.
22
z
= tan 1 2.40
Lt
Figure 2.4 Material discharge incorporating transition angle (Roberts et al., 2004)
For the low-speed belt condition the angle at which material leaves the belt with respect
to the first point of contact between the belt and head pulley is denoted by the angle 0.
For high-speed belt conditions the trajectory commences at the point of tangency
between the belt and the head pulley and can be verified using equation 2.41 and it can
be assumed that v0 = v0l = vb from Figure 2.4. When air resistance is neglected the
trajectory can be determined from equation 2.42 and the radius of curvature of the
resulting trajectory is given by equation 2.43.

vb R p g 0 + cos ( + ) 2.41
b gh
gy 2
x = y tan ( + 0 ) + 2.42
2v0 2 cos 2 ( + 0 )
1.5
v gy
2
2
Rt = 0
1+ tan ( + 0 ) 2.43
g v02y

2.4.10 Trajectory Discussion

A variety of trajectory prediction methods have been presented, ranging in complexity
from basic, Dunlop (1982) and Goodyear (1975) to complex, Booth (1934), Golka
23
(1992; 1993) and Korzen (1989). Some methods include a multitude of parameters such
as C.E.M.A. (1966; 1979; 1994; 1997; 2005) and M.H.E.A. (1986) while others
incorporate parameters which no others address, such as divergent coefficients (Golka,
1992; 1993) and air drag (Korzen, 1989).
Findings by Arnold and Hill (1990) resulting from an experimental investigation found
that the Dunlop (1982) and Booth (1934) methods predicted the trajectory path more
accurately than C.E.M.A. (1966; 1979) and the C.E.M.A. (1966; 1979) and M.H.E.A.
(1986) methods seemed to wrap material too far around the head pulley for the low-
speed belt condition. Further, the Dunlop (1982), Booth (1934) and Korzen (1989)
methods resulted in higher accuracy for high-speed belt conditions. Arnold and Hill
(1990) suggest that if a material exhibits noticeable adhesion, the Korzen (1989) method
should be used but if the material is free-flowing and the material has an average
particle mass larger than 1g it is simpler to use the Dunlop (1982) or Booth (1934)
method.
2.5 UPPER TRANSFER CHUTE DESIGN OPTIONS
2.5.1 Hood Design

The upper portion of a conveyor transfer is often referred to as a hood. The main design
aspects of a hood are to catch the material stream before curving and concentrating the
material. The surface of the hood initially contacted by the material should be designed
for a low angle of incidence, less than 15, which will minimise losses in momentum,
speed and minimise such factors as material attrition, dust generation and wear
(McBride, 1997). It should also be noted that no matter how good a chute design is, if
the quality of manufacture is below par or the design is not followed, blockages can
easily result (McBride, 2000). The sides of the hood are generally angled and tapered
inwards promoting the concentration of material which is beneficial for two reasons,
firstly the stream is more compact when it interacts with the spoon below and secondly
there is less material in contact with the surface of the hood thus reducing the wall
friction generated. Figure 2.5 provides an example of a converging hood arrangement.
24
Figure 2.5 Hood design (McBride, 1997)
As explained by Roberts (2001), a hood can be analysed in much the same way as for a
standard constant radius chute (or transfer spoon), detailed in Section 2.6.3.2. Figure 2.6
presents the force diagram for an element on the chute. Ilic et al. (2007) comments that
currently there are limitations with the analytical method of Roberts, including; the
method is limited to thin stream rapid flow, the ability to analyse more chute
configurations is required and more focus on material characteristics is needed.
Figure 2.6 Inverted curved chute
The radius of curvature of the discharge trajectory is presented in equation 2.44.

However, to ensure that the material trajectory comes in contact with the constant radius
chute, the radius of curvature, Rt, of the trajectory must meet the condition of equation
2.45 (Roberts, 2001).
25
2 1.5
gx
1 + 2
Vb cos
2

Rt = 2.44
g
2
Vb cos
2

Rt R p 2.45
The coefficient of equivalent friction, e, for a chute having a rectangular cross-section

is found using equation 2.46 (Roberts, 1999; Roberts, 2001), for a circular cross-section
by equation 2.47 (Roberts, 2003) and for a chute of varying width by equation 2.48
(Roberts, 2003), see Figure 2.7 for specific details.
Figure 2.7 (a) Rectangular cross-section, (b) circular cross-section and

(c) varying widths (Roberts, 2003)
H
e = 1 + KV 2.46
B
H H
2

e = 1 + a1 + a2 + ... 2.47
B B
where a1 = 0.4 and a2 = 0.2057.
H tan
e = 1 + K v 1 + 2.48
B0 2 s tan
26
Continuity of flow is given by equation 2.49 and for the rectangular cross-section,
equation 2.46 can then be expressed by equation 2.50.
b AV = constant 2.49
C
e = 1 + 1 2.50
v
For a chute having a rectangular cross-sectional area, C1 is given by equation 2.51.
KV v0 H 0
C1 = 2.51
B
If the chutes curved section is of constant radius, R, and E is assumed to be an

averaged constant for the material stream, the velocity at a given point in the chute is
found using equation 2.52. The constant of integration, K, is found by substitution of
the initial conditions, v = v0 and = 0 . It is important to note that equation 2.52 only
applies when the condition of equation 2.53 is satisfied.
2 gR
v=
4 e + 1
2 ( 2 e2 1) sin + 3e cos + Ke 2 e 2.52
v2
sin 2.53
gR
2.5.2 Impact Plates

A deflector or impact plate is sometimes installed within a chute to assist in the
guidance of the falling stream onto the receiving conveyor and can aid in the reduction
of material plugging within the chute (Colijn and Conners, 1972).
2.5.2.1 Method of Lonie

Lonie (1989) states that he has successfully used vertical impact plates positioned in
front of the head pulley over many years and has found that the fluid mechanics
approaches of streamline and control volume best represent the flow characteristics.
From Bernoullis equation,
27
P1 V p21 P2 V p22
+ + h1 = + + h2 + losses 2.54
b 2g b 2g
The assumption is that there is equal pressure throughout the stream and that there are
negligible losses, resulting in a simplification of equation 2.54 to that displayed in
equation 2.55.
V p 2 = V p21 + 2 g ( h1 h2 ) 2.55
The cross-sectional area of the material stream can be determined from equation 2.56.
Q1 = Q2 = b AV
1 p1 = b A2V p 2 2.56
Hence,
V p1
A2 = A1
V p 2
2.57

Once the velocities are known, the resultant force exerted on the impact plate can be
determined from equation 2.58 or its modified form in equation 2.59.
F = Fx2 + Fy2 2.58
where,
Fx = QV p1 assuming that Vp1 is acting horizontally in the x-direction, and
Fy = QV p 2 assuming that Vp2 is acting vertically in the y-direction
Hence,
F = Q V p21 + V p22 2.59
These calculations are based on the premise that the material is leaving the impact plate
and thus the upper section of the transfer chute vertically. If there is any deviation from
vertical then there will be a lateral force present when reaching the lower section of the
28
transfer chute where the stream is constrained. This force is given in equation 2.60 and
the modified form of equation 2.61.
Flateral = QV p 2 sin ( i ) 2.60
Or, alternatively,
Flateral = b A2 sin ( i ) V p21 + 2 g ( h1 h2 ) 2.61
Lonie (1989) further adds that the impact plate should incorporate two wings at 45 to
the flat section of the impact plate, which will form a U-shape to help confine the
material. There will be a small quantity of material which stagnates above the material
flow which will acts as a surface over which material will flow. This stagnating material
will assist in minimising wear of the impact plate.
If the very centre of the vertical material stream is located in-line with the intersection
of the feeding and receiving conveyors then there should be no effect on the material
flow based on the direction of the receiving conveyor (Lonie, 1989).
2.5.2.2 Method of Korzen
2.5.2.2.1 Initial Conditions

An in-depth analysis of material interaction with impact plates has been provided by
Korzen (1988). The analysis was broken up into two cases, that of non-cohesive
materials and cohesive materials.
The first step of the analysis is to determine the initial conditions of the centre of the
material flow, that is, the angle, ip, and velocity, vip, of inflow of the material onto the
impact plate. This procedure is identical whether for non-cohesive materials or cohesive
materials. For high-speed belt conveyors the condition of equation 2.62 must be met and
the conditions d = b and vd = vb result. If on the other hand, a low-speed belt
conveyor is being analysed, the condition of equation 2.63 is met and the discharge
angle and discharge velocity require a complex method of determination which is not
covered by Korzen (1988).
29
vb2
cos b 2.62
Rb g
vb2
< cos b 2.63
Rb g
Continuing with the high-speed condition, the angle of inflow of the material onto the
impact plate, ip, can now be determined from equation 2.64 and the velocity of inflow
of the material onto the impact plate, vip, from equation 2.65.
ip = tan 1 tan d g ( s s0 ) vd2 cos 2 d 2.64
vip = v 2p + vd2 + 2v p vd cos ( 90 d ) 2.65
2.5.2.2.2 Non-cohesive Materials

The conditions of flow for a non-cohesive material have been analysed for three
orientations of impact plate with an impact chute angle, , they being, vertical (=0),
forward tilt (+) and backward tilt (-). In a similar fashion to the method Korzen uses
for trajectories, the outgoing velocity, va, of the material after impacting on the impact
plate is determined using an iterative approach until the relative error between
successive steps is 1% . Equation 2.66 to equation 2.71 describes the iterative method
required to determine va.
Step 1:
m
Aa1 = Ap = 2.66
vip b
va1 = vip sin 2 ( ip + ) p cos 2 ( ip + ) 2.67
Step 2:
m
Aa 2 = 2.68
va1 b
Ap
va 2 = vip sin 2 ( ip + ) p cos 2 ( ip + ) 2.69
Aa 2
30
Step i:
m
Aai = 2.70
va( i 1) b
Ap
vai = vip sin 2 ( ip + ) p cos 2 ( ip + ) 2.71
Aai
2.5.2.2.3 Cohesive Materials

The same three orientations are used for cohesive materials as were used for non-
cohesive materials again using the impact chute angle, . The outgoing velocity, va, of
the material after impacting on the impact plate is determined using an iterative
approach until the relative error between successive steps is 1% . Equation 2.66 to
equation 2.71 describes the method required to determine va. Before the iterative
process commences, the cross-sectional area of the material stream inflow to the impact
plate, Ap, the thickness of the material stream inflow to the impact plate, hp, and the
constant of integration, C, need to be determined from equation 2.72, equation 2.73 and
equation 2.74 respectively. Equation 2.75 to equation 2.86 describes the iterative
method required to determine va.
m
Ap = 2.72
b vip
m
hp = 2.73
b vip bw
2 2 gR cb gR
C =e
4 i ip
vip
1 + 16i 2 ( )
5i sin ip + 4i 2 1 cos ip + w
2 i Ap
2.74

Step 1:
R = R1 = hp 2.75
A ( ) = Aa1 = Ap 2.76
2 gR cbw gR
va1 = Ce 4 i +
1 + 16 i 2 ( )
5i sin + 4 i 2 1 cos
2 i A ( )
2.77
31
where,

in the case of a forward tilting impact plate
2

= in the case of a vertical impact plate
2

+ in the case of a backward tilting impact plate
2
m
ha1 = 2.78
b va1bw
Step 2:
hp + ha1
R = R2 = 2.79
2
m
A ( ) = Aa 2 = 2.80
b va1
2 gR cbw gR
va 2 = Ce 4 i + 2
5i sin + ( 4i 2 1) cos 2.81
1 + 16i 2 i A ( )
m
ha 2 = 2.82
b va 2bw
Step k:
hp + ha( k 1)
R = Rk = 2.83
2
m
A ( ) = Aak = 2.84
b va( k 1)
2 gR cbw gR
vak = Ce 4 i +
1 + 16i 2 ( )
5i sin + 4i 2 1 cos
2 i A ( )
2.85
m
hak = 2.86
b vak bw
Burnett (2000) discovered a discrepancy between the impact plate model of Korzen and
of that observed from experimental measurements. The Korzen model states that the
material will exit from the impact plate/chute tangential to the surface, however, Burnett
observed a degree of rebound from the chute surface which results in impact in the
32
lower section of the chute in a different location and suggests there are faults with the
theory.
2.6 LOWER TRANSFER CHUTE DESIGN OPTIONS
2.6.1 Chute Angles

Page (1991) discusses the use of angles in transfer chutes in some detail. When the
chute is designed to allow for the capture of fine material from scrapers, an angle
greater than 70 should be used to avoid material build-up (Page, 1991). If water is
added to the scenario mixed with the fines present within the material, it becomes much
harder to predict the angles required to successfully remove build-up.
The acceleration or deceleration of material on a straight chute is given by equation 2.87

(Stuart-Dick and Royal, 1991), see Figure 2.8 for the force diagram. For a curved chute,
centrifugal forces are added to the analysis and the resulting acceleration is given in
equation 2.88 (Stuart-Dick and Royal, 1991), see Figure 2.9 for force diagram.
a = g ( sin c cos c tan ) 2.87
V2
a = g ( sin c cos c tan ) tan 2.88
R
It should also be noted that the cross-sectional area of the material stream will change
throughout the chute as the product accelerates and decelerates and therefore the mass
of the material element being analysed (Stuart-Dick and Royal, 1991). This fact should
be taken into account in the analysis process to maintain accuracy.
Figure 2.8 Sliding on a straight chute Figure 2.9 Sliding on a curved chute
33
2.6.2 Rock Box

A rock box can be used to remove material impact from the chute surface (Stuart-Dick
and Royal, 1991). Page (1991) describes two possible designs for a rock box, the first
relies on a layer of material building up on which the material stream impacts, greatly
reducing wear of the transfer chute. Second, is a cascading rock box where a number of
ledges allow material to build-up as it drops through the transfer chute which are used to
reduce the momentum of material before reaching the receiving conveyor.
A rock box should be avoided if there are measurable amounts of moisture present in
the material because this would soon cause build-up and risk blocking the system.
Likewise with predominantly fine materials where a deficiency in large particles will
not aid in the self cleaning process, again risking blockage (Page, 1991).
2.6.3 Spoon Design

The design of the spoon is critical, as the concentrated material must firstly be
accelerated in the direction of the receiving conveyor and must also be centrally loaded
onto the receiving belt to avoid tracking issues (McBride, 1997).
2.6.3.1 Material Flow Through a Loading Chute

If thought is given to the angle of the transfer chute, the speed of the material will
increase as the angle approaches 90, however, this results in a steady decrease in the
horizontal velocity of the material so the angle should be maximised at 60 (Pitcher,
1981; Rozentals, 1983). If the chute angle is decreased, then the horizontal velocity
component increases at the detriment of the overall speed of the material through the
chute and thus the minimum effective angle should be set to 30 (Pitcher, 1981;
Rozentals, 1983). As a result of the slowing down of material, the thickness of the
material stream begins to increase and can reach the point where material completely
fills the chute cross-sectional area which under extreme conditions can choke the system
and cause spillage.
2.6.3.2 Material Flow in a Constant Radius Curved Chute

For fast flow through a chute (i.e. the material is travelling with such velocity that the
chute is not flood fed), the representation of the flow is given by Figure 2.10.
34
Figure 2.10 Constant radius curved chute
If the chutes curved section is of constant radius, R, and E is assumed to be an

averaged constant for the material stream, the velocity at a given point in the chute is
found using equation 2.89. The constant of integration, K, is found by substitution of
the initial condition, v = v0 at = 0 .
2 gR
v=
4 e + 1
2 ( )
1 2 e2 sin + 3e cos + Ke 2 e 2.89
2.6.3.3 Material Flow in a Parabolic Curved Chute

Roberts (2003) makes reference to chutes which have parabolic profiles which are
unable to be analysed directly with the method presented in Section 2.8.5.2. A parabolic
chute profile is presented as in equation 2.90. The corresponding radius of curvature of
this profile is given by equation 2.91.
y = Cx 2 2.90
1 2 1.5
R= 1 + ( 2Cx ) 2.91
2C
From examples provided by Roberts (2003), for the specific case shown, a parabolic
profile results in a better velocity profile than for a constant radius chute when using the
same flow parameters.
35
2.7 MATERIAL FREEFALL

In the region between the hood and spoon of a typical transfer chute there is generally a
zone where the material is in freefall. There are potentially some problems associated
with the freefall of materials including spillage, dust generation, particle degradation,
material turbulence and belt damage (Walde, 2003). The velocity of the material after
falling a distance, h, where impact on the chute occurs can be estimated by equation
2.92 (Roberts, 2001).
V final = Vinitial 2 + 2 gh 2.92
In the confines of a transfer chute it can generally be assumed that the effect of air
resistance will be quite small and as such has been omitted in equation 2.92. If, on the
other hand, air resistance is considered, equation 2.93 should instead be used,
determining the relationship between drop height, h, and velocity at impact, Vfinal,
(Roberts, 2001).
Vinitial
V 2 1 V V V
h= log e
final initial V 2.93
g V final g
1 V

The terminal velocity, V, used in equation 2.92 has been investigated in great detail in
the literature for spherical particles but to lesser extent for non-spherical particles. As a
result, there are numerous approaches, based on the determination of the drag
coefficient, CD, and/or the equation used to determine the terminal velocity. To account
for non-spherical particles, particle sphericity is commonly used as an adjusting factor.
The approach used in the Weber A model for pressure loss prediction in dense phase
pneumatic conveying can be applied (Wypych, 1993). The free settling, or terminal
velocity is determined using equation 2.94.
1
2 gvP ( s f ) 2
V = 2.94
CD AP f
36
The drag coefficient, CD, in equation 2.94 correlates with the Reynolds Number, which
is determined from equation 2.95. Subsequently, the drag coefficient is determined from
one of the conditions of equation 2.96 using an iterative process.
V f d v
Re = 2.95
f
24 3Re
CD = 1 + , for Re 1.0
Re 16
26
CD = 0.4 + , for 1 < Re 104 2.96
Re 0.8
CD = 0.4 , for 104 < Re 105
Once the terminal velocity for a spherical particle has been found, the actual dimensions
of the particle are used to determine the sphericity factor, shown in equation 2.97.
2
6VP 3
= 2.97
Ap
This sphericity factor is then used in conjunction with Figure 2.11 (Marcus et al., 1990),
to determine the shape adjusted terminal velocity, V.
Figure 2.11 The effect of particle shape on terminal velocity (Marcus et al., 1990)
37
The work of Pettyjohn and Christiansen (1948), later expanded upon by Geldart (1990),
considers two distinct ranges of Reynolds Number, the Stokes range (i.e. Re < 0.2)
and the Newtonian range (i.e. 1000 < Re < 3x105). The determination of the terminal
velocity includes the particle sphericity and either equation 2.98 or equation 2.99 is
used, dependent on the Reynolds Number.
( s f ) gd v
2
VST = 0.843log10 2.98

0.065 18
1
4 ( s f ) gd v 2
VN = 2.99
3 ( 5.31 4.88 ) f
Turton and Levenspiel (1986) reviewed previous drag correlations and developed
equation 2.100, which resulted in less error than the existing correlations for the entire
range of Reynolds Number.
24 0.413
CD =
Re
(
1 + 0.173Re 0.657 + )
1 + 16300 Re 1.09
2.100
Turton and Clark (1987) investigated the terminal velocity of spherical particles and
determine the dimensionless terminal velocity, U*, from equation 2.102, in terms of
only the dimensionless particle diameter, d*, equation 2.101.
1
g f ( s f ) 3
d* = d v 2.101
2
1
1 0.824
U* = 0.824 0.824
2.102
18 0.321
+
d*
2
d*

The dimensionless terminal velocity can also be expressed as shown in equation 2.103,
which allows the dimensional terminal velocity of a sphere to be determined. The
method employed by the Weber A model for non-spherical particles can then be
applied.
38
1
f 2 3
U * = V 2.103
g ( s f )
Haider and Levenspiel (1989) presented two methods of determining the drag
coefficient, one for spherical particles, equation 2.104, and one for non-spherical
particles, equation 2.105. For spherical particles, the terminal velocity can then be found
as per the Weber A model. For non-spherical particles, the particle sphericity has been
included in the determination of the four drag equation parameters and therefore no
further adjustments to the resulting terminal velocity are required.
24 0.4251
CD =
Re
(
1 + 0.1806 Re 0.6459 + )
6880.95
2.104
1+
Re
24
CD = (1 + A Re B ) + CD 2.105
Re 1+
Re
where
(
A = exp 2.3288 6.4581 + 2.4486 2 )
B = 0.0964 + 0.5565
(
C = exp 4.905 13.8944 + 18.4222 2 10.2599 3 )
(
D = exp 1.4681 + 12.2584 20.7322 2 + 15.8855 3 )
Karamanev (1996) believes that there is more merit in describing the drag coefficient,
equation 2.106, as a function of the Archimedes Number, equation 2.107, rather than
the Reynolds Number to provide a more explicit solution. On inspection of these two
equations, it is evident that to determine either the drag coefficient or the Archimedes
Number, the other parameter must first be determined. This will result in an iterative
approach to finding the solution.
432 (
1 + 0.0470 Ar 3 ) +
2 0.517
CD = 1
2.106
Ar 1 + 154 Ar 3
3
Ar = Re 2 CD 2.107
4
39
Brown and Lawler (2003) undertook an extensive review of past sphere drag research
and extracted a large selection of meaningful data from the literature to evaluate the
existing correlations. They then developed their own equation for dimensionless settling
velocity which showed much better fit than the existing models for Re < 5000,
equation 2.108. The method of Turton and Clark (1987) can then be followed to
determine the true terminal velocity of spherical particles.
d*2 ( 22.5 + d*2.046 )

U* = 2.108
0.0258d*4.046 + 2.81d*3.046 + 18d*2.046 + 405
Swamee and Ojha (1991) also developed a drag coefficient for spherical particles,
equation 2.109,
0.25
24 1.6 0.72 2.5 0.25
130 40000
CD = 0.5 16 + + + 1 2.109
Re Re Re
Majumder and Barnwal (2004) present drag coefficients for ten distinct ranges of
Reynolds number. Equation 2.110 shows the range which will be required in the
investigations covered in Chapter 10, i.e. 1500 < Re < 12000.
log10 CD = 2.4571 + 2.5558 0.929 2 + 0.1049 3 2.110
Tran-Cong et al. (2004) developed a new drag coefficient for irregularly shaped
particles, shown in equation 2.111, where c is the particle circularity, given by equation
2.112. The use of this drag coefficient is limited to the following ranges, Re < 1500,
0.8 < ds/dv < 1.5 and 0.4 < c < 1.0.
2
d
0.42 s
24 d s 0.15 d s
0.687
CD = dv
1 + Re + 2.111
Re d v c dv d
1.16

c 1 + 42500 s Re
dv
ds
c = 2.112
PP
40
Once the drag coefficient has been determined for either Tran-Cong et al. (2004),
Swamee and Ojha (1991) or Majumder and Barnwal (2004), the Weber A method can
be used to determine the terminal velocity for spherical particles and then the sphericity
adjustment can be applied.
2.8 AIR SUPPORTED BELT CONVEYORS

The concept of supporting a conveyor using air dates back to the late 1800s and early
1900s, however, the designs were not practical and never marketed. They re-emerged
in the 1970s and have continued to this day (C.E.M.A., 2005). The principle is that a
continuous flow of air permeates from a series of holes along the axis of a troughed
plenum chamber creating a cushion between the plenum and the belt.
The AerobeltTM is a commercially marketed air supported belt conveyor which is

modular in construction. With standard head and tail section lengths dependent on the
width of conveyor belt required, the troughed section can be made to suit any length
required by the joining of multiple sections. The troughed section is curved based on the
width of the belt and is supported by a totally sealed unit, see Figure 2.12.
Figure 2.12 Sectional view of an AerobeltTM (Read, 1985)
Holes are located along the centre of the trough over the entire length and air is supplied
to the sealed unit via a centrifugal fan or blower (Thain and Broumels, 1983). An added
bonus of the air cushion is that it is near impossible for material to get caught between
41
the trough and the belt. Coarse materials up to a size of approximately 200 mm can be
successfully conveyed depending on the amount of fines present and speeds of up to 7
m/s are achievable (Read, 1985).
Use of a pressure switch mounted on the sealed unit monitors the plenum pressure. If
this pressure is insufficient to produce an air film between the trough and the belt, the
motor is electrically locked out (Thain and Broumels, 1983).
It is stated by Thain and Broumels (1983) that the material being conveyed should
consist of a breakdown of 75% fine material (less than 100mm) and 25% course
material (100mm and above) to allow the AerobeltTM to satisfactorily operate without
risk to the air cushion.
Advantages to using an air supported belt conveyor include, few moving parts which
reduces maintenance, lower and cheaper specification belts can be used, belt speeds in
excess of 5 m/s are achievable, high surcharge angles are possible which increases the
conveying rate, higher conveyor inclination angles, no belt sag, greatly reduced
frictional and tensional forces (Thain and Broumels, 1983), the smooth trough removing
load disturbances normally encountered with roller idlers, the smooth trough minimises
segregation of load on the belt in turn reducing segregated trajectory off the head pulley
and the trough shape can be optimised to produce a maximum capacity of material. A
safety benefit is found in the fact that there is a substantial reduction in the number of
nip points present (Read, 1985).
A negative of an air supported belt conveyor is that the alignment of belts and pulleys is
more critical. With troughed idlers there is a degree of self-training available due to the
ability to incline them if need be (Read, 1985).
The initial cost of an air supported belt conveyor system is comparable, if not slightly
higher, to conventional roller idler belt conveyors largely due to the highly labour-
intensive fabrication involved in producing the troughs (Read, 1985), however, over the
life of the conveyor there can be up to a 75% reduction in ongoing maintenance costs
(Thain and Broumels, 1983).
42
2.9 DISCRETE ELEMENT MODELLING

The discrete element method (DEM) was developed in the early 1970s for the purpose
of analysis of problems relating to rock and soil mechanics (Cundall and Strack, 1979).
The method involves determining the individual dynamics for all the particles in a given
assembly, firstly by determining the total forces acting on each particle and then
applying Newtons 2nd law of motion. This will yield the velocity and displacement of
each particle and by balancing the angular momentum of each particle, the rotation can
be determined.
The DEM component of this research does not involve the development of a new DEM
code and as such the subsequent literature review will not be an exhaustive list of all the
required formulae to program a source code but instead will focus on some of the key
facets of DEM as a whole.
2.10 APPLICATIONS OF DEM

As DEM has developed and become more widely known, there has been an ever
increasing number of applications to which DEM has been successfully applied, see
Table 2.2.
2.11 CALIBRATION AND VALIDATION OF DEM

From the available literature focussing on the application of DEM, it has been noticed
that there is lack of experimental calibration and validation to justify some of the
findings obtained. Li et al. (2005) found this especially true of users who developed
their own codes. However, there have been some researchers who have calibrated and
validated their DEM investigations with experimental test programs (Theuerkauf et al.,
2003; Pinson et al., 2004; Shiu et al., 2006; van Liedekerke et al., 2006; Song et al.,
2006a; Song et al., 2006b) and others who have only completed validations without
calibration (Xiang and McGlinchey, 2003). It is also important to remember that even
though DEM simulations can prove to be accurate for a particular application or
component of a larger system, there is no guarantee that the DEM is accurate for that
same entire system (McBride et al., 2004). This could be for reasons such as the contact
model used or the particle size and shape selected for the simulations.
43
Table 2.2 Summary of some of the applications DEM has been used to simulate,
including corresponding authors
A range of mills (Cleary, 1998; Cleary and Sawley, 1999; Cleary, 2000;
(including ball, grinding, Cleary and Hoyer, 2000; Cleary, 2001; Govender et al.,
centrifugal, SAG and AG) 2001; Mishra and Murty, 2001; McBride et al., 2004;
Morrison and Cleary, 2004; Chandramohan and Powell,
2005; Cleary et al., 2006; Kwan et al., 2006)
A range of mixers (Cleary, 1998; Cleary et al., 1998; Cleary, 2000; Mustoe et
(including tumbling, al., 2000; Cleary et al., 2002; Kuo et al., 2002; Bertrand et
blade and V) al., 2005; Gantt et al., 2006; Li et al., 2006)
Shear cell testers (Walton, 1982a; Walton, 1982b; Jensen et al., 1999;
(including Jenike, triaxial Theuerkauf et al., 2003; Castro Vazquez et al., 2006; Ji et
and ring, 2D and 3D) al., 2006; Luding, 2006)
Conveyors (Walton et al., 1988; Walton, 1993; Qiu and Kruse, 1997;
(including chute flow, Tillack and Zhang, 1999; Cleary, 2000; Alspaugh et al.,
transfers, bucket 2002; Dewicki and Mustoe, 2002; Dewicki, 2003; Krause
elevators, screw and and Katterfeld, 2004; Pinson et al., 2004; Zhu and Yu,
sampling) 2004; Katterfeld and Grger, 2006; Anderson and Britton,
2007; Ilic et al., 2007)
Pneumatic conveying (Han et al., 2003; Xiang and McGlinchey, 2003;
(including horizontal, McGlinchey et al., 2004; Fraige and Langston, 2006; Golz
vertical and cyclones) and Favier, 2006; Chu and Yu, 2007; Straub et al., 2007)
2D and 3D hoppers (Walton, 1982a; Walton, 1982b; Langston et al., 1995;
Cleary, 1998; Tillack and Zhang, 1999; Masson and
Martinez, 2000; Cleary and Sawley, 2002; Theuerkauf et
al., 2003; Li et al., 2004; Zhu and Yu, 2004; Goda and
Ebert, 2005; Ketterhagen et al., 2006; Markauskas, 2006;
Balevicius et al., 2008)
Pharmaceutical industry (Fu et al., 2006; Pandey et al., 2006; Song et al., 2006a;
Song et al., 2006b)
Some researchers have used high precision instrumentation such as biplanar

angioscopes to track particles in grinding mills (Govender et al., 2001) or Positron
Emission Particle Tracking to track the positions of particles within a rotating drum (Li
et al., 2006) which they have then used to compare to their DEM simulations. Some
researchers relied on experimental work previously performed by others (Liu et al.,
1999; Zhang and Vu-Quoc, 2000; Di Maio and Di Renzo, 2004), while others have
discovered variations between their DEM simulations and experimental findings, put
down to issues such as simplifications of the model used (Tsuji et al., 1993), or how key
components of an overall model were defined (Xiang and McGlinchey, 2003).
High-speed and low-speed conveyor trajectories have been investigated by Nordell

(2003), comparing results of DEM simulations to those generated using the C.E.M.A.
44
method. It was found that the DEM results matched C.E.M.A. well for high-speed
conveying but there is a noticeable difference for low-speed conveying due to the low
stress state of particles present in the flow zone, that is, the flow of the upper path
reclines to the angle of repose of the material in DEM.
Vu-Quoc et al. (2000) performed physical experiments using nearly ellipsoid soybean
particles, where they were dropped from various heights onto materials such as
aluminium, acrylic and glass. Experimental tests were also carried out on soybeans
flowing down an inclined chute. The simulations showed a good qualitative agreement
to the experimental results.
Katterfeld and Groger (2006) validated DEM simulations of transfer chutes and bucket
elevators by performing experimental tests using clayey spathic iron ore on a conveyor
belt investigating free fall impacts. The particles had a maximum diameter of 1 m
falling a distance of 5 m and were simulated using either rigidly clustered cubes or
pyramids. The simulated loads showed good agreement with actual tests. The geometry
of an experimental bucket elevator was reproduced for DEM simulation and the results
showed good agreement with the video footage of the conducted experiments.
The calibration of bench-scale tests, such as angle of repose, has been performed by
Franz (2007) before advancing to the purpose of his research, that being the conveying
of coal through a complex conveyor transfer system consisting of multiple feeding and
receiving belts as well as single and bifurcating transfers. The conclusion of this work
was that the DEM was of great help in evaluating the chute design during the planning
stage. The degree to which the DEM varies to the actual situation should not be
forgotten in this design stage and also be mindful of the existing practical bulk-materials
handling experience at hand. Groger and Katterfeld (2006b) have begun investigating
the angle of repose formed inside a rotating drum from the perspective that both the
coefficient of friction and coefficient of rolling friction will apply in a different way to
other applications, such as in the shearing of a consolidated system. It is their intention
to correlate the results from these two results should find an intersecting point where the
frictions are the same, which can then be used for a wider set of applications.
45
Ilic et al. (2007) compared DEM simulations of a slewing stacker for both the hood and
spoon with the analytical method of Roberts (2001; 2003). Unfortunately no
comparison was made to experimental data.
Coetzee and Els (2009) have investigated silo discharge and bucket filling of corn by
first investigating small scale particle effects. The corn kernels were simulated by
clustering two spheres of differing diameter, which may not seem highly representative;
however, they believe that this was accurate enough for their investigations and also
results in a substantial simulation time saving. Calibration of the DEM software was
via, direct shear tests, compression tests and angle of repose determinations. The overall
results of these investigations were that with initial calibration of material properties,
both the DEM simulation of silo discharge and bucket filling can be modelled
accurately.
2.12 PARTICLE SHAPE IN DEM

Traditionally, two-dimensional discs and three-dimensional spherical particles have
been used to represent particles within DEM simulations but with the ever increasing
computational power comes the ability to generate more realistic particle shapes, in turn
producing even more realistic outcomes. Some researchers are now digitally generating
complex particle shapes using; x-ray micro-tomography (Gan et al., 2004), three-
dimensional particle imaging (Bujak and Bottlinger, 2008) and three-dimensional laser
ranging (Latham et al., 2008) which can capture realistic geometries and be used in
shape libraries. For any complex shape, their simulated formation is not the problem
per-se, rather the contact detection between neighbouring particles and the resulting
calculations for the overlap area, intersection points, points of contact and the normal
and tangential contact vectors (Dziugys and Peters, 2001).
Potapov and Campbell (1998) recorded the time required to simulate 36 particles of
various shapes dropping within a tube, setting spherical particles as the base shape and
assigning a value of 1 as the time taken. The respective times for the other shapes are
displayed in Table 2.3.
46
Table 2.3 Relative time to simulation various shapes (Potapov and Campbell, 1998)
Shape Simulation time

Round 1
Quasi-triangle 1.7
Quasi-square 1.8
Superquadric 12
Square 3.9
The following sections detail particle shapes which have been modelled in DEM
simulations and provide specific examples and issues which may arise.
2.12.1 Circular and Spherical Particles

Without doubt, two dimensional discs or three dimensional spherical particles are the
most commonly used in DEM simulations (Campbell and Brennen, 1985; Walton et al.,
1988; Sadd et al., 1993; Drake and Walton, 1995; Cleary et al., 1998; Iwashita and Oda,
1998; Liu et al., 1999; Sitharam, 1999; Cleary and Sawley, 2002; Dewicki, 2003;
Loughran et al., 2004; Chandramohan and Powell, 2005; Castro Vazquez et al., 2006;
Katterfeld and Grger, 2006). These simulations are much faster to perform than for any
other shape due to any point on the surface of the particle being determined by knowing
the diameter and the coordinates of the centre (Potapov and Campbell, 1998). Further,
the particle overlap is simple to determine by subtracting the distance between particle
centres from the sum of the radii of the two particles.
There are downsides, however, to using 2D discs or 3D spherical particles. Even though
they greatly minimise computational time, simulations can vary to experimental results
if they are being used to represent a particle which is not round (Favier et al., 1999;
Cleary, 2000; Li and Holt, 2005). This statement is backed up by the work of Cleary
(1998) where he simulated near circular particles in a centrifugal mill and found good
agreement with actual data for near spherical particles, however, when salt cubes were
tested and simulated using near circular particles, the results were poor. Although this
result may seem obvious, it is a clear indication that shape is an important factor when
preparing a DEM simulation. Others have addressed the issue of non-sphericity and
attempted to overcome it by using an increased rolling resistance of the particles to
produce realistic flow behaviour and acceptable performance (Zhou et al., 1999; Pinson
et al., 2004).
47
2.12.2 Ellipsoid Particles

The ellipsoid is a shape which produces a good representation of real materials such as
rice and wheat and can be used in both 2D and 3D (Lin and Ng, 1995; Lin and Ng,
1997; Pinson et al., 2004) applications. Favier et al. (1999) highlight that in work by
other researchers, the use of elliptical particles has increased the accuracy of simulations
over using spherical particles when investigating the compression of sand. Vu-Quoc et
al. (2000) clustered spheres together to form ellipsoid particles for work in their
simulations.
2.12.3 Multi-Sphere Particles

Spherodisc particles were used to investigate the discharge of particles from rectangular
hoppers. The experimental and DEM comparisons were performed using Nestle Giant
Smarties having an average diameter of 19.22 mm and a height of 9.75 mm (Li et al.,
2004). Tablet shaped particles were modelled as spherodiscs using multi-sphere
representations of 10, 26, 66 and 178 spheres (Song et al., 2006a; Song et al., 2006b)
and can be seen in Figure 2.13.
Figure 2.13 Spherodisc representations of a tablet, (a) tablet shaped particle, (b) 10
sphere representation and (c) 178 sphere representation (Song et al., 2006b)
The investigations performed by Markauskas (2006) used various complex

axisymmetrical particles all consisting of a number of spheres. Each sphere within a
particle is located on the symmetry axis and are allowed to overlap or have different
diameters, see Figure 2.14 for examples, but are always in a fixed position within the
particle throughout the simulation. Favier et al. (1999) also uses a range of
axisymmetrical particles in their DEM simulations.
48
Figure 2.14 Examples of axisymmetrical particles
Composite particles are produced by combining multiple spheres together to form a

variety of shapes. Nolan and Kavanagh (1995) investigated random packing using a
variety of composite particles, including metal bullet head nails, as shown in Figure
2.15.
Clusters are used to model rough particles by Jensen et al. (1999). An arbitrary shape as
shown in Figure 2.16(a) could be modelled using a circular particle, Figure 2.16(b),
however, this will not produce a representative alternative and so multiple smaller
circular particles are clustered together inside the original shape, Figure 2.16(c). These
joined particles are then defined to behave as a rigid body with no relative movement
between them.
Kruggel-Emden et al. (2008) use a multi-sphere approach which can be used to create
representations of numerous shapes. Their investigations have shown that experimental
and simulated results compare well but are dependent on the orientation of the
composite particles.
Figure 2.15 Bullet head nail composite Figure 2.16 (a) arbitrary particle shape,
particle (Nolan and Kavanagh, 1995) (b) represented by 1 sphere and
(c) represented by multiple spheres
(Jensen et al., 1999)
49
2.12.4 Spherocylinders
A spherocylinder in simple terms is a cylinder with half a sphere attached to either end
as shown in Figure 2.17. The contact model required to simulate spherocylinders is
complex and is explained in detail by Pournin et al. (2005). Numerous other researchers
have also used spherocylinders, including Cleary and Hoyer (2000), Langston et al.
(2004) and Gan et al. (2004).
Figure 2.17 Representation of a spherocylinder (Pournin et al., 2005)
2.12.5 Superquadrics
By using superquadrics, a wide range of particle shapes can be generated by changing
the variables present in equation 2.113 (Cleary et al., 2006), which is the generalised
form of an ellipse.
m m m
x y z
+ + =1 2.113
a b c
The semi-major axes of the superquadric are b/a and c/a and the power m dictates the
particle shape. When m = 2, a sphere results and as m is increased the corners of the
superquadric sharpen until when m = 10 the shape converges to a cube if an aspect ratio
of 1 exists for the semi-major axes (Cleary et al., 2006). A drawback to using
superquadrics is that processing time can increase by a factor of three over spherical
particles (Cleary et al., 2006).
2.12.6 Polygons
Two-dimensional polygons are defined by corners, edges and cosine directors of the
axes whereas 3D polygons (or polyhedrons) are defined by their corners, edges and
50
faces and vectors from the centre of mass locate their corner positions (Hogue, 1998). In
3D, the addition of faces substantially increases the number of contacts possible and
hence the number of computations required. If a mixture of polygons and smooth
particles are to be simulated together, this can prove disadvantageous. A smooth particle
could be defined with many corners to allow for a polygonal contact detection to take
place, however, this will greatly increase computation times (Hogue, 1998).
The use of polygons to define particle shape will yield a more realistic solution,
although, simulation times can be as much as ten times longer than for round particles
and using particles in the form of ellipsoids or superquadrics will also increase
computing time as the particle overlaps need to be determined by solving the non-linear
equations over many iterations (Potapov and Campbell, 1998).
2.13 CONTACT FORCE MODELS

There are numerous contact force models available to simulate particle-particle and
particle-wall interactions. These are all classified as soft sphere contact models where
contact mechanics are used to quantify the displacements, velocities and forces of the
particles. The time step used is only a fraction of an actual collision and more than one
collision is possible at each time step. Also, it is assumed that the particles which collide
do not undergo any deformation but instead overlap to a small degree (Di Maio and Di
Renzo, 2004). The other type of contact model is the hard sphere model where a
collision is seen to be instantaneous. The conservation of energy and momentum and
surface sliding are applied to quantify the velocity of the particles which are colliding
before and after the impact (Di Maio and Di Renzo, 2004). The following sections will
detail several of the more widely recognised contact force models.
2.13.1 Linear Spring-Dashpot Model

Many researchers have used the linear spring-dashpot model, Figure 2.18, including
Haff and Anderson (1993), Hogue and Newland (1994), Jensen et al. (1999), Li et al.
(2004), Asmar et al. (2004), Loughran et al. (2004), Morrison and Cleary (2004),
Bertrand et al. (2005) and Gantt et al. (2006). Generally the model is defined as that
shown in equation 2.114 and equation 2.115, the first component of each is the elastic
force and the second is the damping force.
51
Figure 2.18 Representation of the linear spring-dashpot

model (Asmar et al., 2002)
FN = K N N + CN N 2.114
FT = KT T + CTT 2.115
where,
C N = 2 mij K N , 2.116
ln ( )
= 2.117
2 + ln 2 ( )
mi m j
mij = 2.118
mi + m j
CT = 2 mij KT 2.119
There can be variations to equation 2.114 and in the case of Asmar et al. (2002), they
include a cohesion force in the determination of the normal contact, measured using
Hookes linear spring relationship and being a measure of the attractive force between
two particles, see equation 2.120.
FC = K C C 2.120
Asmar et al. (2002) also show how equation 2.115 can be modified, relative to the
Coulomb friction limit. Once this limit has been reached, the surface contact between
the particles shears and the particles then begin to slide over one another. This is known
as gross sliding. To check whether the Coulomb frictional limit has been reached,
52
equation 2.121 is used. If T > Tmax then sliding occurs and equation 2.123 is used to
determine the frictional force after gross sliding. For all other cases, equation 2.115 is
used as before.
T max = R N 2.121
where
R =
( 2 ) 2.122
2 (1 )
FT = K N N 2.123
2.13.2 Partially Latched Spring Model

Originally used by Walton and Braun (1986), the partially latched spring model is used
to avoid the use of the damping coefficient, see Figure 2.19. This is achieved by using
different stiffness coefficients, KN, for the loading and unloading portions of a collision
(Bertrand et al., 2005). The resulting equations for the normal force component are
given by equation 2.124 and the tangential contact force is given by equation 2.125.
Figure 2.19 Representation of the partially latched spring model

(Walton and Braun, 1986)
K1 for loading
FN = 2.124
K 2 ( 0 ) for unloading
FT = KT T 2.125
53
where

F FT *
K 0 1 T for FT increasing
FN FT *
KT =
2.126
FT * FT
K 0 1 for FT decreasing
FN + FT *
2.13.3 Non-Linear Spring-Dashpot Model

The non-linear spring dashpot model is similar to the linear spring-dashpot model,
however, for the linear model, = 1 and for the non-linear model, = 3/2, see equation
2.127 and equation 2.128.
FN = K N N C N N 2.127

FT = min KT T , FN T CT slip 2.128
T
where
4
KN = RE * 2.129
3
C N = r ( mij K N ) 2 N
1 1
4
2.130
KT = 8 R N G* 2.132
CT = CN 2.133
Hemph et al. (2006) used both the linear model and the non-linear model to model
hopper flow. The results of the simulations showed that there was little difference
between the mass flow rates generated with the two models but the non-linear model
compared slightly better to the experimental results.
2.13.4 Hertz-Mindlin Model

As more research progressed in the area of DEM, investigations into the contact forces
used in the modelling increased. More detailed contact force models were developed
such as the Hertz-Mindlin model (Cundall, 1988; Castro Vazquez et al., 2006).
54
Equation 2.133 and equation 2.134 present the normal and tangential components of the
contact forces.
FN = K N N 2.133
FT = KT T 2.134
where

2G 2 R
KN =
3 (1 v ) N
2.135

2

1
3
2 G 3 (1 v ) R
KT = F 1
3
2.136
2v N

For ball-ball contact

2 Ri R j
R= 2.137
Ri + R j
1
G=
2
( Gi + G j ) 2.138
1
v=
2
( vi + v j ) 2.139
For ball-wall contact
R = Rball 2.140
G = Gball 2.141
v = vball 2.142
2.13.5 Improved Hertz-Mindlin Model

Di Renzo and Di Maio (2005) present a modified Hertz-Mindlin model as an outcome
of obtaining poor results in the analysis of forces, velocities and displacements. When
reviewing the tangential forces, as in equation 2.128 above, it was found that a better
approximation is given by equation 2.142.
55
FT =
2
3( 1
8G R N 2 T ) 2.143
2.13.6 Hertz-Mindlin Model Without Slip

In the work of Tsuji et al. (1992) the normal force component is the same as that for the
standard Hertz-Mindlin contact model explained above, however, for the tangential
force component, if there is slip present at the contact surface the relationship is
complicated. This led to modifying the tangential force component to remove slip from
the analysis. The tangential stiffness then simplifies to that shown in equation 2.144.

2G 2 R
KT =
2v N
2.144

2.13.7 Hertz-Kuwabara-Kono Contact Model

Dutt et al. (2005) detail numerous models for the determination of normal and
tangential contact forces and comment that a previous study found the Hertz-Mindlin
model showed some unusual behaviour for the coefficient of restitution of particles
when compared to experimental results. These findings lead to the development of a
more accurate model, namely the Hertz-Kuwabara-Kono (HKK) model for the normal
contact forces, see equation 2.145. The tangential contact force remains the same as for
the linear spring-dashpot model.
3 1
FN = K N N 2
+ CN N 2 N 2.145
2.13.8 Effect of Rolling Friction

The issue of rolling friction (or rolling resistance) was briefly mentioned in Section
2.12.1. There are still questions over the way in which rolling resistance is defined (Zhu
and Yu, 2004), however, it has been stated by Kondic (1999) that the rolling friction
coefficient, r, is in the order of two magnitudes smaller than both the static and kinetic
friction and thus ignored. A rolling friction value of 1% of the sliding (kinetic) friction
has been used in DEM simulations by Zhu and Yu (2003). Other research has seen
similarly low rolling friction coefficients used, such as r = 0.002 (Yang et al., 2000),
56
r(chute) = 0.05 and r(particles) = 0.01 (Zhang et al., 2004). Andrews (2008) investigated the
simulation of angle of repose by creating a matrix of angle of repose values compared
back to an experimental datum by using multiple combinations of sliding and rolling
friction. The result was a contour map showing combinations of sliding and rolling
friction which would achieve the result found in reality.
2.14 DEM CODES AND LIMITATIONS
2.14.1 DEM Codes Developed by Research Groups

The original research based DEM code was BALL, developed by Cundall (1979) to
model two-dimensional disc assemblies. This later evolved into TRUBAL and since
then, numerous researchers have modified and adapted these codes to their own specific
needs (Dobry and Ng, 1992; Jensen et al., 1999; Sitharam, 1999; Mirghasemi et al.,
2002). Cundall has continued to develop the original BALL code and is now
commercially available as PFC2D and PFC3D, which are explained further in Section
2.13.2.1.
There are also a large number of research groups throughout the world who develop
their own code in either 2D or 3D (Walton, 1993; Wang and Liang, 1997; Asmar et al.,
2002; Pinson et al., 2004; Dutt et al., 2005; Fu et al., 2006; Pandey et al., 2006; van
Liedekerke et al., 2006; Song et al., 2006a; Song et al., 2006b).
Some DEM codes incorporate particle shape libraries (Zhao et al., 2006), which can
store any number of shape geometries, either based on clusters of spheres to produce
particle shapes representative of physical material or by methods such as 3D scans
which produce polyhedral shapes for use in simulations. All particle shapes formed by
spheres will all be based on the same shape primitives, thus have the same contact
detection methods, however, 3D polyhedrals would be based on other shape primitives,
thus having different contact detection methods (Zhao et al., 2006).
57
2.14.2 Commercially Available DEM Software Packages
2.14.2.1 PFC2D and PFC3D

Particle Flow Code (PFC) in 2D and 3D has evolved from the original Ball DEM code
of Cundall (Cundall and Strack, 1979). PFC was one of, if not, the first commercially
available DEM packages and is still used by many researchers (Tillack and Zhang,
1999; Masson and Martinez, 2000; McDowell and Harireche, 2002; Theuerkauf et al.,
2003; Krause and Katterfeld, 2004; McBride et al., 2004; Chandramohan and Powell,
2005; Li and Holt, 2005; Castro Vazquez et al., 2006; Shiu et al., 2006). It has the
capability of simulating just about any discrete particle application and has the added
bonus of the user being able to modify a wide range of internal parameters rather than
having to rely on pre-programmed code. PFC can also utilise multi-processor based
computers if the appropriate software add-on is purchased. Currently, simulations using
300,000 particles are possible with PFC3D (Grger and Katterfeld, 2006a).
2.14.2.2 Chute MavenTM

The Chute MavenTM software has evolved from the original work of Hustrulid (1997)
including the contact model (linear spring-dashpot) and neighbourhood lists for faster
contact detection. The user imports a 3D CAD DXF file with pre-defined layers to
distinguish between the components of the conveyor transfer, such as, belts, skirts,
chutes, and injection points. Within the software, particle properties such as coefficient
of friction between boundaries and particles, coefficient of restitution, particle density,
particle diameter and feed rate are entered. Once a simulation is complete, the results
can be displayed on a fully rotational 3D representation of the transfer and screen
captures or movies can be produced.
2.14.2.3 Chute Analyst

Chute Analyst is a DEM package which was developed by the Overland Conveyor
Company. The analysis process involves developing an accurate three dimensional
CAD representation of the conveyor transfer, determining material properties and
generating a particle description. The simulation is then performed and the results are
evaluated and if modifications are required, the procedure is repeated.
58
2.14.2.4 EDEM
EDEM has been developed by the Edinburgh based company DEM Solutions and as
with other DEM packages, model geometry is imported from a CAD package. EDEM
has a dedicated Particle Factory application for the design and importation of particle
shapes/clusters for use in the simulations. It also contains a database of common particle
properties and wall/boundary materials which can be added to as required (DEM-
Solutions, n.d.). EDEM has the ability to be coupled with FLUENT, a computational
fluid dynamics (CFD) package which allows the modelling of solid-fluid flow, allowing
such applications as fluidized beds and dense phase pneumatic conveying to be
simulated (Golz and Favier, 2006).
2.14.2.5 Newton
Information provided on the ACTek website (ACTek, n.d.) explains that the Newton
DEM software can include cohesive effects to particles and can be applied to many
applications such as silo discharge and transfer chute design. Newton has the ability to
model irregular shaped clusters of particles via the built in particle generation tool and
any particle size distribution can be defined. The coding of Newton can simulate in the
vicinity of half a million particles passing through a transfer chute at any one time in 3
to 4 days (Kruse, 2007).
2.14.3 Factors Influencing Simulation Time

There are several factors which closely dictate the time required to complete a DEM
simulation.
As computer power increases (e.g. processor speed, number of CPU cores and RAM),
faster DEM simulation outputs will result. This will allow simulations to more closely
replicate reality (Pournin et al., 2005). At its inception in the 1970s, DEM was able to
simulate a maximum of 1500 particles (Cundall and Strack, 1979). This number has
steadily increased as computing power has increased and now simulations in the tens to
hundreds of thousands are not uncommon (Xiang and McGlinchey, 2003; Morrison and
Cleary, 2004; Anderson and Britton, 2007).
The Rayleigh number is a measure of the time taken for a shear wave to propagate
through a solid particle, as shown in equation 2.146. The time step used in DEM
59
simulations is a function of the Rayleigh number and is generally based on the

conditions expected of the particles within the system. For densely packed particles,
such as those within a conveyor transfer, it is suggested to use a value of around 10% of
TR, whereas for more loosely packed particles, such as for conveyor trajectories, a value
of around 30% of TR is recommended (DEM Solutions, DEM-Solutions, 2008). The
effect of the variables within the equation for the Rayleigh number will be discussed in
Section 6.3.7.
1
2
rp s
TR = G 2.146
( 0.1631v + 0.8766 )
A wide range of contact models exist, some of which were detailed in Section 2.13, the
linear models generally being numerically simpler than the non-linear contact models.
The complexity of the Hertz-Mindlin contact model requires more computational time
to complete simulations.
When particle shape increases in complexity and the number of required particles
increases substantially, there is a need to utilise parallel processing to reduce simulation
time (Sawley and Cleary, 1999). The use of periodic boundaries can greatly decrease
processing time by simulating only a portion of an overall system. The general premise
is that as particles leave a simulation, they immediately reappear at the start of the
model with the same velocity vector they had at the end, thus creating a steady state. An
example of this is the screw conveyor DEM simulation of Katterfeld and Groger (2006)
where only two pitches of the screw conveyor were used.
Additionally, particle scaling can reduce the time required to complete a DEM
simulation. There is a two fold way in which this can help; by having a larger particle
diameter, the Rayleigh time step will decrease the time needed for each simulation step
and also by having larger particles, there will be less particles in the simulation overall,
also reducing the time required. Some researchers have documented investigations in
this area with positive results (Ji et al., 2006; Grima and Wypych, 2009).
60
2.15 COMPARISONS OF THE CONTACT FORCE MODELS

Luding (2006) states that using simple linear force laws to determine particle-particle
and particle-wall contacts probably will not reproduce reality to the extent required. As
a test of this, a shear cell was simulated in 3D without using complicated interactive
laws and the results reproduced to 80% of the experimental results.
Kruggel-Emden et al. (2006) review numerous contact force models such as the linear
spring-dashpot model and the partially latched spring model. Their intention was to
produce a new improved model by determining a more realistic collision time and
having the ability to model materials which deform plastically or viscoelastically. The
details of the new approach are not presented here but in summary, for the linear spring-
dashpot model new spring stiffness and damping coefficient equations have been
generated and for the partially latched spring dashpot model, new spring stiffness
equations have been generated for the loading and unloading cases. The results of this
new approach have seen the average and maximum error for the coefficient of
restitution decrease by 100% when compared to commonly used models.
Di Renzo and Di Maio (2004) make the point that even though the linear spring-dashpot
model is the most commonly used and that the Hertz-Mindlin model is also used
extensively, there has to date been no substantial comparison or assessment of the
models for the vast variety of applications that DEM is applied to. As a result of this
lack of comparison, Di Renzo and Di Maio undertook a review of three methods, linear
spring-dashpot, Hertz-Mindlin and Hertz-Mindlin (no-slip), along with simulations to
produce direct comparisons of the models. The simulations were conducted on a sphere
undergoing a frictional-elastic impact on a flat wall And the conclusions drawn by Di
Renzo and Di Maio (2004) from these comparisons were as follows;
- the linear spring-dashpot model had very good agreement for the macroscopic
comparison with no major benefit of using either of the Hertz-Mindlin models,
- for microscopic comparisons the Hertz-Mindlin model proved most accurate with
the no-slip version being worse than the linear spring-dashpot model under certain
conditions.
Dziugys and Peters (2001) review in great detail the inter-particle contact forces for
models such as the linear spring-dashpot and Hertz-Mindlin. There are also extensive
61
generalized discussions on normal and tangential contact forces for methods used by a
myriad of other researchers.
62
CHAPTER 3 THE CONVEYOR TRANSFER RESEARCH FACILITY
Chapter
3
THE CONVEYOR TRANSFER RESEARCH
FACILITY
63
3.1 INTRODUCTION
The design of the conveyor transfer research facility commenced in late 2005, involving
the key personnel associated with the ARC Linkage Project titled Quantification and
Modelling of Particle Flow Mechanisms in Conveyor Transfers. This design process
continued past the commencement of this PhD research and became one of the first
focal areas. Flexibility of design in a test facility such as this was vital and could be
achieved in two major ways, one being variable belt speed and the other, the ability to
interchange a variety of conveyor transfers with a minimum of effort.
3.2 KEY DESIGN CRITERIA
3.2.1 Conveyor Belt Speed

By incorporating variable frequency drives (VFD) into the conveyor transfer research
facility, a wide range of belt speeds were achievable. Discussions with key personnel
resulted in an agreement that a maximum belt speed of 7 m/s should be possible to
allow substantial high-speed conveying trials to be performed.
3.2.2 Conveyor Belt Width

It is not uncommon in industry for conveyor belts to be over 1 m in width, allowing
extremely high mass flow rates to be achieved. Incorporating belts of this width in the
proposed conveyor transfer research facility would have been impractical due to issues
such as the available space, power and volume of material required. At the other
extreme, small width belts would also be impractical due to issues such as the
uncertainty of scale-up effects and the additional expense of custom building conveyor
assemblies. Discussions lead to the selection of a 300 mm wide conveyor belt as a
suitable compromise.
3.2.3 Conveyor Belt Surface

Two main styles of belt surface exist, smooth and cleated. The general rule is that
smooth belts are used for conveyors with little to no inclination, whereas cleated belts
are used for conveyors having a substantial inclination angle. The cleats provide some
resistance to material sliding or rolling back down the conveyor. The decision was made
that for any belt conveyor installed in the conveyor transfer research facility with an
64
inclination of more than 10, cleated belts would be selected and for any belt conveyor
with an inclination less than 10, a smooth belt would be used.
3.2.4 Belt Conveyor Selection

In consultation with conveyor equipment companies, it was found that providing a
modular idler style conveyor was possible but there would be additional expenses
required to achieve the desired belt speed of 7 m/s. With strict budgetary constraints this
option was ruled out.
AerobeltTM conveyors showed more promise, with assurances from the supplier,
AerobeltTM Australia, that 7 m/s could be achieved by using VFDs connected directly to
the motor of each conveyor in the conveyor transfer research facility. Further to this,
was the ability to construct the conveyors to any length through the use of standard
trough lengths of 4.75 m and shorter made-to-order lengths as required. The decision
was made to choose AerobeltsTM and the 3D CAD design process was started.
3.3 FACILITY LAYOUT

Several design orientations were proposed with most consisting of some form of manual
return of material to the feed point. The chosen design relied on a recirculating
geometry using three AerobeltTM conveyors. The first and second conveyors were to be
aligned end to end along the same axis, with approximately a 1.5 m drop between them.
The second conveyor feed to feed through an inclined chute onto the third conveyor.
The third conveyor was then to be angled back toward the feed point of the first
conveyor to allow material to be returned to the feed bin. The 3D CAD design of the
conveyor transfer research facility underwent several evolutions with the key points
being;
conveyor 1 being 4 m in length between roller centres, inclined at 5,
conveyor 2 being 6.7 m in length between roller centres, inclined at 23,
conveyor 3 being 11.45 m in length between roller centres, inclined at 23,
conveyor 1 and conveyor 2 are parallel along the same central axis,
conveyor 3 is angled at 2.5 toward the feed bin to allow product recirculation,
the feed bin to have a maximum capacity of approximately 1 m3,
a HoganTM valve under the feed bin allowing for product feed control,
65
a curved chute under the HoganTM valve to promote product flow in the direction
of conveyor travel,
transfer chute 1 to allow for variable geometry conveyor transfers to be installed
for the purpose of testing, quantification and modelling,
transfer chute 2 consisting of a 60 slope to promote product flow, and
transfer chute 3 being angled at 10 to allow product to return to the feed bin.
The final model produced with Pro-Engineer is shown in Figure 3.1, excluding the belts
and the variable geometry conveyor transfer.
Figure 3.1 Final conveyor transfer research facility design
3.3.1 Conveyor Support Frames

2D CAD drawings supplied by AerobeltTM Australia showed the recommended
locations for the support frames. From this information, eleven frames were designed
and constructed from mild steel box section and flat plate.
3.3.2 Transfer Chutes

Three transfer chutes were required in the conveyor transfer research facility. Important
issues that had to be considered included; sufficient angles of the inclined surfaces such
that the velocity of flow was kept high enough to prevent stagnation zones and inlet and
outlet sizes were adequate to prevent blocking or restriction to flow.
66
3.3.2.1 Transfer Chute 1

The first design for transfer chute 1 was a hood and spoon arrangement, shown in
Figure 3.2, both constructed from a combination of 4.5 mm, 6 mm and 10 mm acrylic.
The green surfaces visible in Figure 3.2 indicate a Polystone Ultra liner, which has been
installed to promote flow of material as well as being more wear resistant than the base
acrylic construction. The hood has been designed to be fully adjustable longitudinally
and also perpendicularly up and down with respect to the inclination of the first
conveyor. The tail of the hood was designed to be attached to adjustable arms which
allow rotation of the hood through the rod support. The spoon was also designed to be
fully adjustable in the same way as the hood, although the spoon was designed to move
with respect to the inclination of the second conveyor. An acrylic enclosure was also
been designed, to confine the hood and spoon, to contain any product that may diverge
from the intended flow path through the transfer. The detailed engineering drawings of
the hood and spoon were developed by Grima (2007) and are located in Appendix B.
(a) (b)
Figure 3.2 Hood and spoon geometry of the first design - (a) CAD design,
(b) conveyor transfer research facility

The transfer chute between the second and third conveyor had many constraints put on
it due the geometry of the two conveyors, these included;
both conveyors having a 23 angle of inclination,
the third conveyor being offset by 2.5 to the second conveyor,
67
the outlet position on the second conveyor,

the use of the supplied conveyor inlet on the third conveyor,
the lower sloped surface of the chute having an inclination of no less than 60 to
the horizontal to minimise flow problems,
central feeding of product onto the third conveyor.
The final design of chute 2 is shown in Figure 3.1. One of the constraints listed above
could not be met, that being the sloped surface inclination angle being no less than 60.
Due to the height and width constraints there was no physical way to achieve the 60
requirement, the resultant angle being 58.5. For dry, free flowing products this should
still be acceptable, however, for materials which have some degree of moisture content,
monitoring of the chutes may be needed to avoid material build-up in valley angles,
with the potential of causing blockages.

The transfer chute, at the end of the third conveyor, needed to transition with the top of
the feed bin. Due to the third conveyor outlet not being positioned directly over the
central axis of the feed bin, an offset transfer resulted. The sides of the chute were
designed to be inclined at 10 to the vertical to allow the successful transition between
the chute and the feed bin lid.
3.3.3 Feed Bin

The feed bin also serves the purpose of receiving bin for the conveyor transfer research
facility, as shown in Figure 3.1. In conjunction with the design of chute 3, a square feed
bin was designed, ultimately having sides of 1200 mm, see Figure 3.3. A drawback of a
square bin is the resulting angles of the walls in the hopper section. Where each hopper
wall intersects, a shallow valley angle results, which is not recommended in bin design
as these valleys are prone to material hanging up. As a solution, fillets were inserted
over the valleys to produce steeper angled surfaces which would allow material to flow
more freely.
A HoganTM valve has been installed at the outlet of the feed bin to operate as both the
on/off control and the mass flow rate control. The HoganTM valve had a 290 mm square
68
opening and four rotating paddles which are used to control the flow of material. There
are also two vibrators attached to the side of the valve which can be activated in the case
of a non-free flowing material. A curved feed chute was installed to the underside of the
HoganTM valve to direct the flow of material in the direction of belt travel.
3.4 INSTALLATION OF THE CONVEYOR TRANSFER RESEARCH

FACILITY
The layout of the conveyor transfer research facility was to utilise the split levels of the
Bulk Materials Handling Laboratory floor, and as such, the 3D CAD design was used as
a template to provide the accurate positioning of the support frames. Conveyor 3 was
the first to be installed and due to its length, was delivered in two sections which
required separate assembly onto the frames before being joined, see Figure 3.4. The use
of chain blocks, slings, a forklift and numerous staff aided in the construction phase.
Once conveyor 3 had been successfully positioned, the remaining two conveyors were
installed using the same method.
The splicing of the three conveyor belts was performed by Rydell Industrial Belting
Company using a hot vulcanising process as can be seen in Figure 3.5.
To allow centralised operation of the conveyor transfer research facility, a control panel
was required. The operational requirements of the control panel were developed in
consultation with the contractor, IEE Services Pty Ltd, who was responsible for the
design and construction, see Figure 3.6a and Figure 3.6b.
It is important to note that the operation of the control panel set the longest conveyor to
started first, followed by the second conveyor and finally the feed conveyor. This was
done to ensure that any residual material within the system was cleared to minimise the
chance of material spillage or chute blockage. Consequently, the naming of the
conveyors has been reversed, so from this point onwards the naming convention given
is as follows;
- conveyor 1 is the 11.45 m conveyor
- conveyor 2 is the 6.7 m conveyor
- conveyor 3 is the 4 m conveyor.
69
Figure 3.3 Square feed bin assembly
Figure 3.4 Conveyor in position Figure 3.5 Splicing the cleated belt
Figure 3.6a Control panel front Figure 3.6b Control panel internals
70
Each conveyor has a start button which incorporates a safety mechanism. As each
conveyor is reliant on a fan supplying air to create an air cushion between the belt and
the trough, it is critical that the belt cannot move until a minimum safe pressure is
reached. A gas pressure switch installed on each conveyor controls the minimum
acceptable pressure on each conveyor, set to 0.5 kPa. When a conveyor is started, the
gas pressure switch must first reach the set minimum pressure before the motor can
operate. Each motor is controlled via an SEW inverter and the front of the control panel
has manually adjustable potentiometers to vary the belt speed of each conveyor.
There are three emergency stop switches, one for each conveyor, and regardless of
which is activated, the entire system instantaneously shuts down. The system cannot be
restarted until the emergency stop switch is reset.
Finally, the control panel also has on/off controls for the HoganTM valve vibrators and
additional 240V power outlets to supply power to the data acquisition unit, computer,
high-speed camera and lighting.
3.5 DRY COMMISSIONING

The dry commissioning did not involve the conveying of material due to the hood and
spoon transfer chute still being under construction. The dry commissioning was
performed to allow the belt tensions and belt tracking to be reviewed after the belt
splicing. Belt tensioning was performed by adjusting the lateral position of the tail
pulley of each conveyor. The conveyors were then powered up to check the belt
tracking. Adjustments were made by either repositioning the tail pulley, the snubber
idler on the underside of the conveyor near the tail pulley or the return idler on the
underside of the conveyor near the head pulley.
3.6 FINAL COMMISSIONING

The final commissioning was performed after delivery and installation of the transfer
hood and spoon. Again, the belt tensions and belt tracking were checked but while
material was conveyed. No adjustments were required, however, it was noticed that the
majority of the 18 transition idlers were not rotating and that they would have to be
checked. It was found that the end seals were causing a slight misalignment, but this
was easily remedied by using a press to reposition the shaft and seals.
71
3.7 FEED BIN AND CHUTE LINERS

Once experimental testing had commenced, it was found that elevated noise levels were
being produced when product impacted with the feed bin and second conveyor transfer,
see Figure 3.7. The noise generated in the feed bin could be alleviated by ensuring there
was always a sufficient head of material present, resulting in incoming material
impacting with other material rather than the bin walls. This option was not practical as
approximately twice the volume of material would be required. As a result, the decision
was made to install liners. Polystone Matrox liners were installed by Dotmar. The sheets
of Matrox were mechanically fastened to the sides of the feed bin and second conveyor
transfer and plastic welding was used to join two adjacent sheets of Matrox to each
other to avoid points of material ingress which could damage the liner.
3.8 DUST EXTRACTION

Some materials are naturally dusty and with occupational health and safety (OH&S)
being of high priority, cannot be allowed to propagate dust during transportation. The
initial design of the conveyor transfer research facility incorporated a number of acrylic
panels to enclose the experimental test zone; however, this was not entirely air tight.
The initial material tested was not inherently dusty and so was not an issue, but the
second material was extremely dusty and a solution was required. To remedy this
problem, the design of a dust extraction system was investigated by Roth (2008). The
extent of the dust extraction retrofitting included the following;
- incorporating a 7.5 Hp 3000 cfm Donaldson Shaker unit, see Figure 3.8,
- 250 mm diameter ducting reducing to 100 mm,
- a 100 mm diameter duct servicing the outlet of the second conveyor transfer along
with cardboard covers to reduce air entrainment, see Figure 3.9,
- 100 mm diameter duct branching and reducing to two 80 mm diameter ducts, both
servicing the experiment transfer zone, see Figure 3.10,
- a series of dust curtains at the transfer zone exit to minimise air entrainment,
- cardboard covers above the transfer hood to contain dust,
- a general purpose vacuum cleaner utilised at the feed point to capture dust being
generated in the initial curved chute, see Figure 3.11.
- cardboard covers around the feed point to minimise air entrainment, and
- sealing of the conveyor chute leading into the feed bin.
72
Figure 3.7 Impact of material inside feed bin
Figure 3.8 Donaldson shaker unit Figure 3.9 Extraction after second
transfer
Figure 3.10 Extraction at the transfer Figure 3.11 Extraction at feed point
zone
73
3.9 ADDITIONAL FRAMES

In December 2008, the University of Wollongong was successful in obtaining Federal
funding for a new building to be located on the site of the then current Building 6 which
housed the Bulk Materials Handling Laboratory. Demolition of the existing building
was scheduled to commence in early 2009. Relocation of test equipment, including the
conveyor transfer research facility, was to a property approximately 3 km away. This
new property had a single level floor which required the design of a new frame to
support the feed bin assembly and also required a set of steps to reach the HoganTM
valve to operate the feed control. Appendix B shows the new design of the feed bin
support stand, although, during the construction, the fabricators changed the design due
to time constraints. Figure 3.12 shows the conveyor transfer research facility in its new
location with the final design of the new feed bin support stand.
Figure 3.12 The conveyor transfer research facility installed at new location,
set for trajectory investigations
74
CHAPTER 4 CALIBRATION OF THE CONVEYOR TRANSFER RESEARCH FACILITY
Chapter
4
CALIBRATION OF THE CONVEYOR
TRANSFER RESEARCH FACILITY
75
4.1 INTRODUCTION
To ensure accurate measurement of key facility parameters, it is paramount that
accurate calibration of the test equipment be performed before, during and after testing.
The following sections detail the procedures and show examples of how this was
achieved.
4.2 DATA ACQUISITION

A Data Electronics DataTaker 80 data acquisition unit (DT80) has been used to record
the voltage signals generated by the loadcells during the experimental test program, see
Figure 4.1. The load cell cable was passed through an amplification box before being
connected to the DT80, where DeLogger software on a personal computer recorded the
input as millivolt signals. Microsoft Excel was then used for data manipulation where
the millivolt signals were converted into kilograms using a pre-determined calibration
factor, explained in Section 4.4.
Figure 4.1 DT80 data acquisition unit
4.3 CALIBRATION OF THE CONVEYOR BELT SPEED

Each belt conveyor required calibration of its belt speed to allow accurate and
repeatable tests. It was found that the potentiometer scale controlling the speed of the
three conveyors was not accurate enough and so the digital output from the SEW
inverters was used instead, in conjunction with a laser tachometer reading of the head
pulley shaft RPM for each conveyor. The laser tachometer required a thin strip of
reflective tape to be attached to the rotating shaft and after several seconds of aiming the
tachometer at the rotating strip, the shaft RPM was displayed. The belt speed was
76
determined by multiplying the RPM of the head pulley shaft by the outer diameter of
the belt surface. The resulting data was then manipulated to produce a table of belt
speed versus inverter digital output for each conveyor, shown in Table 4.1.
Table 4.1 Belt speed calibration chart
CONVEYOR 1 CONVEYOR 2 CONVEYOR 3

BELT SPEED DISPLAY BELT SPEED DISPLAY BELT SPEED DISPLAY
m/s RPM m/s RPM m/s RPM
0.00 0 0.00 0 0.00 0
0.50 209 0.50 222 0.50 207
1.00 418 1.00 442 1.00 413
1.50 625 1.50 659 1.50 619
2.00 832 2.00 874 2.00 824
2.50 1037 2.50 1086 2.50 1029
3.00 1242 3.00 1296 3.00 1233
3.50 1446 3.50 1504 3.50 1436
4.00 1648 4.00 1710 4.00 1639
4.50 1850 4.50 1913 4.50 1841
5.00 2051 5.00 2113 5.00 2043
5.50 2251 5.50 2312 5.50 2244
6.00 2449 6.00 2508 6.00 2444
6.50 2647 6.50 2701 6.50 2644
7.00 2844 7.00 2893 7.00 2843
4.3.1 Additional Conveyor Belt Speed Calibration

Due to the relocation and subsequent recommissioning of the conveyor transfer research
facility in early 2009, calibration checks were undertaken on the feed conveyor to
ensure the original calibration was still accurate. The RPM was set from the calibration
found in Section 4.3 and the laser tachometer was again used to determine the rotational
speed of the head pulley shaft. The belt speed was then calculated and checked against
the original value, see Table 4.2. These results showed that the original calibration was
still accurate.
77
Table 4.2 Belt speed calibration check
CONVEYOR 3
ORIGINAL CHECKED
BELT SPEED BELT SPEED
m/s m/s
1.00 0.990
2.00 1.999
3.00 3.044
4.00 4.030
5.00 5.031
6.00 6.044
7.00 6.994
4.4 CALIBRATION OF THE FEED BIN LOAD CELLS

Each of the four sides of the feed bin had a 500 kg shear beam load cell attached. The
four load cells were wired together allowing measurement of a maximum of 2000 kg.
The loadcells were then connected to an amplifying box before being connected to the
data acquisition unit. To calibrate the load cells, the data acquisition system was
programmed to record the incoming signal. The data acquisition unit was started while
no load was acting on the load cells to obtain a zero reading. Masses were progressively
added to the load cells, allowing for a period of steady state to be achieved before
adding the next mass. Visually, a stepped graph of millivolts vs. time results resulted, as
shown in Figure 4.2. A scaling factor was the determined to convert the millivolt scale
into kilograms. An example of a load cell calibration is shown in Figure 4.3.
1.50 250
1.25
200
Millivolts (mV)
1.00
Mass (kg)
150
0.75
100
0.50
50
0.25
0.00 0
0 100 200 300 400 500 0 100 200 300 400 500
Time (sec) Time (sec)
Figure 4.2 Millivolt reading vs. time Figure 4.3 Calibrated mass vs. time
78
4.5 CALIBRATION OF THE HOGANTM VALVE

The discharge of material from the feed bin was through a HoganTM valve. This valve
can be operated through 90 of travel, 0 being closed and 90 being fully open. An
angular scale was produced for 10 incremental positions with 5 increments also
achievable, see Figure 4.4.
Before calibration, the belt speed was set such that it would remove material from the
feed point to avoid build-up of material directly below the HoganTM valve. The data
acquisition unit was then connected to the feed bin load cells and by successively
increasing the opening position of the HoganTM valve, the mass flow rate was
determined. The calibration graph for each test material is shown in Figure 4.5.
There were several observations made with reference to the trends presented in Figure
4.5;
- for polyethylene pellets it was found that at 45 the HoganTM valve was not able
to deliver any more material from the feed bin. This seemed to indicate that
flood feeding had been achieved,
- for both corn and iron ore, the flood feeding seen with polyethylene pellets did
not occur, instead there was a steady increase in mass flow rate,
- calibration for the corn and iron ore did not continue for the full range of the
HoganTM valve as the material feeding onto the conveyor was reaching the point
of over-filling the conveyor belt. This would have resulted in spillage of the
material,
- an alternative method which could have been employed for all test materials was
to remove the feed bin from the conveyor transfer research facility and discharge
material from the feed bin into another hopper where the possibility of over-
filling would not have been an issue. The decision was made to calibrate with
the method described above so that the material would be easily recirculated for
the following test.
79
Figure 4.4 Angular scale for the HoganTM valve
Figure 4.5 HoganTM valve calibration graph for test materials
80
CHAPTER 5 PARTICLE AND BULK CHARACTERISTICS
Chapter
5
PARTICLE AND BULK CHARACTERISTICS
81
5.1 PARTICLE AND BULK CHARACTERISATION AND MEASUREMENT

To quantify the physical properties of the materials to be used for the testing program
and also for the discrete element modelling, a range of tests were undertaken. The
following sections detail those tests with the results provided in Section 5.2 for each test
material. All testing was performed by the author or staff in the Bulk Materials
Handling Laboratory at the University of Wollongong.
5.1.1 Loose-Poured Bulk Density

Loose-poured bulk density is the mass per unit volume that has been measured by
carefully pouring material into a graduated measuring cylinder without causing any
compaction or consolidation, see equation 5.1. Testing followed the standardised
procedure described by Halford (2009).
Masssolids
b = 5.1
Volumetotal
5.1.2 Mean Particle Density

The particle density of a material is its mass divided by its true volume. The mean
particle density of a bulk solid sample is determined by using a manually controlled
stereopycnometer, see Figure 5.1, following the standardised procedure described by
Halford (2009).
Figure 5.1 Stereopycnometer
82
5.1.3 Equivalent Volume Diameter

The equivalent volume diameter is the diameter, Dsv, of a sphere having the same
volume as the particle being measured, Vp, see equation 5.2. Sufficient particle volumes
should be measured to achieve a representative value.
1
6V 3
Dsv = p 5.2

In the case of the three test materials they were relatively mono-sized. The shape of
individual polyethylene pellets was approximately cylindrical so the length and
diameter of 50 particles were measured directly and the shape of corn was
approximately a trapezoidal prism so the maximum and minimum widths, height and
depth were measured for 50 particles. Subsequently the average volume for one particle
was determined for each material. The shape of iron ore particles was more irregular
and dimensions hard to determine in the same way as the other two test materials so 50
particles were submerged in a known quantity of water, then the change in volume
recorded. This change in volume was equivalent to the total volume of the 50 iron ore
particles. The average volume for one iron ore particle could then be determined. Once
the average volumes of the three test materials were determined, equation 5.2 was
employed to find the average equivalent volume diameter of each test materials.
5.1.4 Particle Sphericity

As previously described, particle sphericity can be determined using equation 2.97. For
a perfect sphere, the sphericity is 1, for other shapes the sphericity is a value less than
one. The mean volume of an individual particle, Vp, is first determined along with the
surface area of a sphere, As, having the same volume as the particle.
The above method is sufficient if considering regular shaped particles, however, if

particles are irregular in shape, the particle volume and/or surface area can prove to be
difficult to determine. The method used by Wadell (1935) for determining the sphericity
of quartz particles has been applied in the case of irregular shaped particles. This
method consisted of firstly determining the area of a projected surface of a particle and
the resulting diameter of a circle, dc, with the same area. This was achieved using the
83
Image Pro Plus analysis software. The projected area was chosen based on the largest
projected area of each particle analysed. Secondly, the diameter of the minimum
circumscribing circle containing the particle was found, Dc, see Figure 5.2 for an
example. The sphericity could then be determined using equation 5.3.
Figure 5.2 Sample particle and circumscribing circle
dc
= 5.3
Dc
5.1.5 Particle Size Distribution

The products selected for the experimental work were of a granular nature and sieves
were used to determine the particle size distribution (PSD) of each material. A selection
of sieves was selected spanning the size range of the test materials, see Figure 5.3(a).
Material was then placed in the top sieve and all were shaken for a standard time in the
sieve shaker, Figure 5.3b. The sieves were then removed and each one weighed to
determine the retained sample mass for each size fraction. The percentage of material
retained on each sieve was then used to represent the particle size distribution.
84
(a) (b)
Figure 5.3 (a) Sample of sieves, (b) sieves in mechanical sieve shaker
5.1.6 Coefficient of Restitution

The coefficient of restitution is the ratio of the velocity of a particle after an impact to
the velocity of the particle before impact, as in equation 5.4.
V2
COR = 5.4
V1
The experimental procedure used to determine the coefficient of restitution involved

dropping particles onto different flat surfaces; acrylic, Polystone Ultra, polyethylene and
smooth AerobeltTM. High-speed video was used to capture particles falling and
impacting on each of the surfaces. A potential issue with the measurement of the
coefficient of restitution results from the irregular shape of all materials other than
spheres. Spherical particles would rebound vertically, providing an accurate
measurement of V1 and V2, but irregular shaped particles would potentially rebound
with a varied angular component. This would have a dramatic effect on the determined
coefficient of restitution. To overcome this, only particles which rebounded close to
vertical were used in the analysis, resulting in many disregarded tests. The subsequent
video footage was then analysed using Image Pro Plus to determine the velocity before
and after impact. Using equation 5.4, the coefficient of restitution was then determined
for each surface material.
85
5.1.7 Wall Friction Angle and the Coefficient of Wall Friction

The wall friction angle is measured from a wall yield locus, which is produced using the
standardised procedure for coal (SAA, 1991) to determine the effect of wall materials
on the flow of products flowing from storage bins. The method can also be used to
determine the wall friction angle of surfaces such as those on chutes and conveyor
transfers. The wall yield locus plots the relationship between wall shear stress, w, and
wall normal stress w. At any point along this locus the wall friction angle can be
determined by projecting back to the origin and using the relationship shown in
equation 5.5 and the coefficient of wall friction by equation 5.6.
w
w = tan 1 5.5
w
w
w = 5.6
w
The wall friction angle of the test materials was determined for wall samples of
AerobeltTM, acrylic and Polystone by staff in the Bulk Materials Handling Laboratory
using a Jenike wall friction tester, see Figure 5.4. Samples of most of the wall materials
used in the conveyor transfer rig were readily available (vis. acrylic, Polystone Ultra
and smooth AerobeltTM). However, a special wall sample was prepared for the
polyethylene pellets, as shown in Figure 5.5, to allow the determination of a pseudo
internal friction angle to be determined separately to the regular method, described in
Section 5.1.8.
Figure 5.5 shows the method used to slowly heat a small quantity of polyethylene
pellets to melting within an enclosed box. The molten polyethylene was then allowed to
cool, resulting in the formation of a 150 mm square sheet which was then used to
determine the wall friction angle.
86
Wall sample
Cell
Hanger
Figure 5.4 Jenike shear tester
(a) (b)
(c)
Figure 5.5 (a) Melting of polyethylene pellets to create a test sheet, (b) partially
melted polyethylene pellets and (c) the final polyethylene pellet wall sample
87
5.1.8 Internal Friction Angle

The internal friction angle is a measure of the resistance to shearing of a material when
exposed to a normal force when a shear force is applied. An instantaneous yield locus
test is performed using the standardised procedure for coal (SAA, 1991), which can be
adapted for other materials. To determine the internal friction angle, the Jenike shear
tester is once again used, but in a different configuration to that used for determining the
wall friction angle. The shear cell comprises two parts, the shear base and the shear ring.
The shear base is positioned on the bearing plate of the Jenike shear tester and the shear
ring sits on top of the shear base. The shear base and ring are then filled with material
and covered with a shear lid, see Figure 5.6. A normal force is then applied to the
material through the shear lid by way of a mass hanger, which consolidates the material.
The shear cell is then sheared across the top of the shear base. The end result is a set of
stress data, where shear stress is plotted against normal stress to produce yield loci. The
internal friction angle can then be obtained from these loci. The internal friction angle
was used as an approximation of the particle-particle friction in the Chute MavenTM
DEM software.
FN
shear lid
Fs
shear ring
base
bearing plate
Figure 5.6 Jenike shear tester configuration for IYL test
5.1.9 Static and Kinetic Wall Friction

The coefficient of static friction, s, applies up to the point where slippage first occurs
between the material and wall, while the coefficient of kinetic friction, k, applies once
motion commences. It is generally accepted that k < s (Meriam and Kraige, 1987).
The coefficients of static and kinetic wall friction are required in the determination of
the discharge angles and trajectories for the Booth (1934) and Korzen (1989) analytical
88
methods. In the laboratory, this test was performed on the apparatus shown in Figure
5.7. The experimental test materials were prepared by gluing them to a piece of wood in
a closely packed single layer, examples are shown in Figure 5.8.
These samples were created to represent the layer of material closest to the conveyor
belt, allowing the determination of the coefficient of static wall friction. By slowly
elevating the inclined plane, the angle at which the samples began to slide on the
AerobeltTM was recorded and using equation 5.7, the coefficient of static wall friction
was determined.
Figure 5.7 Inclination tester
(a) (b) (c)
Figure 5.8 Preparation of test material for the friction test with
(a) polyethylene pellets, (b) iron ore and (c) corn
The coefficient of kinetic wall friction was determined directly from the coefficient of
static wall friction using the relation shown in equation 5.8, where Ts and Tk are the
89
static and kinetic frictional resistance on the belt surface (Korzen, 1989). For the
estimates provided for the test materials, the midway value of 0.8 has been chosen.
s = tan 5.7
Tk
= 0.7 to 0.9 5.8
Ts
5.1.10 Particle-Particle Friction

The angle (or coefficient) of particle-particle friction is a measure of the angle at which
particles will begin to slide or roll over themselves in a non-consolidated way,
attempting to replicate the behaviour of materials on a conveyor belt or in a transfer
chute. The method used to determine the angle of particle-particle friction utilised the
same equipment as for the static and kinetic wall friction. The wooden blocks were
inverted so the test materials faced upward, onto which sample of the corresponding
material was placed while the inclination tester was in the horizontal position. The
surface was then gradually inclined until particles began to fall off the test block. This
angle was then recorded and the procedure repeated several times to obtain an average
value. From this average angle, the coefficient of particle-particle friction could be
determined from an equation of the same form as equation 5.7.
5.1.11 Angle of Repose

The angle of repose is the maximum angle of the slope formed when a quantity of
material is poured into a heap, see Figure 5.9a. To determine the angle of repose, a cone
is placed on a flat surface and filled with a representative sample of a material. The cone
is then slowly raised, allowing the material to flow out and gradually form a heap. The
maximum angle of the side of the heap is then measured. In reality the apex of the heap
will be somewhat flattened or curved due to dynamic effects as the falling particle
stream interacts.
5.1.12 Surcharge Angle

The surcharge angle is a dynamic angle of repose and is specific to belt conveyors. As
material is fed onto a conveyor belt, it undergoes dynamic interactions as a result the
method of the method of feeding onto the conveyor (i.e. feed chute, vibrating screens),
90
interaction with skirts, idler spacing or the belt speed. As a result, the surcharge angle
will be less than the angle of repose, See Figure 5.9. To measure the surcharge angle, a
straight line from the outer most point of the material bed to the apex of the material
profile is drawn and the angle of this line to the horizontal is calculated (C.E.M.A.,
2005). In reality the apex of the surcharge heap will be somewhat flattened or curved
due to interactions such as those mentioned above.
Figure 5.9a Angle of repose Figure 5.9b Surcharge angle
5.1.13 Terminal Velocity of Particles

Numerous methods of determining the freefall or terminal velocity of a particle were
presented in Section 2.7. Some of these methods allow for the inclusion of particle
sphericity to obtain a more accurate terminal velocity for a non-spherical particle
whereas others only determine the terminal velocity of spherical particles. In the latter
case, the method employed by Wypych (1993), using a sphericity factor has been
applied to estimate the terminal velocity of the particle based on its sphericity. The
method used to determine the freefall velocity of the test materials, along with the
results, is provided in detail in Chapter 10.
5.1.14 Poissons Ratio and Shear Modulus

Poissons ratio, , is a measure of the ratio of lateral to axial strain, provided they both
fall in the range of Hookes law (Shigley, 1986), see equation 5.9. Poissons ratio is also
required in the setup of the EDEM discrete element modelling.
lateral strain
= 5.9
axial strain
91
The shear modulus, G, (or modulus of rigidity), is an indication of the stiffness of a

material and is required in the setup of the discrete element modelling using EDEM.
The shear modulus is represented by equation 5.10, where E is Youngs modulus (or
modulus of elasticity) and is Poissons ratio. By knowing any two parameters the third
can be found.
E
G= 5.10
2 (1 + )
No test equipment was available to perform tests to determine these two parameters. As
a result, estimations have been made based on other research and this is discussed
further in Section 5.2.4.
5.2 TEST MATERIALS

Throughout the experimental test program numerous products were used. A series of
bench-scale tests were required to determine the particle and bulk properties as
described in Section 5.1. The test data was then used in the analytical modelling and
discrete element modelling simulations as well as being used to create the beginnings of
the particle database.
5.2.1 Polyethylene Pellets

Polyethylene pellets were selected as the first test material due to its durability and
consistency. This material is highly free flowing, non-cohesive, non-adhesive and has a
sphero-cylindrical shape, as seen in Figure 5.10a. The measured particle and bulk
characteristics are listed in Table 5.1.
92
Table 5.1 Particle and bulk characteristics of polyethylene pellets
Loose-poured bulk density 514.5 kg/m3

Particle density (Pycnometer) 919.2 kg/m3
Equivalent volume diameter (based on 50 particles) 4.65 mm
Particle sphericity 0.873
Particle size distribution (2.36 3.35 mm) 2.90 %
Particle size (minimum) 2.36 mm
Particle size (maximum) 4.75 mm
Angle / coefficient of wall friction (smooth AerobeltTM) 25.7 / 0.481
Angle / coefficient of wall friction (4.5 mm acrylic) 21.5 / 0.394
Angle / coefficient of wall friction (6 mm acrylic) 19.1 / 0.346
Angle / coefficient of wall friction (Polystone with grain) 15.5 / 0.277
Angle / coefficient of wall friction (Polystone across grain) 16 / 0.287
Angle / coefficient of wall friction (Polyethylene sheet) 12.5 / 0.222
Angle / coefficient of static wall friction (smooth AerobeltTM) 21.5 / 0.394
Angle / coefficient of kinetic wall friction (smooth AerobeltTM) 17.5 / 0.315 *
Angle / coefficient of internal friction 44 / 0.966
Angle / coefficient of particle-particle friction 16.5 / 0.296
Coefficient of restitution (smooth AerobeltTM) 0.542
Coefficient of restitution (4.5 mm acrylic) 0.626
Coefficient of restitution (6 mm acrylic) 0.668
Coefficient of restitution (Polystone liner) 0.657
Coefficient of restitution
0.650
(Average of 4.5mm, 6mm acrylic and Polystone)
* based on equation 5.8
5.2.2 Yandicoogina Iron Ore

Samples of iron ore were sourced from Rio Tinto owned Hammersley Iron, which
operates the Yandicoogina Mine in Western Australias Pilbara region. This iron ore
was requested in two size ranges, 3 6.3 mm and 6.3 10 mm to investigate possible
differences of conveyor transfer chute flow of different size fractions. The particle
characteristics for the smaller size range were determined first and it was found that the
size range varied slightly to that originally requested, i.e. 2.36 6.3 mm, see Figure
5.10c. Due to dust generation issues, which are explained in detail in Section 9.3.1, only
the 2.36 6.3 mm size fraction was tested. The particle characteristics for this size
fraction are presented in Table 5.2.
93
Table 5.2 Particle and bulk characteristics of iron ore
Loose-poured bulk density 1607 kg/m3

Particle density (Pycnometer) 4039.3 kg/m3
Angle / coefficient of wall friction (Polystone with grain) 15.6 / 0.279
Angle / coefficient of wall friction (Polystone across grain) 16.1 / 0.289
Angle / coefficient of static wall friction (smooth AerobeltTM) 36.5 / 0.740
0.442
(Average of 6mm acrylic and Polystone)
5.2.3 Corn
This material was selected only for use in the trajectory comparisons due to its granular
free-flowing nature along with its flatter shape. The shape of the corn particles made it a
perfect choice to investigate during simulation in the discrete element modelling, see
Figure 5.10b. The measured particle and bulk characteristics are provided in Table 5.3.
(a) (b) (c)

Figure 5.10 (a) Polyethylene pellets, (b) corn, and
(c) iron ore, size range 2.36 6.3 mm
94
Table 5.3 Particle and bulk characteristics of corn
Loose-poured bulk density 691.1 kg/m3

Particle density (Pycnometer) 1362 kg/m3
Angle / coefficient of wall friction (Polystone) 17.4 / 0.313
Angle / coefficient of static wall friction (smooth AerobeltTM) 29 / 0.554
0.618
(Average of 6mm acrylic and Polystone)
5.2.4 Determining the Shear Modulus and Poissons Ratio

The values for the shear modulus and Poissons ratio for the materials being used in this
research were unable to be determined due to a lack of suitable test equipment.
Approximate shear modulus values for the test materials as well as the relevant
conveyor components were sourced from various locations, as referenced in Table 5.4.
These values were then used in the setup of DEM discrete element modelling
simulations. To evaluate the validity of the values quoted, a sensitivity analysis was
completed to determine the effect of varying shear modulus on the DEM simulation
outputs, as presented in Section 6.3.9.1.
95
Table 5.4 Shear modulus values of materials and products
Modulus of Shear
elasticity modulus
Material MPa MPa Source
AerobeltTM (soft PVC) 80 35 6 (Matbase, n.d.)
Acrylic (Perspex) - 1400 (3DCAM, 2009)
Polystone Ultra 1 800 333 7 (Rochling, n.d.)
Polyethylene Pellets - 117 (Archer et al., 1972)
Iron Ore 7551 2 2697 7 (Garg, 1973)
5394 2 1926 7 (Garg, 1973)
Corn 20 480 7 171 7 (Singh et al., 1991)
8.92 3 3.2 7 (Niedzika and Szymanek, 2005)
1
ultra high molecular weight polyethylene (UHMWPE)
2
unfrozen permafrost containing iron ore
3
mean results from a cutting test on four corn variants
4
mean results from a penetration test on four corn variants
5
mean results from a pressing test on four corn variants
6
it is believed that the modulus of elasticity and shear modulus provided by Matbase (n.d.)
are back to front, as when using equation 5.10, the shear modulus should be the lower value
7
calculated using equation 5.10
Likewise, values for Poissons ratio were also sourced from various locations, as
referenced in Table 5.5. An additional sensitivity analysis was completed to determine
the influence of variations in Poissons ratio also, see Section 6.3.9.1. It should be noted
that initial searches failed to locate a representative Poissons ratio for the AerobeltTM
(soft PVC). As this value was crucial to the commencement of the EDEM simulations,
an estimate of 0.45 was chosen, as shown in Table 5.5. Later searches proved more
fruitful, with values being discovered for flexible PVC within the CES EduPack
software (2009) for varying Shore A hardness values, although the quoted values have
been marked as being estimates within the CES EduPack also. As it turns out, these
values varied only marginally to the initial Poissons ratio estimate used in the DEM
simulations. These are also presented in Table 5.5.
96
Table 5.5 Poissons ratio values of materials and products
Poissons Source
ratio
Material
AerobeltTM (soft PVC) 0.45 Estimate
0.48 0.49 2 (CES EduPack, 2009)
0.47 0.49 3 (CES EduPack, 2009)
0.47 0.48 4 (CES EduPack, 2009)
Acrylic (Perspex) 0.37 0.39 5 (Harris and Sabnis, 1999)
Polystone Ultra 1 0.35 (Weihan and Fengnian, 1992)
0.4 (Jin et al., 1999)
0.43 (Bakshi et al., 2007)
0.36 (Karuppiah et al., 2008)
Polyethylene Pellets 0.45 (Archer et al., 1972)
Iron Ore 0.36 (Grgic et al., 2008)
0.41 6 (Garg, 1973)
0.44 6 (Garg, 1973)
Corn 0 and 0.5 7 (Ekstrom et al., 1966)
0.447 (Mohsenin, 1968)
0.4 (Shelef and Mohsenin, 1969)
0.32 0.4 8 (Litchfield and Okos, 1988)
0.4 (Jia et al., 2000)
1
ultra high molecular weight polyethylene (UHMWPE)
2
Flexible PVC with Shore A hardness = 60
3
4
5
at room temperature
6
unfrozen permafrost containing iron ore
7
Ekstrom et al. (1966) could not determine , so investigated the two extremes
8
at 10% moisture content
97
CHAPTER 6 DISCRETE ELEMENT MODELLING SOFTWARE
Chapter
6
DISCRETE ELEMENT MODELLING
SOFTWARE
98
6.1 INTRODUCTION
Two discrete element modelling (DEM) software packages were used in this research,
Chute MavenTM and EDEM, to simulate particle flow through the conveyor transfer
research facility.
At the beginning of this research, the Chute MavenTM DEM simulation software was
purchased as it was specifically developed for modelling belt conveyor systems. It will
be explained in later chapters that limitations were observed using Chute MavenTM with
respect to the required high mass flow rates not capable of be achieved. As a
consequence, no direct comparisons could be made between the high mass flow rates
tested experimentally and simulations by the Chute MavenTM software. The low mass
flow rate simulations were still of value and used as part of comparisons in subsequent
chapters.
A remedy to this limitation was to source another DEM simulation package having the
ability to achieve the desired mass flow rates. The EDEM software package was
sourced which proved more than adequate in generating the high mass flow rates
required to supplement the simulations already performed with Chute MavenTM.
6.2 CHUTE MAVENTM

The programming code behind discrete element modelling can vary greatly in terms of
interaction laws and time integration. For the initial DEM simulation research, the
commercial software Chute MavenTM has been utilised, which uses the linear spring-
dashpot contact model. As with all DEM, the contact model comprises normal and shear
force components. Hustrulid (1997) defined the normal contact stiffness, kn, as that
shown by equation 6.1, where mmax is the largest mass in the simulation, max is the
maximum relative velocity between the interacting particles and boundaries and max is
the maximum overlap (penetration) between any two particles (the default maximum
penetration in Chute MavenTM is set at 5%). The shear stiffness, ks, is defined as being
equal to the normal stiffness.
2
vmax
kn = mmax 6.1
max 2
99
The viscous damping coefficient, cd, is defined by equation 6.2, where two particles
come into contact, denoted by the subscripts 1 and 2. The c term is the partial damping
coefficient of each particle in the collision, defined by equation 6.3, where is the
coefficient of restitution, which was previously determined experimentally and shown
in Tables 5.1, 5.2 and 5.3 for the three test materials. The explicit central difference
scheme is used for the time integration of the DEM code (Hustrulid, 1997).
c '1 c '2
cd = 6.2
c '12 + c '2 2
1 kn m
c ' = 2 ln 6.3

2
1
2 + ln

Five main steps are required in producing a DEM simulation, which are;
- develop a 3D conveyor transfer geometry in a CAD application,
- import the 3D CAD model into Chute MavenTM,
- input the simulation parameters and save the associated setup files,
- run the simulation,
- export the data to MS Excel for further analysis.
The following sections will not repeat that which is already available in the Chute
MavenTM instruction manual (Hustrulid, 2005) but will instead provide concise
overviews of the requirements of Chute MavenTM to produce successful simulations.
However, in the process of learning to use Chute MavenTM and understanding the
numerous parameters required for setting up a simulation, several issues arose which
will be discussed further where appropriate.
6.2.1 Three Dimensional CAD Models

Before starting any DEM simulation, an accurate representation of the conveyor transfer
must be produced. This can be achieved with any three dimensional CAD package
capable of producing a DXF format model. Chute MavenTM requires specific attributes
assigned to this DXF file, the most important is that each component of the three
dimensional model has to be located on a separate layer with a specific naming
100
convention allowing Chute MavenTM to interpret the file correctly. As a minimum, the
following layers are required to simulate a conveyor transfer;
- belt conveyor 1,
- head pulley 1,
- chute 1,
- belt conveyor 2,
- skirt 1 (used to enclose injection zone),
- injection box 1.
Figure 6.1 shows an example of a three dimensional CAD model ready for use in Chute
MavenTM. Other layers can be added if required, being mindful of the required naming
convention. Other necessary attributes on certain layers includes: a velocity vector for
layer belt conveyor 1 which defines the direction of belt travel; a velocity vector is
also required for layer belt conveyor 2; and a diagonal vector is required to be drawn
on the layer injection box 1 which defines the three dimensional space where particles
are generated. Generally, the injection layer is positioned inside the layer skirt 1 which
directs the material onto the first belt conveyor; however, other injection points can be
included in a model. In the case of a rock box, it may prove more computationally
efficient to include one or more injection zones within the dead region of the rock box
to speed up the time required to reach steady state conditions within the simulation.
Figure 6.1 Three dimensional CAD model of a conveyor transfer
101
The initial aim of the 3D CAD design process was to reproduce the conveyor transfer
research facility exactly so as to produce the most accurate simulations possible. This,
as it turns out, was not going to be possible due to a limitation in the way the geometry
of a conveyor belt is defined. On the conveyor transfer research facility, the transition
zone leading to the head pulley has an inclination angle of 2.5 more than the 5 angle of
the main troughed section of the conveyor. As such the 3D CAD model was designed to
represent this. When the model was imported into Chute MavenTM, the software would
crash and after some investigation it was discovered that the extra angle at the transition
was the cause. This meant that the entire belt conveyor had to be designed at one angle.
If an inclination angle of 5 was used, the discharge of material would not be
representative of the experimental geometry so the decision was made to model the
entire first conveyor at 7.5 to mimic the angle of the transition, as this is the angle at
which the belt became tangent with the head pulley.
6.2.2 Model Parameters

There are a number of model parameters required to be entered on the Chute MavenTM
Model Parameters window, as shown in Figure 6.2. The simulation time and output
time interval are self explanatory. The coefficient of friction between particles proved
hard to quantify and is explained further in Section 6.2.2.1. and the coefficient of
restitution is explained in Section 5.1.6. The maximum penetration (introduced earlier
in section 6.2) relates to the contact between particles and the boundaries. It defines the
physical overlap of particles allowable in the computations and is one factor used to
determine the spring stiffness in the linear spring-dashpot contact force calculations.
The critical time step factor is recommended to be kept set at 5%.
During the simulation process, at each one second of simulation time, a restart file is
saved. This is in the unlikely event of a system crash or a power failure. The simulation
can then be restarted from that point to completion. Restart files can also be used as a
means to shorten the length of time required to perform a simulation. If only the
geometry of the conveyor transfer differs and the belt conveyor and injection points
remain constant, previous simulations of material feeding onto the conveyor belt up to
the point of discharge from the head pulley can be simulated and from then on can be
used as a restart file to simulate through any number of transfer chute geometries for the
same belt speed.
102
Figure 6.2 Chute MavenTM model parameters
6.2.2.1 Coefficient of Friction Between Particles

The definition given in the Chute MavenTM instruction manual states;
A rough approximation of an appropriate value can be taken

from the tan of the surcharge angle. (Hustrulid, 2005)
The fact that a clear cut definition of the coefficient of friction between particles was
not given raised some doubt over the validity of using this estimate. As well as this, in
reality, the surcharge angle formed by the material changes slightly at different belt
speeds as a result of the interaction between the material, the belt and the side skirts.
This could, however, be overlooked if the term rough approximation is applied. The
initial instinct of the author was to obtain the results of an instantaneous yield loci (IYL)
test to determine the internal friction angle of the material, explained in Section 5.1.8.
For polyethylene pellets, this angle was found to be 44, which produced a coefficient of
friction of 0.966, see Table 5.1. This value was deemed too high and more appropriate
for use in bin and hopper design where high compressive stresses are commonplace.
Further thought led to investigating the possibility of measuring the friction generated
by sliding one particle over another. This in itself would be quite difficult, especially if
trying to utilise the Jenike shear cell test apparatus. However, using a sheet of
polyethylene, a wall yield loci (WYL) test was performed to determine the wall friction
103
angle, as explained in Section 6.1.7. The result was a wall friction angle of 12.5, giving
a coefficient of friction of 0.222. This value was judged to be more realistic and is on
par with the default value of 0.3 (Hustrulid, 2005). Further discussion on this and a
sensitivity analysis are provided in Section 6.2.7.
6.2.3 Simulation Data

The specific information required to be assigned to the numerous layers within the 3D
CAD model are entered on the simulation data window, an example of which is shown
in Figure 6.3. The time on and time off entries for all elements refers to the time in
seconds when the item appears and disappears during the simulation. The friction for
the belts refers to the coefficient of wall friction between the material being tested and
the AerobeltTM, as explained in Section 5.1.7. The belt speed is also entered for each
belt in the simulation. Generally there is only one injection point in a simulation
although more than one can be included, as explained previously. For each injection box
the mass flow rate, particle density and minimum and maximum radius are required.
Figure 6.3 Chute MavenTM simulation data
The simulation setup also requires a value for a parameter termed particle restrain. In
reality particles are able to both roll and slide, sometimes independently and at other
times in combination. As a form of crude calibration, Chute MavenTM defines this
ability to roll or slide as particle restrain. A value of 100% particle restrain corresponds
104
to 100% of the particles within a simulation sliding, while a value of 0% particle

restrain corresponds to 0% of the particles within a simulation sliding (i.e. 100% are
rolling). The Chute MavenTM User Manual recommends that 100% particle restrain
should be used to provide a conservative result and that if the particles are set to 0%
particle restrain, then a very loose free-flowing material would be the end result
(Hustrulid, 2005). Another important point to make is that if a value of particle restrain
is selected somewhere between the two extremes then simulated particles are randomly
assigned, based on that percentage, to either roll or slide for the duration of the
simulation without changing. In reality this parameter is hard to measure, as under
different flow conditions variations in the degree of rolling or sliding of particles will
exist. The most direct method of quantification relies on visual inspection of particles
flowing through the transfer to investigate the degree of rolling and sliding there is
between particles and the boundary surfaces.
The remaining layer data in Figure 6.3 refers to other model features such as chutes,
skirts and liners. Each of these requires a coefficient of friction between the material
being tested and the corresponding surface material.
6.2.4 Performing a Simulation

Once the model parameters and simulation data have been entered, two files need to be
saved. The first is the DEM file, which is read back into Chute MavenTM at a later date
for further editing if required. The DEM file also embeds the DXF model into it so the
original file is no longer required. The second file is a DIN file, which is the file
required to execute the DEM simulation. Once these two files are saved, the simulation
is executed by selecting Run DIN Model from the file menu.
6.2.4.1 Optimisation of Simulation Time

The time required to perform a simulation is dependent on a number of factors,
including;
- computer processor speed,
- computer RAM,
- real time length of simulation,
- time steps per second of simulation time,
- particle size distribution, and
105
- the number of concurrent simulations being performed.

The computer workstation dedicated to the DEM simulations for this research is a
DELL Precision 670 which has the following main specifications;
- 2x Intel Xeon 3.4GHz single core processors,
- 2Gb RAM, and
- 2x 250Gb hard drives striped to form a 500Gb hard drive.
The inclusion of dual processors in the computer workstation allows for two instances
of Chute MavenTM to operate at 100% capacity at one time, one in each processor. If
more simulations are started, their use is distributed evenly across the two processors,
but this is to the detriment of processing speed. The length of a simulation combined
with the number of time steps per second will influence the time required to complete a
simulation. The particle size distribution of the material being simulated will also
impact on simulation time. Using a combination of mass flow rate, particle density and
particle size, the number of particles to be generated at each time step is determined.
The smaller the particles are, the more particles need to be generated at each time step
to obtain a particular flow rate and thus increases the number of calculations required.
The sensitivity analysis detailed in Section 6.2.7, comments on the influence of the
particle size distribution on both simulation time and variation to the observed flow
characteristics.
6.2.5 Interpreting Results

On completion of a simulation a POS file is generated which contains all the pertinent
information for the simulated particles for each time step. This file is viewed by opening
the DEM file associated with this simulation and then the POS file is loaded. The
simulation can be stepped through one time step at a time or one second of simulation
time at once. Video outputs or screen snapshots of the simulation can also be generated
by selecting the relevant option from the file menu.
The particles are assigned a spectrum of colours dependant on the velocity at which they
are travelling and it is visually straight forward to estimate the particle velocity at any
given location through the simulation. There is also an option to export all the particle
data for each time step to a Microsoft Excel spreadsheet for further analysis. Each
particle is assigned a particle ID in each time step, unfortunately the way the Chute
106
MavenTM software has been written, one particular particle does not keep the same
particle ID throughout the simulation. This unfortunately makes it impossible to track
one particle throughout the entire simulation.
6.2.6 Calibration of DEM at the Bench-Scale Level

Accurate calibration of the DEM simulation package is a must, as without this, there
can be no confidence that the outputs generated are accurate representations of that
which is occurring in reality. To overcome this uncertainty, it is highly recommended
that a series of bench-scale experimental tests be performed followed by corresponding
DEM simulations as validation. A range of static and dynamic tests were undertaken to
calibrate the Chute MavenTM discrete element modelling package and are explained
below. It should again be emphasised that Chute MavenTM can only simulate spherical
particles, thus results will in all likelihood vary slightly to the experimental equivalents
due to particle shape effects.
6.2.6.1 Slump Model

An experimental slump test was developed (Kamaras, 2007) to determine the particle
restrain of polyethylene pellets on a section of smooth conveyor belt under static
conditions. An acrylic tube (NB = 64 mm, L = 184 mm) was placed on the conveyor
belt and filled with material. Using a Redlake X3 MotionPro high-speed digital video
camera, the formation of the heap was recorded while the acrylic tube was slowly lifted
from the belt surface. A representative outcome is shown in Figure 6.4.
Figure 6.4 The results of an experimental slump test (Kamaras, 2007)
A corresponding three dimensional CAD model of the acrylic tube was produced,
comprising 23 ring cells and imported into Chute MavenTM. Each of these cells can be
removed, simulating the lifting of the tube. Two simulations were performed, for 0%
and 100% restrain. By interpolating the experimental results within the DEM results,
107
based on heap diameter, heap height and time to form the heap, a percentage restrain
was obtained for each of the three measured quantities and an overall particle restrain of
63% for the slump test was determined. A further DEM simulation was then performed
with this restrain and compared to the experimental equivalent, Figure 6.5. The
compared results showed an averaged error of 8.5%.
Figure 6.5 Comparison of (a) experimental slump test and

(b) DEM slump test with restrain = 63% (Kamaras, 2007)
6.2.6.2 Hopper Model

To evaluate the particle restrain under dynamic conditions, a small acrylic hopper was
produced, with the hopper outlet positioned 75 mm above a flat section of conveyor
belt. The hopper was filled with polyethylene pellets then the hopper outlet was opened,
allowing a heap to form on the belt. An experimental result is shown in Figure 6.6.
Figure 6.6 The results of an experimental hopper test (Kamaras, 2007)
108
The same method of DEM simulation was performed as for the slump test with the
particle restrain again varying from 0% to 100%. The interpolated results of the
experimental test to the results of the DEM simulation gave an overall particle restrain
of 88%. A further DEM simulation was then performed with this particle restrain and
compared to the experimental equivalent, Figure 6.7. The compared results showed an
averaged error of 5.2%.
Figure 6.7 Comparison of (a) experimental hopper test and

(b) DEM hopper test with restrain = 88% (Kamaras, 2007)
6.2.7 Sensitivity Analysis

Two of the simulation variables explained previously have some uncertainty associated
with them due to either the provided description in the Chute MavenTM instruction
manual or the way in which they need to be quantified. The coefficient of friction
between particles and the percentage particle restrain have both been investigated for
the trajectory geometry where only the influence of the conveyor belt on the particles is
present. The sensitivity analysis has been performed using the particle characteristics for
the polyethylene pellets and the findings obtained have been assumed to be
characteristic of other geometries and materials which were used throughout the
research program.
109
6.2.7.1 Trajectory Geometry

A 3D CAD model similar to that shown in Figure 6.1 was been used for these
comparisons. The main difference was the transfer chutes were removed and the
injection box was set further away from the head pulley of the feed conveyor. Three
groups of comparisons were made and are shown in Tables 6.1, 6.2 and 6.3.
Table 6.1 Sensitivity analysis based on 100% particle restrain
Belt Coefficient Material Particle size Percentage

speed of friction flow rate distribution restrain
(m/s) between (t/h) (radius) (%)
particles (mm)
0.5 0.222 0.111 1.675 2.375 100
Set 1
0.5 0.966 0.111 1.675 2.375 100
1.0 0.222 0.222 1.675 2.375 100
Set 2
1.0 0.966 0.222 1.675 2.375 100
2.0 0.222 0.444 1.675 2.375 100
Set 3
2.0 0.966 0.444 1.675 2.375 100
2.25 0.222 0.500 1.675 2.375 100
Set 4
2.25 0.222 5.000 1.675 2.375 100
2.25 0.966 0.500 1.675 2.375 100
Set 5
2.25 0.966 5.000 1.675 2.375 100
2.25 0.222 5.000 1.675 2.375 100
Set 6
2.25 0.966 5.000 1.675 2.375 100
The first group of comparisons, shown in Table 6.1, were all based on particles having
100% particle restrain. The mass flow rate for set 1, set 2 and set 3 was set at a low
value to achieve a thin trajectory curve to represent the lower trajectory stream.
Observations for these three sets showed that there was no difference to the position of
the trajectory stream for a given belt speed when only changing the coefficient of
friction between particles. Set 4 and set 5 compared a low mass flow rate with a higher
mass flow rate to investigate the influence of the coefficient of friction between
particles when there were more particles present in the material stream. The lower
trajectory curve for both sets again showed no difference. Figure 6.8 and Figure 6.9
showed the two trajectory curves for set 4. Set 6 compared the coefficient of friction
between particles for the higher mass flow rate, the result being identical curves. For all
six sets presented in Table 6.1, the particle velocity was also compared to determine if
either by changing the coefficient of friction between particles or the mass flow rate,
there was any appreciable change to the stream velocity. The conclusion was that there
110
was no change to the particle velocity at corresponding vertical displacements in the

trajectory streams.
The second group of comparisons, shown in Table 6.2, were all based on particles
having 50% particle restrain. Both set 7 and set 8 had a low mass flow rate, again to
produce a single particle stream to represent the lower trajectory curve. For set 7 it was
noticed that for a low coefficient of friction between particles, there were more particles
dropping away from the trajectory stream. For a high coefficient of friction between
particles, the trajectory stream widened as the vertical displacement increased. For set 8
there was again a noticeable increase in particles dropping away from the trajectory
stream for a low coefficient of friction between particles, as can be seen in Figure 6.10
and Figure 6.11. Other than this difference, the trajectory streams for both set 8
conditions were identical. The particle velocities throughout the trajectory streams for
set 7 and set 8 were also identical and were also identical to the sets of Table 6.1.
Table 6.2 Sensitivity analysis based on 50% particle restrain

particles (mm)
1.0 0.222 0.25 1.675 2.375 50
Set 7
1.0 0.966 0.25 1.675 2.375 50
2.0 0.222 0.50 1.675 2.375 50
Set 8
2.0 0.966 0.50 1.675 2.375 50
The last group of comparisons compared sets which only varied by the percentage of
particle restrain, as shown in Table 6.3. When comparing the two tests of each set, it
was found that the same trends were present for all four sets. For 50% particle restrain
there was noticeably more scatter of particles in the trajectory stream as well as more
particles dropping away from the stream than for the 100% particle restrain cases.
Comparing the particle velocities for each of the sets again found that the particle
velocity was consistently the same.
111
Figure 6.8 Set 4 trajectory curve for Figure 6.9 Set 4 trajectory curve for
mass flow rate, ms = 0.5 t/h mass flow rate, ms = 5 t/h
Figure 6.10 Set 8 trajectory curve with Figure 6.11 Set 8 trajectory curve with
coefficient of friction between coefficient of friction between
particles, p = 0.222 particles, p = 0.966
112
Table 6.3 Sensitivity analysis based on comparisons between

50% and 100% particle restrain

particles (mm)
1.0 0.222 0.25 1.675 2.375 50
Set 9
1.0 0.222 0.25 1.675 2.375 100
1.0 0.966 0.25 1.675 2.375 50
Set 10
1.0 0.966 0.25 1.675 2.375 100
2.0 0.222 0.50 1.675 2.375 50
Set 11
2.0 0.222 0.50 1.675 2.375 100
2.0 0.966 0.50 1.675 2.375 50
Set 12
2.0 0.966 0.50 1.675 2.375 100
The overall conclusions made for the trajectory geometry sensitivity analysis is that the
coefficient of friction between particles has little to no influence on the resulting
trajectory stream. Also, with respect to the percentage material restrain, the more rolling
that is included in the particle restrain, the more particles deviate from the main flow
stream of material.
6.2.7.2 Effect of Particle Size Distribution

The particle size distribution (PSD) of the polyethylene pellets used for the sensitivity
analysis in Section 6.2.7.1 ranged in radius from 1.675 mm to 2.375 mm, which
accounted for 97.1% of the material. As mentioned in Section 6.2.4.1, the smaller the
particle size, the more time simulations require to complete. As a further analysis, the
effect of changing the particle size distribution to a radius of 2.00 mm to 2.375 mm
(85.37% of the material) on the simulation time as well as the flow characteristics was
investigated. The results of twelve tests have been summarised in Table 6.4 with
observations following.
The first and second simulations of set 13 showed;

- an identical particle velocity at the same vertical displacement from the head
pulley,
- an identical trajectory stream profile,
- very few particles deviated from the main trajectory stream, and
- both had a discharge angle of 38.9.
113
The third simulation of set 13 was a repeat of the second and it was found that there
were an identical number of particles generated and a frame by frame comparison
showed there to be an identical match, that is to say when the particles were initially
generated, they must be determined from a random number generator which reset each
time a simulation was commenced. The simulation time for the 2.00 mm to 2.375 mm
PSD was 2.5 times quicker than for the 1.675 mm to 2.375 mm PSD.
Table 6.4 Sensitivity analysis based on comparisons between

97.1% and 85.37% particle size distributions
Belt Coefficient Material Particle size Percentage Number Time

speed of friction flow rate Distribution restrain of (min)
(m/s) Between (t/h) (radius) (mm) (%) Particles*
particles
1.0 0.222 0.25 1.675 2.375 100 2100 337
Set 13 1.0 0.222 0.25 2.000 2.375 100 1720 132
1.0 0.222 0.25 2.000 2.375 100 1720 135
2.0 0.966 0.50 1.675 2.375 100 2958 269
Set 14
2.0 0.966 0.50 2.000 2.375 100 2389 130
2.0 0.222 5.00 1.675 2.375 100 29538 6240
Set 15
2.0 0.222 5.00 2.000 2.375 100 23930 2992
2.0 0.966 0.25 1.675 2.375 50 2109 205
Set 16
2.0 0.966 0.25 2.000 2.375 50 1700 97
1.0 0.966 0.25 1.675 2.375 50 2037 336
Set 17
1.0 0.966 0.25 2.000 2.375 50 1676 166
Set 18 1.0 0.966 7.5 1.675 2.375 100 63287 31680
* Number of particles refers to highest count during the steady state period
The two simulations of set 14 showed;

pulley,
- an identical stream profile, and
- high-speed discharge.

pulley,
- an identical stream profile, and
- high-speed discharge.
114

pulley,
- an identical stream profile,
- high-speed discharge, and
- increased presence of particles deviating from the main trajectory stream due to
the percentage particle restrain being reduced.

pulley,
- an identical trajectory stream profile,
- both had a discharge angle of 38.8, and
- increased presence of particles deviating from the main trajectory stream due to
the percentage particle restrain being reduced.
Regardless of the change in percentage particle restrain for set 13 and set 17, the
simulations showed;
- there is no change to the particle velocity at the same vertical displacement,
- effectively the same discharge angle, and
- the same trajectory path (lower path due to small mass flow rate)
Set 14, set 15 and set 16 are compared based on each set being generated using the same
belt speed and showed;
- the lower trajectory stream profile to be the same (the upper trajectory could not
be compared due to varying mass flow rates), and
- all simulations had the same particle velocity at the same vertical displacement
from the head pulley.
For set 18 only one simulation was performed due to the simulation time required for a
mass flow rate of 7.5 tph. The findings showed;
pulley as for set 13 and set 17,
- no particles deviated from the trajectory stream,
115
- a discharge angle of 38.7 (effectively the same as set 13 and set 17),
- a lower trajectory stream practically identical to set 13 and set 17, and
- a noticeable divergence of the trajectory stream as the vertical displacement
increases (thickness of material stream at discharge was 30 mm and at a drop
height of 1365 mm was 55.8 mm).
It should also be noted at this point that in the above sensitivity analysis, the highest
mass flow rate simulated was 7.5 tph. It was found that at higher mass flow rates that
Chute MavenTM would crash at random times during the simulations. Feedback was
sought from the software vendor, with the initial indication being that arrays within the
source code may not have been set large enough to contain the number of particles
required in the simulations being executed. Unfortunately, no solution to the problem
was provided, thus limiting the tangible outputs possible.
6.3 EDEM
The EDEM discrete element modelling software by DEM Solutions required a more in-
depth setup before a simulation could be generated. The main steps involved are listed
below and additional information on each is provided in the following sections,
- set the global model parameters
- define the particles
- define the geometry
- define the domain
- create the particle factory
- run the simulation
- analyse the simulation.
6.3.1 Computing Power

EDEM could successfully operate on the existing computer workstation, detailed in
Section 6.2.4.1, making full use of the two processors due to its multi-processor/multi-
core coding. Additionally, to fully utilise the software, a new DELL workstation was
purchased with the following main specifications;
- 2x Intel Xeon 3.0GHz processors, each with 4 cores,
- 16Gb RAM, and
- 2x 250Gb SAS hard drives striped to form a 500Gb hard drive.
116
With the extra processing power and RAM, simulations could be completed in a much
quicker time with the added benefit that many more particles could be simulated in a
given simulation.
6.3.2 Global Model Parameters

The global parameters include the following,
- simulation name and description,
- choosing the contact model to use in the simulation, there are several available
including, linear-spring and Hertz-Mindlin, with the ability to add additional
contact models if required,
- setting gravity and the properties of the materials used in the geometry and the
particles properties (i.e. Poissons ratio, shear modulus and particle density), and
- defining the interaction between materials, such as particle to particle and particle
to wall.
6.3.3 Particle Definition

EDEM has the ability to not only simulate with spherical particles, but to incorporate
shape. This is possible by importing CAD models of particles, into which any number
of spherical particles can be placed to represent a real particle. It is important to
remember that the more spherical particles needed to produce a realistic particle, the
longer time a simulation will require to complete. Scaling factors can be applied to the
particles as well.
6.3.4 Defining the Geometry

A three dimensional CAD model of the conveyor geometry needs to be imported, in
STP format, into the EDEM software. Each of the layers can then be named, such as
belt, hopper, hood and spoon. The belt is assigned as a moving plane with the velocity
appropriate for a given simulation to mimic that occurring on the conveyor transfer
research facility. Particles enter the simulation through a factory plate. This needs to be
defined for both size and position. It is a virtual surface as it is not a physical part of the
system.
117
6.3.5 Defining the Domain

The domain is the volume in which particles are visible and simulated within a system.
By default, the domain will be created based on the extents of the boundaries of the
imported CAD model; however, the domain can be manually set to any size larger or
smaller. This can be of benefit if a particular section of an entire system is to be
investigated only because particles are only simulated within the domain. Having a
smaller domain means that fewer particles will be produced, thus reducing the time
required for simulation.
6.3.6 Particle Factory

The particle factory defines how the particles are introduced during the simulation. Two
methods can be used; the first is to have an infinite number of particles produced at a
pre-defined rate per second and the second method is to have a pre-defined number of
particles introduced over the entire simulation at a set number of particles per second.
The particle factory is aligned with the factory plate which has already been created and
the particles should be generated randomly over the entre plate area.
6.3.7 Running a Simulation

Before starting a simulation, a number of time options need to be set. The Rayleigh time
step is calculated, using equation 2.146. For a given simulated particle, the particle
radius, rp, particle density, s, shear modulus, G, and Poissons ratio, , are constant, but
what would be the effect of altering any of these? If the particle radius is increased (i.e.
scaled up), this has a linear effect on the resulting Rayleigh number. If the shear
modulus is altered, then being the denominator in a square root function, this has the
potential to alter the Rayleigh number substantially. Modifying the Poissons ratio will
have a small effect only. The particle density should remain constant. Section 6.3.9
deals with these issues further.
The total simulation time and the frequency of data being output to file is also required.
The total number of simulation time steps within a simulation is the product of the
Rayleigh time step and the total simulation time. In reality, the data produced from
every time step is not required, a value of 0.1 second may be more than adequate. The
more frequently the data is saved, the slower the simulation will run.
118
Within the simulation domain there is a grid. This grid is used in the contact detection
algorithms and the larger an individual grid is, the more particles it can theoretically
contain. The more particles there are in a simulation, the more collisions that can occur,
which in turn slows down the simulation. It is recommended that the grid size be kept to
a minimum of two times the minimum particle radius to keep the simulation progressing
with minimal speed reduction.
Once these values have been entered, the simulation is ready to start. An additional
feature of the EDEM software is the ability to stop, or pause, the simulation at any stage
and restart. This is especially useful to check that a simulation is progressing how it was
intended, instead of having to wait until the entire simulation has finished before being
able to review it. This feature can also be of great benefit if the simulation has already
completed what was required before completing the preset simulation time, saving
precious time and also allowing subsequent simulations to be started sooner.
6.3.8 Simulation Analysis

On completion of a simulation the EDEM Analyst can be evoked, from which a
multitude of functions can be applied to the data. The view of the geometry can be
altered from solid to opaque to wire mesh. The particle template which was used in the
initial definition of the particles can be applied to the simulation to give a realistic look
to the output. The colour of the particles can also be varied to represent such parameters
as kinetic energy or particle velocity.
The data can also be extracted from the EDEM software as a comma delimited text file
which can be read by Microsoft Excel where post processing can occur.
6.3.9 Sensitivity Analysis

There are numerous particle properties which need to be set during the setup of DEM
simulations and to justify the validity of the outputs generated, these parameters must be
realistic. The validity of some of these parameters has already been investigated in the
sensitivity analysis performed using Chute MavenTM, but key parameters which are only
used in EDEM, namely the shear modulus an Poissons ratio, have been investigated
due to the inability to determine exact values for the test materials, instead, relying on
estimates obtained in the literature, as summarised in Section 5.2.4.
119
6.3.9.1 Sensitivity Analysis for Polyethylene Pellets

As an example of the potential to alter the Rayleigh time step, three calculations will be
performed using the data in Table 6.5. The particle density has been set constant at
919.2 kg/m3 and the particle radius will remain constant at 2.375 mm for the three trials.
Table 6.5 Variables used to investigate the sensitivity of the Rayleigh time step
Shear Modulus, Poissons Ratio, Rayleigh Time
G, (GPa) Step (s)
Trial 1 1.17 x 108 0.45 2.20 x 10-5
Trial 2 1.17 x 106 0.45 2.20 x 10-4
Trial 3 1.17 x 108 0.35 2.24 x 10-5
What is clear from this comparison is that for a 100-fold reduction in the shear modulus
there is a 10-fold reduction in the Rayleigh time step (with all other variables kept
constant). It is also clear to see that for a change in the Poissons ratio from 0.45 to 0.35,
there is only a very minimal change in the Rayleigh time step.
The system geometry used for the sensitivity analyses is shown in Figure 6.12 and
consists of the conveyor belt feeding onto the transfer spoon with the belt speed being
set to 1 m/s for all simulations.
Figure 6.12 System geometry used for sensitivity analysis
120
Both spherical particles and shaped particles have been considered in this sensitivity
analysis. Fifteen simulations were generated for the spherical particles and fifteen
simulations were generated for the shaped particles as summarised in Table 6.6. The
particle radius has been kept constant and is that found in Table 5.1.
The highlighted grey boxes in Table 6.6 indicate the values which were changed in each
simulation. The first nine simulations were performed by systematically altering the
shear modulus for the materials within the system geometry, i.e. the Polystone liner in
the transfer spoon, the acrylic transfer spoon, the AerobeltTM and the polyethylene
pellets. The data for test 1 is the data used in the EDEM simulations contained in
subsequent chapters but to gauge whether these parameters were representative; each
shear modulus value was reduced by a factor of ten in the proceeding simulation until
the order of magnitude reached 106. Tests 10 to 12 applied the shear modulus values
used in test 1 (as the upper limit) and altered the Poissons ratio from 0.45 to 0.40 to
0.35 to investigate the effect on simulation outputs. Tests 13 to 15 applied the shear
modulus values used in test 9 (as the lower limit) and altered the Poissons ratio from
0.45 to 0.40 to 0.35 to investigate the effect on simulation outputs. Figure 6.13 displays
a steady-state output for each spherical particle simulation. The following observations
were noted as a result of the spherical particle simulations;
- when comparing the particle streams for tests 1 to 9, there was a slight reduction
in the width of the stream at the point of discharge from the head pulley (6.9%)
and very little difference to the width of the stream near the exit of the transfer
spoon (2.6%), indicating that the variation in shear modulus does not have a
major influence over the dynamics of the stream,
- when comparing test 1 to tests 10 to 12, where all shear modulus values were
identical and only Poissons ratio was varied, it was found that there was a
minimum reduction in the width of the stream at the point of discharge from the
head pulley (3.4%) and no difference to the width of the stream near the exit of
the transfer spoon, indicating that the variation in Poissons ratio has very little
influence over the dynamics of the stream,
121
Polystone Acrylic AerobeltTM Polyethylene Pellets
Rayleigh Time Step Time Time Step (sec)

TEST Shear Shear Shear Shear (sec) Step
Poissons Poissons Poissons Poissons
Modulus Modulus Modulus Modulus (%)
Ratio Ratio Ratio Ratio
(Pa) (Pa) (Pa) (Pa)
Spherical Shaped Spherical Shaped
8 9 7 8 -5 -5 -6
1 3.33x10 0.4 1.4x10 0.4 3.5x10 0.45 1.17x10 0.42 2.21x10 2x10 30 6.64x10 6.01x10-6
2 3.33x107 0.4 1.4x109 0.4 3.5x107 0.45 1.17x108 0.42 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
3 3.33x106 0.4 1.4x109 0.4 3.5x107 0.45 1.17x108 0.42 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
4 3.33x106 0.4 1.4x108 0.4 3.5x107 0.45 1.17x108 0.42 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
5 3.33x106 0.4 1.4x107 0.4 3.5x107 0.45 1.17x108 0.42 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
6 3.33x106 0.4 1.4x106 0.4 3.5x107 0.45 1.17x108 0.42 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
7 3.33x106 0.4 1.4x106 0.4 3.5x106 0.45 1.17x108 0.42 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
8 3.33x106 0.4 1.4x106 0.4 3.5x106 0.45 1.17x107 0.42 7x10-5 6.3x10-5 30 2.1x10-5 1.9x10-5
9 3.33x106 0.4 1.4x106 0.4 3.5x106 0.45 1.17x106 0.42 2.21x10-4 2x10-4 30 6.64x10-5 6.01x10-5
10 3.33x108 0.45 1.4x109 0.45 3.5x107 0.45 1.17x108 0.45 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
11 3.33x108 0.4 1.4x109 0.4 3.5x107 0.4 1.17x108 0.4 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
12 3.33x108 035 1.4x109 035 3.5x107 035 1.17x108 035 2.21x10-5 2x10-5 30 6.64x10-6 6.01x10-6
13 3.33x106 0.45 1.4x106 0.45 3.5x106 0.45 1.17x106 0.45 2.21x10-4 2x10-4 30 6.64x10-5 6.01x10-5
Table 6.6 Sensitivity analysis settings for polyethylene pellets
14 3.33x106 0.4 1.4x106 0.4 3.5x106 0.4 1.17x106 0.4 2.21x10-4 2x10-4 30 6.64x10-5 6.01x10-5
15 3.33x106 035 1.4x106 035 3.5x106 035 1.17x106 035 2.21x10-4 2x10-4 30 6.64x10-5 6.01x10-5
122
(a) Test 1 (b) Test 2 (c) Test 3 (d) Test 4 (e) Test 5 (f) Test 6
(g) Test 7 (h) Test 8 (i) Test 9
(j) Test 10 (k) Test 11 (l) Test 12 (m) Test 13 (n) Test 14 (o) Test 15
Figure 6.13 Steady-state EDEM outputs from the 15 sensitivity analysis tests for
spherical particles
identical and only Poissons ratio was varied, it was found that there was no
change in the width of the stream at the point of discharge from the head pulley
123
and no difference to the width of the stream near the exit of the transfer spoon,
indicating that the variation in Poissons ratio has no influence over the
dynamics of the stream at these low shear modulus values,
- it has also been noted that there was no change to the Rayleigh time step when
changing the shear modulus values for the materials making up the conveyor
geometry but there was an increase in the Rayleigh time step once the shear
modulus of the polyethylene pellets was reduced. It was found that when the
shear modulus was reduced by a factor of 100 times, the Rayleigh time step
decreased by a factor of 10. This has the effect of speeding up simulations
substantially,
- there was very little change to the Rayleigh time step when altering only the
Poissons ratio values.
These findings were also applicable to the results for the shape based particle
simulations, the steady-state outputs from the shaped particle simulations are shown in
Figure 6.14.
- when comparing the particle streams for tests 1 to 9, there was a slight reduction
in the width of the stream at the point of discharge from the head pulley (6.7%)
and very little difference to the width of the stream near the exit of the transfer
spoon (4.2%), indicating that the variation in shear modulus does not have a
major influence over the dynamics of the stream,
difference in the width of the stream at the point of discharge from the head
pulley and no difference to the width of the stream near the exit of the transfer
spoon, indicating that the variation in Poissons ratio has very little influence
over the dynamics of the stream,
change in the width of the stream at the point of discharge from the head pulley
and no difference to the width of the stream near the exit of the transfer spoon,
indicating that the variation in Poissons ratio has no influence over the
dynamics of the stream at these low shear modulus values,
124
- the same findings for the Rayleigh time step apply to the shaped particle
simulations as were found with the spherical particle simulations.
(a) Test 1 (b) Test 2 (c) Test 3 (d) Test 4 (e) Test 5 (f) Test 6
(g) Test 7 (h) Test 8 (i) Test 9
(j) Test 10 (k) Test 11 (l) Test 12 (m) Test 13 (n) Test 14 (o) Test 15
Figure 6.14 Steady-state EDEM outputs from the 15 sensitivity analysis tests for
shaped particles
125
The results of these two sets of sensitivity analyses have indicated that varying the shear
modulus and Poissons ratio of the materials contained within an EDEM simulation do
not greatly influence the dynamic behaviour of the particle stream, with the exception of
the shear modulus of the particles where reducing the shear modulus of the polyethylene
pellets saw an increase in the Rayleigh time step, which speeds up simulation time. This
will be as a direct result of softening the particles, which in turn means that
particle/particle and particle/geometry collisions will take longer to occur and thus a
larger time step can be used and still capture the dynamics of the collision.
An additional comment which can be made as a result of this analysis is that

approximate values for shear modulus and Poissons ratio can be used in lieu of specific
values for the test materials being used and still be able to obtain reliable simulation
outputs, especially if the facility for determining these values are not available. In
saying this, the values selected should be based on similar materials, keeping the order
of magnitude as representative as possible.
The sensitivity analysis presented has been applied for polyethylene pellets and it has
been assumed the findings/trends obtained can be adapted to the other test materials
without repeating the process specifically for iron ore or corn. Understandably, this is a
substantial call to make as there are a variety of differences between the test materials
(e.g. particle shape, size and density, shear modulus, etc.). Time constraints were an
additional factor in making this decision. Further research is warranted in this area as
ultimately, investigations of this type are vital in defining the particle behaviour of the
discrete element simulations.
126
CHAPTER 7 CONVEYOR DISCHARGE ANGLES
Chapter
7
CONVEYOR DISCHARGE ANGLES
127
7.1 INTRODUCTION
As discussed in Chapter 2, there are a number of flow mechanisms common in
conveyor transfers. The first flow mechanism of importance is that at the conveyor
discharge. This is a vital parameter when low-speed conveying conditions apply,
dictating the angle at which material will leave the head pulley and begin its trajectory.
For high-speed conveying conditions, the discharge angle will always be at the point of
tangency between the belt and the head pulley.
This chapter presents three approaches to determining the material discharge angle;
experimental measurements of conveyor discharge angles taken from the conveyor
transfer research facility, those predicted by trajectory methods, and finally the
discharge angles measured from DEM simulations.
7.2 CRITICAL BELT SPEEDS

The critical belt speed, touched on in Chapter 2, refers to the point of transition from
low-speed to high-speed conveying. The equations presented in each of the trajectory
models have been reorganised and are presented in Table 7.1.
Table 7.1 Critical belt speeds for the various methods
gRb 2 cos b
C.E.M.A. and M.H.E.A. Vb ,cr = 7.1
Rc
Booth and Golka (lower) Vb ,cr = gRb cos b 7.2
2h
Golka (upper) Vb ,cr = gRb 1 + cos b 7.3
Rb

Korzen Vb ,cr = gRc cos b + a 7.4
h
Goodyear Vb ,cr = gRc cos b 7.5
As previously explained, Golka (1992; 1993) determined two distinct discharge angles
(for lower and upper trajectory streams) which also implied there were two critical belt
speeds. The method by Korzen (1989) incorporated an adhesive stress component,
128
however, if the adhesive stress equals zero, the equation is identical to that of Goodyear
(1975). For the Dunlop method (1982), the critical belt speed is graphically determined
and the inference provided by the worked examples is that the lower and upper
discharge angles are identical. There are limitations to its determination with only
pulley diameters between 312 mm and 1600 mm able to be used.
7.3 CONVEYOR DISCHARGE ANGLE MODEL COMPARISONS

An arbitrary set of parameters, see Table 7.2, has been selected typical of conveyor belt
designs and also material characteristics, taken from several sources (Korzen, 1989;
Golka, 1993; C.E.M.A., 2005). The purpose of generating this data was to produce a
preliminary investigation of conveyor discharge angles from the range of analytical
models explained in Section 2.4. This data has no link to the conveyor transfer research
facility or the test materials used. Parameters not included in Table 7.2 include; belt
inclination angle, b, head pulley diameter, Dp, and belt velocity, Vb, as these
parameters are varied to allow a range of comparisons to be made.
Table 7.2 Parameters used for comparisons

Belt width, bw 0.762 m
Belt thickness, bt 0.01 m
Surcharge angle 20
Trough angle 20
Coefficient of static friction, s 0.5
Coefficient of kinetic friction, k 0.42
Product density, b 2000 kg/m3
Specific gravity of bulk solids, 19.62 kN/m3
Adhesion stress, a 0 kPa
Centroid height (CEMA 1,2,4,5), a1 0.04064 m
Centroid height (CEMA 6), a1 0.04191 m
Centroid height (MHEA86), a1 0.04 m
Material height (CEMA 1,2,4,5), h 0.09652 m
Material height (CEMA 6), h 0.10287 m
Material height (MHEA86), h 0.096 m
7.3.1 Effect of Belt Inclination Angle on Critical Belt Speed

A quantitative comparison of the critical belt speeds has been undertaken for the
discharge angle methods and is presented in Figure 7.1 for a range of head pulley
diameters. These figures clearly show the variation in belt speed as belt inclination
129
angle increases. Although the actual variation of belt speed for any given method and
pulley diameter is quite small (i.e. 0.1 to 0.25 m/s), there will be an effect on the
speed at which the transition from low-speed to high-speed conditions occurs. The
lower stream Dunlop (1982) curves do not appear to follow the trend of any one method
as is evident in Figure 7.1b and Figure 7.1c, where the Dunlop curve follows the Golka
L (1992; 1993) / Booth (1934) curve and then the C.E.M.A. (1966; 1979; 1994; 1997;
2005) / M.H.E.A. (1986) curve respectively. The upper Dunlop (1982) curve loosely
follows the upper Golka (1992; 1993) curve in Figure 7.1a and Figure 7.1b but no upper
critical belt speeds could be determined for the 1.5 m diameter pulley due to the actual
diameter of the outer surface of the material being 1.726 m which is outside the limits of
the Dunlop graphical method.
7.3.2 Effect of Belt Speed and Pulley Diameter on Conveyor

Discharge Angle
Applying the parameters of Table 7.2 to the discharge angle methods for a horizontal
conveyor, comparisons were made for four head pulley diameters, 0.5 m, 1.0 m, 1.5 m
and 2.0 m and for belt velocities ranging from 0.50 m/s to 3.25 m/s in 0.25 m/s
increments. Figure 7.2 displays a range of belt velocities (0.5 m/s to 1.75 m/s) for a
head pulley diameter of Dp = 0.5 m with the following observations being made:
a) Figure 7.2a shows two distinct groupings of discharge angles with slight
variations evident in both groups;
b) As the belt velocity increases, there is a noticeable spread in the discharge angles,
moving away from the initial two groupings, see Figure 7.2a to Figure 7.2e;
c) Figure 7.2e shows that some methods, C.E.M.A. (1966; 1979; 1994; 1997; 2005),
M.H.E.A. (1986) and Golka (upper) (1992; 1993) have already reached high-
speed conditions (i.e. discharge at the point of tangency between the belt and head
pulley);
d) Golka (upper) (1992; 1993) reaches high-speed conditions based on the calculated
tangential velocity of the upper stream, whereas Golka (lower) (1992; 1993) is
still under low-speed conditions;
e) Figure 7.2f shows that all discharge angle methods are now under high-speed
conditions (Vb = 1.75 m/s), referring to Figure 7.1a for the critical belt velocity
where the transition from low-speed to high-speed conditions occurs;
130
(a) (b) (c)
Figure 7.1 Critical belt speed for (a) Dp = 0.5 m, (b) Dp = 1.0 m, (c) Dp = 1.5 m
CEMA 1,2,4,5 BOOTH DUNLOP

CEMA 6 GOLKA GOODYEAR
MHEA86 KORZEN
Dimensions in metres
(a) Vb = 0.50 m/s (b) Vb = 0.75 m/s (c) Vb = 1.00 m/s
(d) Vb = 1.25 m/s (e) Vb = 1.50 m/s (f) Vb = 1.75 m/s
Figure 7.2 Variation in discharge angle based on belt speed

for a pulley diameter of Dp = 0.5 m
131
f) Referring to the two discharge angles for Dunlop (1982) in Figure 7.2e, there is an
indication that there is a high convergence of the lower and upper paths based on
the assumption made that two distinct discharge angles should be determined (as
explained previously). There will in all likelihood be a crossing of the lower and
upper trajectory streams which in reality would not occur. Of course, this situation
only occurs under low-speed conditions. For high-speed conditions (Figure 7.2f),
the streams are parallel.
Further comparisons are presented in Figure 7.3, Figure 7.4 and Figure 7.5 displaying
other combinations of pulley diameter and belt speed. Additional observations have
been made from these comparisons:
g) For a constant belt velocity, the discharge angle for a given method increases as
the pulley diameter increases, see Figure 7.3a to Figure 7.3c or Figure 7.4a to
Figure 7.4c;
h) As previously stated, the upper discharge angle for the Dunlop (1982) method is
displaying a high-speed condition once a pulley diameter of 1.5 m has been
reached, due to limitations with the Dunlop graphical method;
i) Enforcing the statement made in (b) above, as belt velocity increases for a given
pulley diameter, there is more spread in the discharge angles determined, see
Figure 7.3b and Figure 7.4b and also Figure 7.3c, Figure 7.4c and Figure 7.5c;
j) As the belt velocity / pulley diameter ratio decreases for a given belt velocity, the
two groupings of discharge angles become more defined, see Figure 7.3a to
Figure 7.3c and Figure 7.4a to Figure 7.4c;
k) As pulley diameter increases there is a wider range of belt velocities available to
produce low-speed conditions.
132
CEMA 1,2,4,5 BOOTH DUNLOP

CEMA 6 GOLKA GOODYEAR
MHEA86 KORZEN
Dimensions in metres
(a) Dp = 1.0 m (b) Dp = 1.5 m (c) Dp = 2.0 m
Figure 7.3 Variation in discharge angle based on pulley diameter

for a belt speed of Vb = 1.00 m/s
(a) Dp = 1.0 m (b) Dp = 1.5 m (c) Dp = 2.0 m

(a) Dp = 1.0 m (b) Dp = 1.5 m (c) Dp = 2.0 m

133
7.3.3 Effect of Static and Kinetic Friction

The coefficient of static and kinetic friction is only used in the determination of the
discharge angle of material leaving the head pulley of a conveyor belt under low-speed
conditions in the Booth (1934) and Korzen (1989) methods. Increasing these frictional
values will result in material wrapping around the head pulley to a larger degree,
whereas if the frictional values are decreased, the material will wrap around the head
pulley to a lesser degree before discharge.
7.3.4 Effect of Adhesive Stress on Conveyor Discharge Angle

The Korzen (1989) method includes adhesive stress, however, for the comparisons
presented above, the adhesion stress has been set to zero to keep comparisons consistent
against the other methods. As an indication of the effect of varying the adhesive stress, a
pulley diameter of 1.0 m was selected and a range of adhesive stresses from 0 to 2 kPa
was applied, see the results in Figure 7.6. It can clearly be seen that as the adhesion
stress is increased for any given belt speed, the resulting discharge angle increases. Also
as the adhesive stress increases there is also an increase in the range of belt speeds
before the transition from low-speed to high-speed conveying occurs. If the adhesion
stress is increased to 2.5 kPa for this set of comparisons, a discharge angle of 91.3
results for a belt speed of 0.5 m/s which is obviously at an angle past the most
horizontal point on the head pulley and would in actual fact result in material doubling
back on itself.
Figure 7.6 Effect of adhesive stress on discharge angle
134
7.4 DETERMINING THE CONVEYOR DISCHARGE ANGLE
7.4.1 Experimental Determination

As explained in Section 2.3, the conveyor discharge angle is the angle at which material
leaves the conveyor head pulley. The discharge angle is affected by both the diameter of
the head pulley and the belt speed. The diameter of the head pulley on the conveyor
transfer research facility has an OD of 240 mm including the rubber lagging. This
diameter has also been used in the analytical and numerical models (discussed later).
Under low-speed conveying conditions, material stays in contact with the belt as it
travels around the head pulley for a certain angle of wrap, dependant on belt speed. For
high-speed conveying conditions, the material leaves the conveyor at the point of
tangency between the belt and the head pulley. The experimental determination of the
discharge angle for each test material was performed for various belt speeds. The
conveyor transfer was removed from the transfer zone to allow the material to drop
unhindered from the feed conveyor to the receiving conveyor, see Figure 7.7.
To allow accurate comparisons between the methods explained in Section 2.3, the
conveyor belt needs to be loaded to full capacity. This requires the determination of the
edge distance, which is the width of belt on either side of the material profile which is
left exposed, to minimise spillage, especially through the transition zone where material
slumps to some degree as the belt flattens and wraps around the head pulley. The
method of C.E.M.A. (2005) for determining edge distance, c, is given by equation 7.6,
where b is the width of the belt. For a 300 mm wide belt, the edge distance is
approximately 39 mm.
c = 0.055b + 0.9 7.6
Belt speeds were selected between the range 0.5 m/s and 2.25 m/s in 0.25 m/s steps.
Once the belt speed had been set, the HoganTM valve was opened and adjusted, for each
belt speed, to obtain an edge distance of 39 mm. The intention of this investigation was
to view bulk behaviour rather than individual particle behaviour. A standard 25 frame
per second digital video camera, mounted perpendicular to the discharge, was used to
capture the material flow. A single frame from each video was then extracted and
analysed manually using Adobe Photoshop to determine the discharge angle for each
135
velocity. Divisions were marked on the acrylic panels at 0, 30, 60 and 90, 0 being at
the tangent between the belt and head pulley, accounting for the inclination angle of the
conveyor. For slow-speed conveying, the angle of discharge is measured from a vertical
reference as shown in Figure 7.8.
Figure 7.7 Setup for determining material discharge angle and trajectory
Figure 7.8 Low-speed discharge of polyethylene pellets at Vb = 1.0 m/s
7.4.2 Determination by Analytical Method

Four of the methods of determining the conveyor discharge angle allow direct
calculation based on the experimental head pulley diameter of 240 mm, Booth (1934),
Golka (1992; 1993), Korzen (1989) and Goodyear. However, C.E.M.A. (1966; 1979;
1994; 1997; 2005), M.H.E.A. (1977; 1986) and Dunlop (1982) do not, as the head
pulley diameter is not within the range of either the tabulated data or graphical limits.
C.E.M.A. (2005) provides a series of equations which allows for the approximate
136
determination of the centroidal height, a1, of the cross sectional area of the material
stream at discharge, combining these equations results in that shown in equation 7.7,
where C is the width and h is the height of the material stream at discharge.
a1 =
(13C h + 12h )
2 3
7.7
( 32C + 24h )
2 2
Based on this equation, both the C.E.M.A. method and the M.H.E.A. method can be
included in these comparisons. As the earlier editions of the C.E.M.A. guide (1966;
1979; 1994; 1997) do not include formulae to determine the centroid height, the same
values of a1 and h have been used for all editions. This results in the same discharge
angle for all methods. This is also true of the M.H.E.A. (1986) guide where the same a1
and h values have been used.
7.4.3 Determination by Discrete Element Modelling

The discharge angle can be measured from the graphical outputs of the DEM
simulations for either Chute MavenTM or EDEM. In Chute MavenTM, this is achieved by
aligning the graphical simulation output to a profile view of the material flow from the
conveyor head pulley and creating a still image. Using Image Pro Plus, the angle at
which material leaves the head pulley can then be measured. In EDEM, graphing
functions are used to produce the xy-coordinates of the material flow at the discharge
point.
7.5 CONVEYOR DISCHARGE ANGLE FOR POLYETHYLENE

PELLETS
7.5.1 Experimentally Determined Conveyor Discharge Angles

As described in Section 7.4, the conveyor discharge angle is analysed from a single
frame of the video footage taken for each belt speed. Table 7.3 summarises the
discharge angles for the range of belt speeds tested. At a belt speed of 2.00 m/s, the
material is discharging from the point of tangency between the belt and the head pulley,
indicating high-speed conveying conditions have been reached.
137
Table 7.3 Experimentally determined discharge angles for polyethylene pellets
Belt Speed (m/s) Discharge Angle*

0.50 65.1
0.75 61.6
1.00 39.1
1.25 16.6
1.50 5.2
1.75 0
2.00 tangency
2.25 tangency
* Discharge angle measured from the vertical
in the direction of pulley rotation
7.5.2 Analytically Determined Conveyor Discharge Angles

The analytical models used to determine conveyor trajectories (see Chapter 8) have also
been used to determine the conveyor discharge angles of the polyethylene pellets, using
the relevant data from Table 5.1. The results are presented in Table 7.4 and as is
evident, the point at which high-speed conveying commences is substantially lower than
the experimental results have indicated.
Table 7.4 Discharge angles for polyethylene pellets determined from

the trajectory models
Vb (m/s) C.E.M.A./M.H.E.A.* Booth* Golka*# Korzen* Goodyear*

0.50 75.1 52.1 78.0 49.3 80.8
0.75 54.7 40.6 62.2 38.3 68.9
1.00 tangency 20.0 34.0 23.0 50.1
1.25 tangency tangency tangency tangency tangency
* Discharge angle measured from the vertical in the direction of pulley rotation
#
Discharge angle for lower trajectory stream
7.5.3 DEM Conveyor Discharge Angles

The trajectory curves generated for the Chute MavenTM sensitivity analysis in Section
6.10.1 have been utilised to determine the required conveyor discharge angles.
Snapshots of the trajectory profile were taken at the conveyor discharge point for the
lower trajectory boundary and the discharge angles were measured graphically then
averaged to produce the results shown in Table 7.5.
138
Table 7.5 Discharge angles determined from the Chute MavenTM DEM simulations
Chute Maven
Vb (m/s) Discharge Angle*
0.50 65.4
0.75 61.0
1.00 39.5
1.25 4.1
1.50 0
1.75 tangency
in the direction of pulley rotation
7.5.3.2 EDEM
The discharge angle simulations produced with EDEM were produced after the
trajectory simulations discussed in Chapter 8. As such, the method used to produce the
simulations is not included in this section as it is detailed completely in Section 8.4.3. In
summary, EDEM has been used to produce a series of discharge angle simulations for
both spherical particles and particles for which particle shape has been considered. An
example of a simulation output is shown in Figure 7.9 where a region of analysis
(referred to as a bin) has been selected. This bin has been created along the
longitudinal axis of the conveyor and is 40 mm wide to provide a pseudo two-
dimensional representation of the particle stream to allow direct comparison with the
discharge models.
bin
Figure 7.9 Example of the discharge of polyethylene pellets using EDEM with
spherical particles
139
Subsequent analysis of the simulation data with MS Excel has resulted in the discharge
angles presented in Table 7.6. Comparing the results for the spherical and shaped
particles has shown that there is at most a 1 variation for a corresponding belt speed.
Table 7.6 Discharge angles determined from the EDEM simulations

for polyethylene pellets
EDEM EDEM
Discharge Angle* Discharge Angle*
Vb (m/s) (spherical) (shaped)
0.50 58 59
0.75 52 53
1.00 37 37
1.25 3 4
1.50 -2 -2
1.75 -4 -3.5
2.00 tangency tangency
7.5.4 Comparison of Conveyor Discharge Angles for Polyethylene

Pellets
Combining the four sets of data presented above results in the comparisons shown in
Figure 7.10.
Figure 7.10 Comparison of polyethylene pellet conveyor discharge angles from

experiments, trajectory models and DEM
140
There is a wide spread of results, especially when comparing the results of the discharge
angles obtained from the trajectory models. The experimentally measured discharge
angles follow a similar trend to the DEM discharge angles; however, there is some
deviation. The Chute MavenTM prediction of the discharge angles has an almost
identical result to the experimental discharge angle up to a belt speed of Vb = 1 m/s,
after which Chute MavenTM under predicts. From these comparisons, there does not
seem to be any model or simulation method which is capable of predicting the
experimental discharge angle for polyethylene pellets.
7.6 CONVEYOR DISCHARGE ANGLE FOR IRON ORE

Due to the dusty nature of the iron ore tested experimentally, a full range of belt speeds
was not possible. A belt speed of Vb = 2 m/s was experimentally tested and from video
footage of the material discharge, it was found that the material was exhibiting high-
speed conveying conditions, i.e. discharge at the point of tangency between the belt and
head pulley.
7.7 CONVEYOR DISCHARGE ANGLE FOR CORN

To provide further comparisons between the various methods of determining conveyor
trajectories, it was decided that a second product be trialled. Corn was selected due to its
larger size and also its non-spherical shape, which will be investigated in the DEM
simulations produced by EDEM.
7.7.1 Experimental Conveyor Discharge Angles

The video footage which was captured to determine the experimental discharge angles
for the polyethylene pellets was low resolution so the decision was made to use a digital
SLR camera to capture the discharge angles for the corn experiments. The photographs
were analysed using the same method as for the video footage and the results are shown
in Table 7.7. The results showed that at a belt speed of 2.00 m/s, the material is
discharging from the point of tangency.
141
Table 7.7 Experimentally determined discharge angles for corn
Belt speed (m/s) Discharge angle*

0.50 65.6
0.75 54.4
1.00 35.4
1.25 17.3
1.50 7.5
1.75 0
2.00 tangency
7.7.2 Analytically Determined Conveyor Discharge Angles

The analytical trajectory models have again been used to determine the conveyor
discharge angles for corn. The results are presented in Table 7.8 and the point at which
high-speed conveying commences is identical to that found for the polyethylene pellets.
Table 7.8 Discharge angles for corn determined from the trajectory models
Vb (m/s) C.E.M.A./M.H.E.A.* Booth* Golka*# Korzen* Goodyear*

0.50 75.1 53.6 78.0 50.4 80.8
0.75 54.7 41.7 62.2 38.7 68.9
1.00 tangency 20.4 34.0 22.9 50.1
1.25 tangency tangency tangency tangency tangency
#
Discharge angle for lower trajectory stream
7.7.3 DEM Conveyor Discharge Angles

Due to the limitations in producing high mass flow rate conveyor trajectories with
Chute MavenTM, the decision was made not to generate any DEM simulations for corn
with this software.
7.7.3.2 EDEM
EDEM has been used to produce a series of discharge angle simulations for corn for
both spherical particles and 6-sphere shaped particles, designed to approximate the true
shape of a corn kernel. The method used is detailed in Section 8.6.3. A bin has again
been used to extract only data from the region of interest and has been analysed using
142
MS Excel to produce the results shown in Table 7.9. The comparison between spherical
particles and those considering shape have a very close agreement for corresponding
belt speeds. An additional simulation was completed using a 12-sphere representation of
a corn kernel to determine if the accuracy of the represented particle had any influence
on the results. A single discharge angle simulation was performed for a belt speed of 1
m/s and on analysis, it was found that the discharge angle was identical to that obtained
using the 6-sphere approximation. From this result it was decided that there would be no
tangible gain in using the 12-sphere shaped particle and in fact would increase the
overall simulation times required to obtain what could be done with a simpler
representation.
Table 7.9 Discharge angles determined from the EDEM simulations for corn
EDEM Discharge EDEM Discharge

Angle* Angle*
Vb (m/s) (spherical) (shape)
0.50 62 63
0.75 53 56
1.00 40 39.5
1.25 5 4
1.50 -1 1
1.75 -3 -2.5
2.00 tangency tangency
7.7.4 Comparison of Conveyor Discharge Angles for Corn

Combining the three sets of data presented above yields the comparisons shown in
Figure 7.11. The experimental results compare quite well to the DEM results at low belt
speeds but then deviate away. There are no analytical models which come close to
representing the experimental results. There is a wide spread of predictions within the
various analytical models, with no two models predicting the same trend. The results for
the spherical and shaped particle DEM simulations provide a nearly identical result.
This comparison seems to indicate that there is no method capable of predicting the
experimentally determined discharge angles for corn.
143
Figure 7.11 Comparison of corn conveyor discharge angles from experiments,

trajectory models and DEM
7.8 DISCUSSION
An experimental investigation of conveyor discharge angles has been completed for
polyethylene pellets and corn. The results have shown that for both materials, the belt
speed at which high-speed conveying commences is between 1.75 and 2.00 m/s. The
profile of the material discharge shown in Figure 7.8 is indicative of the method used
for all test materials and belt speeds and minimises the effect of parallax error. In the
next chapter there is a discussion in Section 8.4.1.3 regarding the curving of the
trajectory stream after discharge, however, on close inspection of the underside of the
trajectory stream in Figure 8.19, it can be clearly seen that the profile is flat at the point
at which the angle is measured. This indicates that any curving of the trajectory stream
(which occurs at varying distances from the point of discharge based on belt speed), has
not influenced or affected the measurement of the experimental discharge angle.
Seven different discharge angle methods have been presented utilising a wide range of
formulae and in the case of Dunlop (1982), a graphical approach has been used. With
such a range of formulae it is inevitable that there are differences between the discharge
angles produced by each method, which has clearly been shown, yet they cannot all be
correct.
It was found that when comparing horizontal and inclined conveyor geometries there is
no difference to the discharge angle until nearing the critical belt speed at which point,
144
as the belt inclination angle increases, the belt speed reduces slightly at the transition
from low-speed to high-speed conditions.
The methods presented by C.E.M.A. (1966; 1979; 1994; 1997; 2005) and M.H.E.A.
(1986) are the same, varying only in the fact that one is based on imperial units while
the other used metric units respectively and also some minor adjustments to the
tabulated data in different editions of the C.E.M.A. guide. The Korzen (1989) method
incorporates adhesive stress into its determination of the discharge angle unlike any of
the other methods.
Discrete element modelling simulations have also been generated to investigate how
accurately the discharge angle of materials can be predicted. For polyethylene pellets
both Chute MavenTM and EDEM have been used and for corn only EDEM was used.
When considering the discharge angles determined from the three methods for each
material, there is no discharge model which accurately predicts the experimental results.
This is also true of the discrete element modelling simulations, although there is some
agreement between Chute MavenTM and the experimental discharge angles for
polyethylene at low belts speeds before diverging and also agreement between EDEM
and the experimental corn discharge angles at low belt speeds before diverging.
The fact that there is no clear method of predicting the conveyor discharge angle has
flow on effects when considering the path of the resulting conveyor trajectories, which
will be discussed in Chapter 8, for materials and their subsequent interaction with a
conveyor transfer hood.
145
CHAPTER 8 CONVEYOR TRAJECTORIES
Chapter
8
CONVEYOR TRAJECTORIES
146
8.1 INTRODUCTION
The accurate determination of the material trajectory from a conveyor head pulley is
extremely important in the design of conveyor transfers. Incorrect trajectory path
predictions can result in several detrimental issues arising within a conveyor transfer,
including; particle attrition, chute wear, dust generation, spillage, chute blockage and
excessive noise, all due to incorrect design parameters.
There is much information presented in the literature discussing the design of the hood
(or upper) section of a transfer chute. Of critical importance is the material stream flow
into or onto these sections. For one, the angle of incidence should be kept sufficiently
low to minimise any reduction in velocity of the material as it flows through the chute.
Without an accurate prediction of a trajectory path there is no way to adequately
determine this angle of incidence. The profile of the trajectory stream is also of
importance in the design of a conveyor transfer, where one trajectory method might
result in a converging stream while another might produce a diverging stream or one of
constant depth.
The determination of the discharge angle, detailed in Chapter 7, at which the material
stream leaves the head of a conveyor, dictates the trajectory the particle stream takes.
This chapter will first investigate and review various aspects of the trajectory methods
available in the literature, detailed in Section 2.4. These methods vary considerable with
respect to the number of individual parameters being used in the determination of the
material trajectory, from the very basic to complex iterative approaches. The
investigation includes altering values of the parameters exhibiting variability,
determining the extent to which they influence the trajectory being generated. The
profile of the material stream through the trajectory path also is discussed. These
trajectory methods are evaluated for both horizontal and inclined conveyors for a range
of belt speeds and pulley diameters to evaluate both low-speed and high-speed
conveying conditions.
The second part of this chapter will focus on the methods used to determine the
experimental trajectories on the conveyor transfer research facility. Predicted
trajectories from the various analytical methods and the trajectories produced from
DEM simulations will be generated based on the experimental geometry. This section
147
will also provide comparisons between each approach. Part of the sensitivity analysis
detailed in Section 6.2.7 has been based on the DEM output for trajectories generated
with Chute MavenTM. This work showed that there were limitations with respect to the
number of particles which could be simulated successfully and as a result, it was not
possible to produce simulations with the high mass flow rates required. For this reason,
Chute MavenTM, was not used in this section, instead EDEM was utilised, as it could
simulate the required mass flow rates.
8.2 CONVEYOR TRAJECTORY MODEL COMPARISONS

The clearest way to compare one trajectory method against another is to use a set of
arbitrary parameters, as shown in Table 8.1. These parameters are representative of
conveyor belt designs and also material characteristics, taken from several sources
(Korzen, 1989; Golka, 1993; C.E.M.A., 2005). This data has no link to the conveyor
transfer research facility or the test materials. There are many parameters which could
be altered to produce a wide range of comparisons, however, as this would prove to be
highly exhaustive, the majority of parameters have been set as constant with variations
only made to belt speed, pulley diameter and belt inclination angle.
Table 8.1 Conveyor parameters used for comparisons
PARAMETERS VALUES
Belt Width, wb 0.762 m
Belt Thickness, bt 0.0111125 m
Belt Speed, Vb 1.25, 3.00 and 6.00 m/s
Pulley Diameter, Dp 0.5, 1.0 and 1.5 m
Surcharge Angle 20
Trough Angle 20
Belt Inclination Angle, b 0 and 10
Divergent Coefficient, Lower, 1 0.1
Divergent Coefficient, Upper, 2 -0.1
Coefficient of Friction, Static, s 0.5
Coefficient of Friction, Kinetic, k 0.42
Adhesion Stress, a 0 kPa
Equivalent Spherical Particle Diameter 0.001 m
Atmospheric Temperature, Tatm 20 C
Atmospheric Pressure, Patm 101 kPa
Air Viscosity, f 1.80E-05 Ns/m2
Product Bulk Density, b 2000 kg/m3
148
8.2.1 Low-Speed Conveyor Trajectory Comparisons

For a low-speed belt condition (i.e. Vb = 1.25 m/s), the material being conveyed will
discharge after wrapping around the head pulley to a calculated or graphically
determined discharge angle, d. It was found that as the pulley diameter increases, an
increasing number of methods converge to a similar discharge angle. This results in the
number of unique trajectory curves being reduced as the pulley diameter increases.
Where trajectory curves for different models vary by no more than 2% for both the
lower and upper trajectories, they have been grouped together for clarity.
In the case shown in Figure 8.1 and Figure 8.2, there were nine unique trajectory curves
produced from the methods and as such the representation of the lower and upper paths
were separated onto two graphs to allow clear comparisons to be made.
Out off interest it was found that for a belt inclination angle of either b = 0
(horizontal) or b = 10, there is no variation to the discharge profile for the
corresponding pulley diameters, which can be seen in Figure 8.3 and Figure 8.4. This is
due to the fact that for the low-speed condition, the material wraps around the head
pulley to a determined angle of discharge regardless of whether the material travels
horizontally or at an inclined angle beforehand. Also evident is the grouping of some
trajectory curves reducing the number of unique curves to six and as a result the lower
and upper paths can be displayed on the one graph.
8.2.2 High-Speed Conveyor Trajectory Comparisons

In order to investigate the possible changes in trajectory profile due to belt speed, two
high-speed belt conditions were selected, Vb = 3.0 m/s and Vb = 6.0 m/s. A comparison
was made between the discharge velocities and pulley diameters and is presented in
Table 8.2 (only for Vb = 3.0 m/s). As can be seen from all the values for the lower
trajectory limit, the discharge velocity is equal to the belt speed, the variations lie with
the upper trajectory discharge velocities. The slight variations with the C.E.M.A. (1966;
1979; 1994; 1997; 2005) and M.H.E.A. (1986) methods result from the variation in
material height as the upper discharge velocity is calculated based on a ratio of
velocities and radii.
149
Figure 8.1 Low-speed, horizontal Figure 8.2 Low-speed, horizontal

conveyor, lower path, conveyor, upper path,
pulley diameter, Dp = 0.5 m, pulley diameter, Dp = 0.5 m,
belt velocity, Vb = 1.25 m/s belt velocity, Vb = 1.25 m/s
Figure 8.3 Low-speed, Figure 8.4 Low-speed,

horizontal conveyor, inclined conveyor,
pulley diameter, Dp = 1.0 m, belt inclination angle, b = 10
belt velocity, Vb = 1.25 m/s pulley diameter, Dp = 1.0 m,
belt velocity, Vb = 1.25 m/s
The upper discharge velocity for the Booth (1934) method is identical to the C.E.M.A.
(2005) due to the material height from C.E.M.A. (2005) being assumed. For both belt
speeds selected for this comparison, the Golka (1992; 1993) method uses case 2,
150
explained previously, and as such the discharge velocity for both the lower and upper
trajectories is equivalent to the belt speed. Korzen (1989) by default uses the belt speed
as the discharge velocity for the lower, central and upper trajectories as previously
stated. The methods of Dunlop (1982) and Goodyear (1975) do not calculate discharge
velocity and as such have been omitted from Table 8.2.
Table 8.2 Discharge velocities versus pulley diameter for a belt speed of Vb = 3.0 m/s
DISCHARGE VELOCITY, Vd, (m/s)

TRAJECTORY
Dp = 0.5 m Dp = 1.0 m Dp = 1.5 m
METHOD
C.E.M.A. 1, 2, 4, 5 L: 3.000 U: 4.109 L: 3.000 U: 3.566 L: 3.000 U: 3.380
C.E.M.A. 6 L: 3.000 U: 4.182 L: 3.000 U: 3.604 L: 3.000 U: 3.405
M.H.E.A. 1986 L: 3.000 U: 4.103 L: 3.000 U: 3.563 L: 3.000 U: 3.378
KORZEN L: 3.000 U: 3.000 L: 3.000 U: 3.000 L: 3.000 U: 3.000
BOOTH L: 3.000 U: 4.182 L: 3.000 U: 3.604 L: 3.000 U: 3.405
GOLKA L: 3.000 U: 3.000 L: 3.000 U: 3.000 L: 3.000 U: 3.000
When comparing the high-speed trajectories produced for horizontal belts and inclined
belts, although somewhat small, there is a difference, see Figure 8.5 and Figure 8.6.
Figure 8.5 High-speed, Figure 8.6 High-speed,

horizontal conveyor, inclined conveyor,
pulley diameter, Dp = 1.0 m, belt inclination angle, b = 10
belt velocity, Vb = 3.00 m/s pulley diameter, Dp = 1.0 m,
belt velocity, Vb = 3.00 m/s
151
This is due to the trajectories for the high-speed conditions commencing at the tangent
point of the belt and the head pulley and clearly if the belt has an initial inclination then
the trajectory will start earlier than for the horizontal equivalent and will result in the
trajectory for the inclined condition projecting slightly further than for a horizontal
geometry.
In Figure 8.5 and Figure 8.6 the inclusion of air drag by Korzen (1989) results in a
trajectory prediction markedly lower than for all other methods. If air drag is neglected
in the Korzen (1989) method, the resulting trajectory prediction is located amongst the
other methods.
Previously it was explained that Golka (1992; 1993) uses divergent coefficients to
obtain a better approximation of the lower and upper trajectory paths. Golka et al.
(2007) provided detail of how these divergent coefficients are determined and can be
used to more accurately predict the conveyor trajectories. If the divergent coefficients
are neglected in the calculations, the resulting trajectories are identical to the Korzen
(1989) method when air drag is neglected as the equations are identical.
For both high-speed cases examined, the early C.E.M.A. (1966; 1979; 1994; 1997) and
M.H.E.A. (1986) methods generate the highest trajectory curve for all pulley diameters
and as the discharge velocity increases, the variation from the other curves becomes
more defined.
Figure 8.7 represents another of the fast-speed conditions (i.e. Vb = 6.00 m/s) and as is
evident, there is another reduction in the number of unique trajectory curves, now only
having four. The curves generated from the Korzen (1989) method incorporating air-
drag are now dramatically diverging from the other curves.
This would seem to warrant further investigation because as it stands, it would be hard
to justify applying this method when there are eight other methods which do not
produce a curve anywhere close to it.
152

pulley diameter, Dp = 1.5 m, belt velocity, Vb = 6.00 m/s
8.3 INFLUENCES ON CONVEYOR TRAJECTORY PROFILES
8.3.1 Effect of Belt Inclination Angle

When varying the belt inclination angle in the comparisons above, initially there was no
visible difference to the trajectories from a horizontal belt to those of an inclined belt for
low-speed conditions. There are five unique equations across the trajectory methods
above to determine the critical angle, as explained in Section 7.2. The same findings
that are presented in Section 7.2 also apply for conveyor trajectories and will not be
repeated. For high-speed conveying conditions there will be differences between the
trajectories generated for horizontal belts versus inclined belts in that the trajectories
will start before the vertical most point of the head pulley for inclined belts. This will
mean that for inclined belts, the trajectory will have an initial vertical component of
velocity.
8.3.2 Effect of Static and Kinetic Friction

To evaluate the effect of varying the coefficient of static friction on the results of the
Booth method (1934), nine values between 0.1 and 0.9 were selected and using the
parameters of Table 8.1, nine profiles have been determined for both the lower and
upper trajectories and displayed for low-speed, Figure 8.8, and high-speed conditions,
Figure 8.9. For the Korzen method (1989), both, s and k are used. To investigate the
153
effects of varying s and k on the resultant trajectories from the Korzen method, the
combinations listed in Table 8.3 have been used, keeping in mind the general
understanding that k < s. This results in 36 individual profiles for both the lower and
upper trajectories as shown in Figure 8.10 and Figure 8.11.
Dimensions in millimetres
Figure 8.8 Booth Vb = 1.5 m/s Figure 8.9 Booth Vb = 3.0 m/s
Dimensions in millimetres
Figure 8.10 Korzen Vb = 1.5 m/s Figure 8.11 Korzen Vb = 3.0 m/s
154
Table 8.3 Combinations of coefficient of static and kinetic friction

used for the Korzen method (1989)
s
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0.1 9 9 9 9 9 9 9 9
0.2 9 9 9 9 9 9 9
0.3 9 9 9 9 9 9
0.4 9 9 9 9 9
k 0.5 9 9 9 9
0.6 9 9 9
0.7 9 9
0.8 9
0.9
For the low-speed conditions of both methods, the profiles generated form a band that
increases as the vertical displacement increases. At a vertical displacement of 3.5m, the
range is given in Table 8.4. For high-speed conditions only one unique trajectory curve
was produced for both methods, compare Figure 8.9 with Figure 8.11. This is due to the
coefficient of static and kinetic friction only being used in the determination of low-
speed discharge angles.
Table 8.4 Range of trajectory profiles for low-speed conditions by Booth (1934) and
Korzen (1989)
Trajectory Booth Korzen

Lower 1280 to 1470 mm 1255 to 1345 mm
Upper 1385 to 1575 mm 1445 to 1545 mm
8.3.3 Effect of Divergent Coefficients

As previously mentioned, Golka (1992; 1993) uses divergent coefficients to aid in the
prediction of conveyor transfers, but without explanation of how they are quantified.
More recently Golka et al. (2007) supplied a table explaining the range of divergent
coefficients to use under various conditions. The maximum divergent coefficient
recommended is 0.2. A range of divergent coefficient values ranging from 0 to 0.2
have been applied to the parameters of Table 8.1 using a belt speed on 3 m/s to produce
155
upper and lower trajectory streams, the results shown in Figure 8.12. It is evident that
that this one parameter alone has a dramatic affect on the resulting trajectories.
Figure 8.12 Variation in trajectories based on different divergent coefficients
8.3.4 Effect of Particle Shape and Size

Again, the parameters of Table 8.1 have been applied and only the equivalent spherical
particle diameter, dk, has been varied. As an illustration of this, three values of dk have
been selected, see Table 8.5. The resulting trajectory curves are displayed in Figure 8.13
and it can be seen that a change in the equivalent spherical particle diameter does have
an effect. Each figure includes the no air drag condition for comparison. Figure 8.13a
plots the trajectory curves for the smallest particle size and shows the largest variations.
Figure 8.13c plots the trajectory curves for the largest particle size (i.e. 1 g) and shows a
near identical path to the no air drag trajectory. This corresponds with the claim made
by Korzen (1989) that air drag can be neglected for particles over 1 g in mass.
Table 8.5 Selected equivalent spherical particle diameters for comparison
dk Unit mass of
(mm) particle (g)
0.5 0.0001
1 0.001
10 1.05
156
If a particle size distribution (PSD) was present for a given conveying condition, a
combination of the results obtained in Figure 8.13a, Figure 8.13b and Figure 8.13c
would result. Due to the likelihood of some degree of segregation occurring as a result
of this PSD, there would be a tendency for the fines to behave as shown in Figure 8.13a
while the coarser material would behave more like that shown in Figure 8.13c.
(a) dk = 0.5 mm, (b) dk = 1.0 mm, (c) dk = 10 mm,

dm = 0.0001 g dm = 0.001 g dm = 1.05 g
Figure 8.13 Effect of particle size distribution on Korzen (1989) method
8.3.5 Effect of Adhesive Stress

In Section 7.3.3, the effect of adhesive stress on discharge angle was presented. The
resulting discussion also applies for the trajectory profile and will not be repeated here.
8.3.6 Effect of Bulk Density

The Korzen method (1989) is the only one to incorporate material bulk density into the
equations. Bulk density is required to determine the bulk specific gravity of the material
for the adhesive component, the discharge velocities of the lower and upper trajectory
streams and the unit particle mass. Using the parameters of Table 8.1, for a belt speed of
Vb = 3 m/s and only varying the bulk density, it was found that as bulk density is
reduced, the trajectory profile becomes shallower and by increasing it, the trajectory
profile converges on the profile generated when air drag is neglected, as shown in
Figure 8.14. It should also be noted that for the no air drag condition, the bulk density is
157
not used in the determination of the trajectory path, therefore no bulk density has been
defined for this condition on Figure 8.14.
Figure 8.14 Effect of bulk density on trajectory profile, Vb = 3 m/s
8.4 CONVEYOR TRAJECTORIES OF POLYETHYLENE PELLETS
8.4.1 Experimental Conveyor Trajectories

To allow conveyor trajectories to be investigated, the conveyor transfer research facility
was configured to allow uninterrupted flow of material from the head of the feed
conveyor onto the receiving conveyor below by removing the hood and spoon.
8.4.1.1 Preliminary Setup

The trajectory was photographed while under steady-state conditions and videoed using
a standard 25 frame per second digital video camera. Belt speeds were tested between
0.5 m/s and 2.25 m/s at 0.25 m/s increments. This upper belt speed limit was due to the
trajectory stream falling in close proximity to the end acrylic cover containing the
material. An example of the captured trajectory stream can be seen in Figure 8.15.
158
Figure 8.15 Trajectory, Vb = 1.5 m/s, ms = 24 tph
Using known distances on the framework of the conveyor transfer research facility;
measurements were to be taken to produce the trajectory profile of the lower and upper
streams. In reality this proved to be difficult, as a degree of parallax error was evident in
the image captured. This parallax error creates inaccuracy in the measured trajectory
profiles and therefore it was decided this series of tests would be used for demonstration
purposes. More thought was required to produce a more accurate method of extracting
the trajectory profile data.
8.4.1.2 Laser Scanning

Another method of profiling the conveyor trajectory was trialled, with equipment from
Bluescope Research being tested by an undergraduate Mechanical Engineering thesis
student (Andrews, 2008). An optical laser connected to an X-Y frame was positioned
above the head pulley of the discharge conveyor. The laser moved via stepper motors,
which controlled linear slides and was connected to a laptop via a data acquisition card
to record the signals. The size of the X-Y frame and linear slides meant that this
arrangement was not ideal for obtaining the trajectory profile of the upper surface as not
enough of the trajectory could be recorded. The results are shown in Figure 8.16.
The profiling of the lower trajectory stream could not be approached in the same way,
instead of using the X-Y frame, the laser was fixed to an existing cross-brace and a
159
stepper motor was used to rotate the laser to scan the trajectory profile. The laser being
used had an operating focal length of 0.5 to 6 m and as such had its own limitations. If
the trajectory stream fell too close to the laser, i.e. within 0.5 m, the laser could not
detect the profile.
The decision was made that the laser scanning method was not feasible with the lasers
available and as such was disregarded as a suitable method of determining the lower and
upper trajectory profiles.
Figure 8.16 Laser scanned upper trajectory profile (Andrews, 2008)
8.4.1.3 Final Setup

An enhancement of the preliminary trajectory setup was produced, Section 8.4.1.1,
involving the addition of a 100 mm square grid placed behind the trajectory stream.
Also included in this phase of the testing was the addition of a trajectory hopper,
designed to manually slide along the receiving conveyor allowing capture of the
trajectory stream and smooth delivery of material onto the receiving conveyor. This
trajectory hopper allowed higher belt speeds to be tested, beyond the limiting 2.25 m/s
of the preliminary trajectory testing. All extraneous framework was removed to give the
most uninterrupted view of the trajectory possible and the final arrangement can be seen
in Figure 8.17.
The addition of the trajectory hopper allowed testing of belt speeds up to the maximum
achievable on the conveyor transfer research facility, resulting in tests being performed
using belt speeds between 1 m/s and 7 m/s in 1 m/s increments. Low mass flow rates
were tested to generate a thin particle trajectory stream to provide a lower trajectory
160
stream only and high mass flow rates were tested, with the edge distance set to
maximum for each belt speed tested to produce both lower and upper trajectory streams
to allow comparisons with the equivalent analytical methods and DEM simulations.
Table 8.6 summarises the range of experimental tests performed.
Figure 8.17 Final conveyor trajectory test arrangement
Table 8.6 Experimental trajectory test setups
Belt Speed Low Feed Rate High Feed Rate

(m/s) (tph) (tph)
1 2.6 19
2 2.6 31
3 2.6 37.8
4 2.6 37.8
5 2.6 37.8
6 2.6 37.8
7 2.6 37.8
The same mass flow rate was used for belts speeds of 3 m/s and higher due to the
HoganTM valve being open such that material was flood feeding onto the conveyor.
Each test performed was videoed, capturing the entire stream and each test was
photographed, but this time not capturing the overall trajectory, but small sequential
sections to minimise any potential parallax error, see Figure 8.18. A significant finding
resulting from the experimental testing is that the underside of the trajectory stream
161
does not stay flat. As product moves along the conveyor through the troughed section,
the material is forced into a curved geometry, however, once the transition zone is
reached, the profile of the material changes. Through the transition zone the underside
of the material profile changes from the profile of the trough to finally flat, when
material reaches the head pulley and discharges.
Figure 8.18 Example grid referencing, Vb = 2 m/s, ms = 2.6 tph
Figure 8.19 shows the flat nature of the underside of the material stream at the point of
discharge and the beginning of the curving to the underside of the trajectory stream.
This flattening of the material through the transition zone causes a degree of lateral
downward velocity to some of the material which continues after discharge, forming
what has been termed wings.
The formation of these wings is evident in the trajectory stream shown in Figure 8.20.
The material present in this region of the trajectory stream is not as densely packed as
the main body of the trajectory and as such the influence of air drag effects will be more
pronounced.
The individual photographed sections of the trajectory stream were then analysed and
compiled to produce overall trajectories. The results of the experimental trajectory
analyses are presented in Figure 8.21 and Figure 8.22.
162
Figure 8.19 Flat underside of the trajectory stream at the point of discharge
for a belt speed of Vb = 4 m/s and mass flow rate of ms = 37.8 tph
Figure 8.20 Trajectory wings for a belt speed of Vb = 4 m/s and

mass flow rate of ms = 37.8 tph
163
Figure 8.21 Experimental polyethylene pellet trajectories for low mass flow rates
Figure 8.22 Experimental polyethylene pellet trajectories for high mass flow rates
An observation made for the low and high feed rate results is that for the higher belt
speeds, there is an overlap of the trajectory streams. It was deduced that there were two
probable causes;
- the particles were reaching their terminal velocity at discharge while being
conveyed at high belt speeds, or
- the particles were undergoing a degree of slip on the conveyor belt before
discharge, thus causing a reduced particle discharge velocity.
164
Of the two most likely possibilities, it was decided that particle slip was more likely and
was investigated. The Redlake X3 MotionPro high-speed digital video camera was
positioned parallel to the particle stream to film the particles discharging from the head
pulley. Image Pro Plus was used to analyse the resulting video footage to determine the
average particle velocity for each belt speed. An additional step in this process was to
look at the particle velocity in the lower half of the stream versus the particle velocity in
the upper half of the stream. This was to determine if there was any inter-particle
motion within the bulk particle stream while being conveyed. An example of this
analysis is shown in Figure 8.23 with the yellow line indicating the approximate
midpoint of the stream height. The final results of this slip analysis are presented in
Figure 8.24.
Figure 8.23 Comparison of belt speed to material discharge velocity

(the yellow line represents the distinction between the lower and upper
halves of the particle stream)
The measured particle velocity for both the lower half and upper half are shown with
respect to the corresponding belt speed and as can be seen, there is good agreement up
to a belt speed of 5 m/s, but there is a dramatic drop off of particle velocity once a belt
speed of 6 m/s and beyond is reached. For the higher belt speeds, it can also be seen that
the upper half of the particle stream has a slightly lower particle velocity than the lower
half equivalent. These findings would seem to account for the observations seen in
Figure 8.21 and Figure 8.22 with respect to the overlapping of the trajectory stream
profiles at high belt speeds.
165
Figure 8.24 Comparison of belt speed to material discharge velocity
Due to this reduced discharge velocity for a belt speed of 6 m/s and 7 m/s, comparisons
with the trajectory models and the EDEM simulations will be limited to a maximum
belt speed of 5 m/s.
8.4.2 Analytically Determined Conveyor Trajectories

Considering the information provided in Figure 8.24, the decision was made to only
produce analytical based trajectories up to and including a belt speed of 5 m/s. Applying
the parameters for the experimental geometry as well as the particle characteristics for
polyethylene pellets, the analytical methods explained in Section 2.4 have been used.
Some minor adjustments have been made to these methods by using the material height
at discharge, h, and centroid height, a1, as supplied by C.E.M.A. (1966; 1979; 1994;
1997; 2005) and M.H.E.A. (1986) methods and which have been measured directly by
experiment. The generated conveyor profiles for the numerous methods and belt speeds
are presented in Figure 8.25.
On review of the trajectories generated for each belt speed investigated, the following
observations have been made;
- for a belt speed of Vb = 1 m/s, low-speed conveying conditions apply and each of
the trajectory methods generates a separate trajectory profile for the lower and
upper boundaries,
166
- the Golka method with and without the divergent coefficients used, produce
nearly identical profiles,
- the C.E.M.A. (1966; 1979; 1994; 1997) and M.H.E.A. (1986) methods produce
identical profiles for each of the belt speeds investigated,
- for a belt speed of Vb = 2 m/s, high-speed conveying conditions apply,
- for a belt speed of Vb = 2 m/s, the C.E.M.A. (2005) and Goodyear (1975) methods
produce identical profiles,
- for a belt speed of Vb = 2 m/s, the Golka (2007) method without divergent
coefficients and the Korzen (1989) method without air drag produce identical
profiles,
- for a belt speed of Vb = 2 m/s, the C.E.M.A. (1966; 1979; 1994; 1997) and
M.H.E.A. (1986) methods clearly produce the largest trajectory and continue to
do so for the higher belt speeds also,
- For all belt speeds exhibiting high-speed conditions, the Golka (2007) method
without divergent coefficients falls symmetrically inside the C.E.M.A. (2005)
method, while the Golka method with divergent coefficients falls symmetrically
outside the C.E.M.A. method,
- as belt speed increases, there is a noticeable merging of several trajectory
methods,
- for a belt speed of Vb = 3 m/s, the same trajectory method groupings exist as for
the Vb = 2 m/s case,
- for a belt speed of Vb = 3 m/s, the Korzen (1989) method incorporating air drag is
beginning to diverge from the other trajectory methods and is falling closer to the
conveyor head pulley,
- for belt speeds of Vb = 4 m/s and Vb = 5 m/s, the same trajectory method
groupings apply and exhibit the same trends for both,
- the Korzen (1989) method incorporating air drag is now more noticeably falling
closer to the conveyor head pulley than any of the other methods.
167
Figure 8.25a Analytically determined conveyor trajectories for Vb = 1 m/s
168
Figure 8.25b Analytically determined conveyor trajectories for Vb = 2 m/s

Figure 8.25c Analytically determined conveyor trajectories for Vb = 3 m/s
169
Figure 8.25d Analytically determined conveyor trajectories for Vb = 4 m/s

Figure 8.25e Analytically determined conveyor trajectories for Vb = 5 m/s
8.4.3 DEM Conveyor Trajectories

The limitations of Chute MavenTM have already been discussed in Section 6.2.7.2 and
the decision was made to use EDEM to produce the trajectory simulations due to its
170
ability to generate the required high mass flow rates to provide the widest range of
comparisons.
8.4.3.1 Scope of Simulations

EDEM has been utilised to produce trajectory simulations for low and high mass flow
rates matching those which were tested experimentally, see Table 8.6, but only up to
and including a belt speed of 5 m/s for the reasons specified in Section 8.4.1.3.
8.4.3.2 Particle Geometry

The effect of particle shape was investigated by producing two sets of simulations, one
set using spherical particles and the other set using shaped particles. Spherical particles
have been created with the EDEM particle factory, see Figure 8.26a, based on
equivalent volume diameter and applying a normal size distribution accounting for
97.1% of the experimentally measured size range (i.e. d = 3.35 mm to d = 4.75 mm).
Shaped particles, consisting of two spherical particles overlapped, see Figure 8.26, have
also been created based on the mean diameter and mean length of the polyethylene
pellets and the same normal size distribution as for spherical particles has been applied.
When the particles are created within the EDEM particle factory, the volume and mass
of the generated particles are calculated from the inputted particle density.
(a) Spherical (b) 2-particle shaped
Figure 8.26 Particle representations of polyethylene pellets used in EDEM
8.4.3.3 Calibration of the Mass flow Rate

Input of the mass flow rate for a given simulation is not as straight forward as entering
the tonnes per hour of material required. EDEM required the user to enter the total
171
number of particles to be generated during the entire simulation and also the number of
particles to be generated per second. To do this accurately, a mass flow rate calibration
needs to be completed. Six simulations were performed where particles were generated
for 1 second for both the spherical and shaped particles, as summarised in Table 8.7.
Using the graphing options within EDEM, the total mass of particles for each
simulation was produced. Figure 8.27 shows the box which was generated, having an
open top, allowing particles to fill the bin via the injection plane positioned centrally
above. The mass flow rate was then determined by plotting the data with MS Excel and
curve fitting a function to both sets of data to produce equations for the relationship
between the number of particles and mass flow rate, as shown in Figure 8.28. These
equations could then predict the number of particles required to produce mass flow rates
comparable to those used experimentally, see Table 8.8.
Table 8.7 Mass flow rate calibration simulation of polyethylene pellets in EDEM
Spherical Shaped
Number of Generation Mass of Mass flow Mass of Mass flow
particles time particles rate particles rate
(sec) (kg) (tph) (kg) (tph)
1000 1 0.041 0.146 0.035 0.126
2500 1 0.102 0.366 0.087 0.314
10000 1 0.408 1.467 0.349 1.257
20000 1 0.814 2.929 0.698 2.511
50000 1 2.035 7.327 1.733 6.240
100000 1 4.065 14.633 3.400 12.239
Injection plane
Domain
Box
Particles
Figure 8.27 Mass flow rate calibration with EDEM simulating 100,000 particles
172
Figure 8.28 Calibration curves for mass flow rate of polyethylene pellets
Table 8.8 Required number of polyethylene pellets to achieve the experimental

mass flow rates
Mass flow rate Number of spherical Number of shaped

(tph) particles required particles required
2.6 17752 20941
19 129818 154826
31 211818 252791
37.8 258284 308304
8.4.3.4 Conveyor Geometry

The required three dimensional CAD model imported into EDEM can consist of several
individual component models, giving the flexibility of quick alterations when required.
A three dimensional CAD model of the feed conveyor was created and imported into
EDEM. The conveyor is inclined at 7.5 to represent the angle of the conveyor through
the transition zone, which is the effective angle at which the material will discharge
from the conveyor under high-speed conditions. Separate to the conveyor model, a three
dimensional CAD model of the experimental feeding chute has also been created and
three instances of this chute have been imported into the same EDEM simulation at
varying distances along the feed conveyor. In reality, the feeding chute is 3.5 m from
the conveyor head pulley axis, however, the feeding chute model was installed 1.5 m
back from the head pulley for a belt speed of 1 m/s, 2.5 m back for a belt speed of 2 m/s
and 4 m back for belt speeds of 3 m/s, 4 m/s and 5 m/s, see Figure 8.29 for the final
conveyor geometry. The rationale behind having multiple feed chutes is to reduce the
173
simulation time for the lower mass flow rates, as steady state conveying would be
achieved in a shorter conveyor distance. Only one feed chute is used in a given
simulation, a setting within the setup allows for individual elements to be made virtual,
meaning although they are still visible, they are no longer solid entities and particles can
flow through them without interference.
Figure 8.29 Conveyor geometry imported into EDEM
8.4.3.5 Particle Parameters

In the setup of the simulation parameters, there are a number of particle parameters
required which have not been determined as part of the characterisation detailed in
Chapter 5. The particle density of the belt has been approximated based on a sample of
the belt installed on the conveyor transfer research facility as it comprises a soft PVC
top layer as well as a fabric ply beneath. Section 2.14.8 provided a discussion on the
quantification of the coefficient of rolling friction and has been defined as 1% of the
static friction, as kinetic friction is not a parameter used by EDEM. The coefficient of
particle-particle friction is listed in Table 5.1 and Section 5.1.10 details its method of
determination.
The modelled belt is inclined at 7.5 from the horizontal to account for the angle of the
transition region, requiring the velocity of the belt to be broken into X and Y
components and the breakdown is presented in Table 8.9 and is used for all the
spherical and shaped particle simulations.
174
Table 8.9 Belt speed settings for all EDEM simulations
Belt Speed Belt Speed X Belt Speed Y

(m/s) (m/s) (m/s)
1 0.991 0.131
2 1.983 0.261
3 2.974 0.392
4 3.966 0.522
5 4.957 0.653
8.4.3.6 Low Mass Flow Rate EDEM Trajectory Simulations

DEM simulations were completed for both spherical and shaped particles,
corresponding to the experimental tests detailed in Table 8.6. On completion of the
simulations, the data was extracted from several regions to obtain specific displacement
and velocity data. These regions are selected by the user and are known as bins and
isolate three dimensional regions within the simulation domain. Figure 8.30 shows the
four bins used for data extraction.
Bin 1 is used on the low mass flow rate simulations and encompasses the entire
trajectory stream from 200 mm before trajectory discharge. Bin 2 selects a 30 mm wide
slice of particles along the longitudinal axis of the conveyor to represent the two
dimensional behaviour of the trajectory. This data will allow direct comparison with the
trajectory models. Bin 3 takes a 40 mm square vertical cut at the top most point of the
head pulley to allow extraction of the particle discharge velocity.
Bin 3
Bin 3
Bin 2 Bin 2
Bin 1 Bin 4
(a) (b)
Figure 8.30 Bins used for data extraction for (a) low mass flow rate simulations and
(b) high mass flow rate simulations
175
For the low and high mass flow rate cases, bin 2 and bin 3 are identical. Bin 4 is used on
the high mass flow rate simulations and is half the width of bin 1, to capture one half of
the trajectory stream in the longitudinal direction of the conveyor. When the particle
data is extracted to MS Excel, there is a limitation of 65536 particles due to this being
the maximum number of rows allowed within a single spreadsheet. For the low mass
flow rate simulations, there are a relatively small number of particles generated in the
simulation and therefore the entire trajectory stream can be captured, however, for the
high mass flow rate simulations, there are more particles generated than there are rows
in the spreadsheets so only one half of the trajectory stream is extracted, on the
assumption that the trajectory stream is symmetrical.
The belt speeds from Table 8.9 were used to generate ten simulations, five each for
spherical and shaped particles, each having a mass flow rate of 2.6 tph. The data from
bin 3 was plotted first to determine if there is any slip between the belt and particles, the
results are presented in Figure 8.31 and Figure 8.32. The horizontal distance 0 mm,
corresponds to the longitudinal axis of the conveyor and velocity data for all particles
20 mm either side has been plotted. Linear trend lines were fitted to each set of data
along with each equation, the y intercept being the approximate mean particle velocity
for each case. It can be seen that as the belt speed increases, the velocity of the particles
at the point of discharge begins to drop. This is more evident for the spherical particle
simulations, which is to be expected as the spherical particles are freer flowing than the
shaped particles. The 1 m/s belt speed case for both spherical and shaped particles are
the only two sets of data which can be said to be accurate for further use.
This discrepancy between particle and belt velocity means that the resulting trajectories
for these simulations are not truly representative of the belt speed. To achieve a higher
particle discharge velocity, the belt speed needs to be increased. Using equation 8.1,
new belt speeds were predicted to generate a particle discharge velocity to match the
experimental belt speeds. Tables 8.10 and 8.11 provide a summary of the original
simulation particle discharge velocities as well as the additional simulations performed
once equation 8.1 was applied.
Vb 2
Vb ,new = 8.1
Vdisch arg e
176
Figure 8.31 Graph of discharge velocities versus width of belt for spherical particles
Figure 8.32 Graph of discharge velocities versus width of belt for shaped particles
As an example, the second row of Table 8.10 will be explained. Firstly the belt speed
for the DEM simulation was defined, in this case Vb = 2 m/s. The conveyor belt was
assumed to be at an inclination angle of 7.5, based on the transition zone, and as such,
the x and y components of the belt velocity were determined via trigonometry. After the
simulation ended, the magnitude of the particle discharge velocity was determined and
compared directly to the initial belt speed and noted as a percentage. If the absolute
percentage difference was within 2%, the simulation was deemed adequate and the data
obtained was accepted. If the absolute percentage error was greater than 2%, then a new
adjusted belt speed was determined with the assumption that the trend was linear.
This new belt speed was used to create a further DEM simulation which is shown on the
third line of Table 8.10. The original particle velocity at discharge had a 9.13% shortfall
and after the adjustment had a substantially smaller error of 1.01%.
177
Table 8.10 Belt speeds used to generate the correct particle discharge velocities
for spherical particle simulations for the low mass flow rate
Belt Velocity Velocity Velocity Factor Adjusted

Speed X Y at (% of belt Belt
(m/s) (m/s) (m/s) discharge speed) Speed
(m/s) (m/s)
1 0.991 0.131 0.981 98.10 -
2 1.983 0.261 1.817 90.87 2.201
2.201 2.182 0.287 1.980 98.99 -
3 2.974 0.392 2.695 89.82 3.340
3.340 3.311 0.436 3.001 100.03 -
4 3.966 0.522 3.454 86.35 4.632
4.632 4.592 0.605 3.927 98.18 -
5 4.957 0.653 3.879 77.58 6.445
6.445 6.390 0.841 4.037 80.73 7.983
for shaped particle simulations for the low mass flow rate
Belt Velocity Velocity Velocity Factor Adjusted

Speed X Y at (% of belt Belt
(m/s) (m/s) (m/s) discharge speed) Speed
(m/s) (m/s)
1 0.991 0.131 1.013 101.28 -
2 1.983 0.261 1.864 93.20 2.146
2.146 2.128 0.280 2.037 101.83 -
3 2.974 0.392 2.798 93.27 3.216
3.216 3.188 0.420 2.997 99.90 -
4 3.966 0.522 3.647 91.19 4.387
4.387 4.349 0.573 3.962 99.04 -
5 4.957 0.653 4.390 87.80 5.695
5.695 5.646 0.743 4.232 84.64 6.728
As can be seen from the data in Tables 8.10 and 8.11, the application of equation 8.1
proved to be linear up to a belt speed of 4 m/s, however, for spherical and shaped
particles with a belt speed of 5 m/s, the equation did not follow the previous trend, with
a value still 20% lower than required. Due to time constraints, no further simulations
were performed to attempt to correct this shortfall.
If the outputs from the DEM simulations are compared to those of the experimental
results, see Figure 8.24, it is clear that experimentally there was an almost exact match
178
with the belt speed and particle discharge velocity. The fact that this was not seen in the
DEM simulations is likely due to a number of issues;
- the coefficient of rolling resistance has been set at 1% of the coefficient of static
friction, as noted in Section 8.4.3.5. This is perhaps too low a value, causing
excessive rolling of the particles with respect to the motion of the belt, and
- as well as rolling of the particles, there could also be a degree of slip between
the particles and the belt.
Although the adjusted belt speeds were now for the most part achieving the correct
particle discharge velocities, they were not ideally comparable to the results obtained
either experimentally or from the trajectory models due to the fact that the simulated
belt speeds were now somewhat higher than those actually tested experimentally.
Further investigations into these issues are explained in Section 8.4.3.8.
8.4.3.7 High Mass Flow Rate EDEM Trajectory Simulations

Even though the procedure used in Section 8.4.3.6 proved less than ideal, it was applied
to the simulations for the high mass flow rates as the simulations performed in this
section were completed before any simulation data analysis was undertaken. In
hindsight, early analysis would have picked up this discrepancy and moves to rectify
this could have been implemented earlier. This not being the case, the results of the high
mass flow rate simulations showed the same trend, however, there was still a noticeable
discrepancy present after the second iteration of simulations for belt speeds of 3 m/s and
above, as shown in Table 8.12. The procedure for determining the adjusted belt
speeds is the same as that for the low mass flow rate simulations discussed in Section
8.4.3.6.
Only spherical particle trajectories were simulated for the high mass flow rate as the
discrepancy between material discharge velocity and the corresponding belt speed were
becoming too large to obtain meaningful results.
179
for spherical particle simulations for the high mass flow rates
Belt Velocity Velocity Velocity at Factor Adjusted

Speed X Y discharge (% of belt Belt Speed
(m/s) (m/s) (m/s) (m/s) speed) (m/s)
1 0.991 0.131 1.0323 103.23 0.969
0.969 0.961 0.126 1.0128 101.28 -
2 1.983 0.261 1.9044 95.22 2.100
2.1 2.082 0.274 2.0020 100.10 -
3 2.974 0.392 2.7813 92.71 3.236
3.236 3.208 0.422 2.9229 97.43 3.321
4 3.966 0.522 3.3584 83.96 4.764
4.764 4.723 0.622 3.6442 91.11 5.229
5 4.957 0.653 3.7791 75.58 6.615
6.615 6.558 0.863 4.2150 84.30 7.847
8.4.3.8 Further Investigation of Rolling Friction

Due to the adjustments to the belt speed required to achieve the correct particle
discharge velocity in the simulations, outlined in Sections 8.4.3.6 and 8.4.3.7 above,
thought was given to the possibility that this particle slip could be due to an under-
estimation of the rolling friction.
In an attempt to prove or disprove this theory, a series of sensitivity simulations were

performed for the low mass flow rate cases for both spherical and shaped particles.
Initially the 5 m/s belt speed case was selected as this belt speed showed the largest
variation in material discharge velocity in comparison to belt speed and potentially took
the greatest number of iterative belt speed adjustments to obtain the correct discharge
velocity. Six coefficient of rolling friction values were chosen, 0.05, 0.10, 0.15, 0.20,
0.25 and 0.30, while all other particle and system parameters were kept the same as for
the original DEM simulations of Sections 8.4.3.6. Once steady-state conveying had
been achieved within each simulation, bin 3 was selected and the particle data for a
suitable time step was extracted and analysed, with the results presented in Table 8.13.
These results show that for a coefficient of rolling friction of 0.30, a near perfect
agreement was obtained when comparing the particle discharge velocity to the
corresponding belt speed. This coefficient of rolling friction was then applied to low
mass flow rate simulations with belt speeds of 1, 2, 3 and 4 m/s, again without
180
modifying any other parameters. Subsequent analysis of the particle discharge velocity
is shown in Table 8.14.
Table 8.13 Results of rolling friction sensitivity simulations for polyethylene pellets
Spherical Shaped
Particle % of Particle % of
Coefficient Velocity at Belt Velocity at Belt
Belt Speed of Rolling Discharge Speed Discharge Speed
(m/s) Friction (m/s) (m/s)
5 0.05 4.568 91.4 4.677 93.5
5 0.10 4.612 92.3 4.871 97.4
5 0.15 4.928 98.6 4.954 99.1
5 0.20 4.889 97.8 4.963 99.3
5 0.25 4.973 99.5 4.978 99.6
5 0.30 4.981 99.6 4.978 99.6
Table 8.14 Additional rolling friction sensitivity simulations for polyethylene pellets
Spherical Shaped
Belt Coefficient Velocity at Belt Velocity at Belt
Speed of Rolling Discharge Speed Discharge Speed
1 0.30 1.013 101.3 1.011 101.1
2 0.30 1.995 99.7 1.995 99.8
3 0.30 2.992 99.7 2.992 99.7
4 0.30 3.993 99.8 3.993 99.8
These results also showed a near perfect agreement when comparing the particle
discharge velocity to the original belt speed. This tabulated data has also been graphed
and is shown in Figure 8.33 and Figure 8.34.
181
Figure 8.33 Graph of discharge velocities versus width of belt for spherical particles
using a 0.3 coefficient of rolling friction
Figure 8.34 Graph of discharge velocities versus width of belt for shaped particles
using a 0.3 coefficient of rolling friction
These findings have shown that for this particular conveying application, the
assumption that the coefficient of rolling friction can be set to 1% of the static friction
(in lieu of the kinetic friction) is not valid. In hindsight, the influence of rolling
resistance should perhaps have been investigated before commencing this series of
polyethylene pellet DEM simulations. For further materials, rolling friction will be
investigated initially.
182
8.4.3.9 Re-Visiting Low Mass Flow Rate EDEM Trajectory

Simulations
The trajectory profiles for the ten simulations discussed in Section 8.4.3.8, using the
adjusted coefficient of rolling friction of 0.3, were plotted, after extracting the data from
bin 1, and are shown in Figure 8.35 for the spherical and shaped simulations. From the
comparisons presented, it is clear to see that there is negligible difference between the
results produced for spherical and shaped particles. Also of note is the increase in
particles diverging from the main trajectory stream as the belt speed increases. This was
also observed in the experimental test program, indicating that the particle dynamics in
EDEM are providing a good approximation of that occurring in reality.
Figure 8.35 Low mass flow rate EDEM simulations for spherical and shaped
particles using 0.3 coefficient of rolling friction
8.4.3.10 Re-Visiting High Mass Flow Rate EDEM Trajectory

Simulations
The adjusted coefficient of rolling friction of 0.3 was then applied to the high mass flow
rate simulations carried out in Section 8.4.3.7 with the results presented in Figure 8.36.
It must be noted that due to the number of particles being generated for the high mass
flow rate simulations, only one side of the trajectory stream has been extracted for
analysis via bin 4. Identical trends to the low mass flow rate simulations are present for
the high mass flow rate simulations.
183
Figure 8.36 High mass flow rate EDEM simulations for spherical and shaped
particles using 0.3 coefficient of rolling friction
8.4.3.11 Trajectory Simulation Comparison of 1% and 0.3 Coefficient

of Rolling Friction
It has been shown above that the DEM simulations of polyethylene pellets achieve the
same output whether spherical or shaped particles are used. This being the case, only the
spherical particles have been used in the following comparison. The simulation data for
the 1% coefficient of rolling friction has been extracted based on the adjusted belt speed
to obtain the correct material discharge velocity. The 0.30 coefficient of rolling
friction data is as presented in the previous sections. The material discharge velocity
results presented in Figure 8.11a showed that the adjusted belt speeds produced material
discharge velocities within 2% of the corresponding belt speed; however, it is clear in
Figure 8.37 that there is substantially more spread of the particles within those
trajectory streams.
For the high mass flow rate simulations there is a noticeable discrepancy between the
trajectory streams for the two approaches used to define the coefficient of rolling
friction, especially at the high belt speeds, see Figure 8.38. This was due to the effect of
particle slip in the simulations using a coefficient of rolling friction of 1% and the
particles not having the ability to achieve steady-state flow on the conveyor belt before
discharge.
184
Figure 8.37 Comparison of the low mass flow rate EDEM simulations of spherical
particles using 1% coefficient of rolling friction and 0.30 coefficient of rolling friction
Figure 8.38 Comparison of the high mass flow rate EDEM simulations of spherical
particles using 1% coefficient of rolling friction and 0.30 coefficient of rolling friction
8.4.4 Conveyor Trajectory Comparisons for Polyethylene Pellets

Experimentally, it has been shown that wings develop at the lateral extremities of the
trajectory stream for the higher mass flow rates due to a lateral velocity component
being introduced to the material as it passes through the transition zone on the conveyor
belt. Experimental comparisons with the trajectory models could not be achieved
directly as the models provide a two dimensional representation of the trajectory stream,
hence there is no way to account for the wings. This has led to the following sets of
direct comparisons being made; the experimental upper trajectory boundary being
compared with the upper trajectory boundary predicted from the models, experimental
185
trajectories versus full stream EDEM simulations for low and high mass flow rates and
trajectory models versus EDEM simulations (thin longitudinal slice only along the
centreline).
Figure 8.39 plots the experimentally determined upper trajectory boundaries for belt
speeds ranging from Vb = 1 m/s to 5 m/s. Also on this graph are the trajectory model
predictions for the corresponding belt speeds. It can be seen that for Vb = 1 m/s, the
experimental trajectory closely follows the Booth (1934) method. For belt speeds of Vb
= 2 m/s, 3 m/s, 4 m/s and 5 m/s, the experimental trajectory follows the trajectory model
grouping of for C.E.M.A. (2005), Goodyear (1975), Korzen (no air drag) (1989), Golka
et al. (no divergent coefficients) (2007) and Booth (1934). There are some minor
variations between these curves which is most likely due to the analysis method used in
the experimental testing.
When considering the EDEM trajectories produced and displayed in Figure 8.40, the
wings observed experimentally were also present in the simulations. This indicates that
the simulations were able to capture the dynamics of the material flow well, mimicking
that occurring in reality. Figures 8.40 and 8.41 provide comparison graphs of the
experimentally generated trajectories and the corresponding EDEM simulations. As is
clear in Figure 8.40, the experimental curves fit almost identically for all five belt
speeds investigated. Figure 8.41 shows the results for the high mass flow rates and it is
evident that there is some variation present for all belt speeds, most noticeably at the
lower trajectory boundary which is probably due in part to the formation of the wings.
186
Figure 8.39 Upper trajectory boundary for the high mass flow rates for the experimental data and trajectory models
187
Figure 8.40 Low experimental trajectories super-imposed over the low

mass flow rate EDEM trajectories for spherical and shaped particles
with 0.3 coefficient of rolling friction

with 0.3 coefficient of rolling friction
EDEM produces three dimensional outputs which do not allow direct comparison with
the two dimensional trajectory models. To remedy this, data binning was used during
post processing. A 40 mm wide slice was taken along the length of the conveyor and
down the centre of the trajectory stream which was then extracted for comparison with
the trajectory models. Figure 8.42 shows the trajectory model results based on the range
of belt speeds tested experimentally. For the low-speed conveying condition, Vb = 1
m/s, an inset image has been included showing a close up of the bottom of the stream.
188
For the low-speed conveying condition, the Booth (1934) method shows the best
agreement with the simulation data although the stream is slightly wider. The high-
speed conveying conditions see several trajectory models producing nearly identical
streams and have been grouped together. The EDEM simulation data fits extremely well
with the trajectory models for C.E.M.A. (2005), Goodyear (1975), Korzen (no air drag)
(1989), Golka et al. (no divergent coefficients) (2007) and Booth (1934).
8.5 CONVEYOR TRAJECTORIES OF IRON ORE

Due to the highly dusty nature of the iron ore tested experimentally, there was no
feasible way to produce experimental trajectories. With the open nature of the conveyor
transfer research facility, capture of the product by the trajectory hopper would have
seen substantial dust generated during both the flight of material and also after impact
with the trajectory hopper. As no experimental trajectories were produced, it was
deemed unnecessary to produce analytical trajectories or DEM trajectories.
8.6 CONVEYOR TRAJECTORIES OF CORN
8.6.1 Experimental Conveyor Trajectories

The conveyor trajectories for corn were performed using the test arrangement described
in Section 8.4.1.3. Calibration of the feed bin load cells and Hogan valve feed rates
were carried out as described in Section 4.4 and 4.5. Table 8.15 summarises the corn
feed rates for each belt speed tested.
Table 8.15 Experimental trajectory test setups
Belt Speed High Feed Rate

(m/s) (tph)
1 19.6
2 33.2
3 45.9
4 55.0
5 65.8
189
Figure 8.42 High mass flow rate trajectory streams for the trajectory models and EDEM simulations
190
The same experimental procedure was used as for the polyethylene pellets, with
sequential photos of each trajectory stream analysed and compiled to produce the results
shown in Figure 8.43. A maximum belt speed of Vb = 5 m/s was tested for the corn as at
this speed the trajectory stream was beginning to lose integrity and spillage was
resulting. The wings which were observed during the polyethylene pellet tests were
also observed when the experimental corn trajectories were produced.
Figure 8.43 Experimental corn trajectories for high mass flow rates
As with the polyethylene pellet testing, the possibility of particle slip was investigated.
Reviewing the method used previously, there was no direct way to measure the belt
speed from the high-speed video footage, other than relying on the setting of the
variable speed drives at the beginning of each test. The Redlake X3 MotionPro high-
speed digital video camera was mounted vertically above the discharge point of the
conveyor as shown in Figure 8.44a. The video camera was aligned so as to capture the
flow of one side of the particle stream to allow sufficient resolution to obtain accurate
results as shown in Figure 8.44b. Divisions of 50 mm were drawn along the entire
length of the conveyor belt to provide both the scaling factor and a marker to track the
speed of the belt and although hard to see, are visible in Figure 8.44b.
191
(a) (b)
Figure 8.44 (a) Vertical positioning of the Redlake X3 MotionPro high-speed

digital video camera for analysis of the particle discharge velocity and
(b) an example of corn for Vb = 1 m/s
The results of the discharge velocity analysis are shown in Figure 8.45 and indicate a
perfect match for belt speed and particle discharge velocity. The test for a belt speed of
Vb = 5 m/s is in fact approximately 4.8 m/s which is most likely due to the variable
speed drive not being adjusted exactly.
Figure 8.45 Comparison of the particle speed of corn and the belt speed at the
discharge point of the conveyor
192
Regardless of this, for this belt speed, the particles are discharging at the same speed as
the belt. From these results, it can be concluded that the trajectory profiles presented in
Figure 8.43 are accurate for the five belt speeds tested. At belt speeds above 5 m/s, the
trajectory stream was losing integrity and thus there would have been difficultly
defining the experimental trajectory accurately.
8.6.2 Analytically Determined Conveyor Trajectories

As with the polyethylene pellets, analytical based trajectories were only produced up to
and including a belt speed of 5 m/s as a result of the particle slip investigations. The
analytical methods described in Section 2.4 have again been used. The generated
conveyor profiles for the numerous methods and belt speeds are presented in Figure
8.46.
On review of the trajectories generated for each belt speed investigated it was found that
the majority of observations were similar to those for the polyethylene pellets, the key
observations being;
- for a belt speed of Vb = 1 m/s, low-speed conveying conditions apply and each of
the trajectory methods generates a separate trajectory profile for the lower and
upper boundaries,
- the Golka method with and without the divergent coefficients used, produce
nearly identical profiles,
- the C.E.M.A. (1966; 1979; 1994; 1997) and M.H.E.A. (1986) methods produce
identical profiles for each of the belt speeds investigated,
- for a belt speed of Vb = 2 m/s, high-speed conveying conditions apply,
- for a belt speed of Vb = 2 m/s, the C.E.M.A. (2005) and Goodyear (1975) methods
produce identical profiles,
- for a belt speed of Vb = 2 m/s, the Golka (2007) method without divergent
coefficients and the Korzen (1989) method without air drag produce identical
profiles,
- for a belt speed of Vb = 2 m/s, the C.E.M.A. (1966; 1979; 1994; 1997) and
M.H.E.A. (1986) methods clearly produce the largest trajectory and continue to
do so for the higher belt speeds also,
- For all belt speeds exhibiting high-speed conditions, the Golka (2007) method
without divergent coefficients falls symmetrically inside the C.E.M.A. (2005)
193
method, while the Golka method with divergent coefficients falls symmetrically
outside the C.E.M.A. method,
- as belt speed increases, there is a noticeable merging of several trajectory
methods,
- for a belt speed of Vb = 3 m/s, the same trajectory method groupings exist as for
the Vb = 2 m/s case,
- for a belt speed of Vb = 3 m/s, the Korzen (1989) method incorporating air drag is
beginning to diverge from the other trajectory methods and is falling closer to the
conveyor head pulley,
- for belt speeds of Vb = 4 m/s and Vb = 5 m/s, the same trajectory method
groupings apply and exhibit the same trends for both,
- the Korzen (1989) method incorporating air drag is now more noticeably falling
closer to the conveyor head pulley than any of the other methods.
194
Figure 8.46a Analytically determined conveyor trajectories for Vb = 1 m/s
195
Figure 8.46b Analytically determined conveyor trajectories for Vb = 2 m/s

Figure 8.46c Analytically determined conveyor trajectories for Vb = 3 m/s
196
Figure 8.46d Analytically determined conveyor trajectories for Vb = 4 m/s

Figure 8.46e Analytically determined conveyor trajectories for Vb = 5 m/s
8.6.3 DEM Conveyor Trajectories

Only EDEM has been used to produce the simulated trajectories as the required high
mass flow rates can be achieved and also the effect of shape can be investigated. A
197
maximum belt speed of 5 m/s has been simulated to keep the results in line with those
obtained experimentally and from the analytical models.
8.6.3.1 Particle Geometry

The natural shape of a corn kernel is more of a flat rectangular geometry and requires
more detail to construct a representative model to use within EDEM. Measurements of
particle thickness, length and width have been taken for 50 particles to obtain an
equivalent volume diameter for the spherical simulations and an average particle shape
with which to model the shaped particles. It has been previously stated that the
simulation time increases noticeably as more spherical particles are used to generate a
clustered particle. As a result, this led to the generation of a 6-sphere cluster, as shown
in Figure 8.47a. As can be seen, there are substantial curves and valleys present,
however, the shaped particle simulations in the following sections have been based off
this particle representation. To investigate whether there was any benefit of modelling a
more accurate particle, a 12-sphere representation was also generated, see Figure 8.47b.
This particle was used to produce a simulation as part of the discharge angle
comparisons in Section 7.7.3.2.
The size distribution of the corn is shown in Table 5.5 and the size range from 4.75 mm
to 9.5 mm has been used, accounting for 98.9% of the sample.
(a) 6-sphere representation (b) 12-sphere representation
Figure 8.47 Particle representations of corn used in EDEM
198
8.6.3.2 Calibration of the Mass flow Rate

The calibration procedure followed the same method used for polyethylene pellets in
Section 8.4.3.3. The results for the shaped particles apply to both the 6-sphere and 12-
sphere representations as the mass of each particle is based on the template volume
rather than the volume of the individual spheres used in creating each representation.
The results are summarised in Table 8.16. The relationship for the number of particles
versus material mass flow rate was then determined to predict the number of particles
required to produce mass flow rates comparable to those used experimentally, see Table
8.17. The data in Table 8.17 also shows a near identical match for number of spherical
particles compared to the shaped particles. This is because the determination of the mass
of the shaped particles is based on the template volume, which is the same volume used
to determine the equivalent volume diameter for the spherical particles. The only
differences will be as a result of the random generation of particles within the size
distribution specified.
Table 8.16 Mass flow rate calibration of corn in EDEM
Spherical Shaped
Number of Generation Mass of Mass flow Mass of Mass flow
particles time particles rate particles rate
(sec) (kg) (tph) (kg) (tph)
1000 1 0.362 1.302 0.375 1.350
2500 1 0.916 3.297 0.938 3.378
10000 1 3.720 13.392 3.770 13.572
20000 1 7.387 26.594 7.513 27.046
50000 1 18.646 67.125 18.739 67.462
Table 8.17 Required number of corn grains to achieve the experimental

mass flow rates
Mass flow rate Number of spherical Number of shaped

(tph) particles required particles required
19.6 14659 14505
33.2 24785 24586
45.9 34241 34001
55.0 41016 40747
65.8 49058 48753
199
8.6.3.3 Conveyor Geometry

The conveyor geometry is identical to that used for the polyethylene pellet trajectories,
but for all simulations, the feed chute furthest from the head pulley has been used to
ensure that steady-state conveying has been achieved for all belt speeds, as summarised
in Table 8.9.
8.6.3.4 Calibration of Rolling Friction

Before starting the discharge angle simulations, detailed in Chapter 7, or the trajectory
simulations, the effect of varying the coefficient of rolling friction was investigated. A
sensitivity analysis was conducted for both spherical and 6-sphere shaped particles
using the conveyor belt CAD geometry.
Initially the 5 m/s belt speed case was selected as this belt speed would potentially show
the largest variation in material discharge velocity in comparison to belt speed. Six
coefficient of rolling friction values were chosen, 0.05, 0.10, 0.15, 0.20, 0.25 and 0.30,
while all other particle and system parameters were kept constant as per the data in
Table 5.3. Once steady-state conveying had been achieved within each simulation, bin 3
was selected and the particle data for a suitable time step was extracted and analysed,
with the results presented in Table 8.18.
Table 8.18 Results of rolling friction sensitivity simulations for corn
Spherical Shaped
Coefficient Velocity at Belt Velocity at Belt
Belt Speed of Rolling Discharge Speed Discharge Speed
5 0.05 4.3455 86.9 4.9916 99.8
5 0.10 4.729 94.6 4.9888 99.8
5 0.15 4.9389 98.8 4.9955 99.9
5 0.20 4.9875 99.75 4.9929 99.9
5 0.25 4.9919 99.84 4.9964 99.9
5 0.30 4.9905 99.81 4.9977 99.95
These results showed that for spherical particles, a coefficient of rolling friction of 0.20
or higher produced a near perfect agreement when comparing the particle discharge
velocity to the corresponding belt speed. It was decided that a coefficient of rolling
200
friction of 0.25 should then applied to simulations with belt speeds of 1, 2, 3 and 4 m/s,
again without modifying any other parameters. Subsequent analysis of the particle
discharge velocities are shown in Table 8.19. For the shaped particle simulations, it was
found that even at a coefficient of rolling friction of 0.05, the particle discharge velocity
was nearly identical to the corresponding belt speed (99.8%). This result will be as a
direct result of the particle shape being of a rectangular flat geometry which restricts the
rolling potential of the particles. Even though it was in all likelihood safe to choose a
coefficient of rolling friction of 0.05 in the subsequent simulations, the decision was
made to err on the conservative side and a value of 0.15 was applied to simulations with
belt speeds of 1, 2, 3 and 4 m/s, again without modifying any other parameters.
Subsequent analysis of the particle discharge velocities are shown in Table 8.19.
Table 8.19 Additional rolling friction sensitivity simulations for corn
Spherical Shaped
Particle Particle
Belt Coefficient % of Coefficient % of
Velocity at Velocity at
Speed of Rolling Belt of Rolling Belt
Discharge Discharge
(m/s) Friction Speed Friction Speed
(m/s) (m/s)
1 0.25 1.0243 102.43 0.15 1.0195 101.95
2 0.25 1.9995 99.98 0.15 2.0011 100.06
3 0.25 2.9964 99.88 0.15 2.9975 99.92
4 0.25 3.9971 99.93 0.15 3.9972 99.93
These findings have again shown that for this particular belt conveying application, the
assumption that the coefficient of rolling friction can be set to 1% of the static friction
(in lieu of the kinetic friction) is not valid.
8.6.3.5 High Mass Flow Rate EDEM Trajectory Simulations

The decision was made to only produce experimental trajectories with a high mass flow
rate and this has also been applied to the EDEM simulations. Simulations were
generated for the spherical particle representations of corn using a coefficient of rolling
friction of 0.25, for belt speeds of 1 m/s, 2 m/s, 3 m/s, 4m/s and 5 m/s. Simulations were
also generated for the 6-sphere shaped particle representation of corn using a coefficient
of rolling friction of 0.15 for belt speeds of 1 m/s, 2 m/s, 3 m/s, 4m/s and 5 m/s. Data
was extracted from each of these simulations using the existing bins, once steady-state
201
conditions had been achieved. The results of the ten simulations have been combined
and displayed in Figure 8.48, representing the entire stream width. The results show that
there is negligible difference between the results obtained when comparing spherical
and shaped representations of corn.
Figure 8.48 High mass flow rate EDEM simulations for spherical and shaped
particles
8.6.4 Conveyor Trajectory Comparisons for Corn

Due to wings forming at the lateral extremities of the trajectory streams, experimental
comparisons with the trajectory models could not be achieved directly as the trajectory
models have no way to account for the wings. This has led to the following sets of
direct comparisons being made; the experimental upper trajectory boundary being
compared with the upper trajectory boundary predicted from the models, experimental
trajectories versus full stream EDEM simulations for high mass flow rates and
trajectory models versus EDEM simulations (using a thin longitudinal slice only along
the centreline).
Figure 8.49 plots the experimentally determined upper trajectory boundaries for belt
speeds ranging from 1 m/s to 5 m/s. Also on this graph are the trajectory model
predictions for the corresponding belt speeds. For a belt speed of 1 m/s, the
experimental trajectory closely follows the method of Booth (1934). For belt speeds of
2 m/s, 3 m/s, 4 m/s and 5 m/s, the experimental trajectory follows the trajectory model
202
grouping of for C.E.M.A. (2005), Goodyear (1975), Korzen (no air drag) (1989), Golka
et al. (no divergent coefficients) (2007) and Booth (1934).
When considering the EDEM trajectories produced and displayed in Figure 8.50, the
wings observed experimentally were also present in the simulations. This indicates that
the simulations were able to capture the dynamics of the material flow well, mimicking
that occurring in reality. Figure 8.50 provides a comparison of the experimentally
generated trajectories and the corresponding EDEM simulations. As is clear in Figure
8.50, the experimental curves for the upper trajectory boundary fit almost exactly with
the EDEM results, however, there is a slight variation at a belt speed of 5 m/s which
could be as a result of the experimental trajectory beginning to lose integrity, making
the profiling of the stream slightly inaccurate. The lower trajectory stream from the
experimental results is accurate for a belt speed of 1 m/s, however, as the belt speed
increases, the effect of the wings becomes more evident. There is a noticeable deviation
from the trajectory streams from EDEM which would tend to indicate that the wings
were not reproduced in the simulations to the degree they existed experimentally.
As previously mentioned EDEM produces three dimensional outputs which do not

allow direct comparison with the two dimensional trajectory models. To remedy this,
data binning was used during post processing. A 40 mm wide slice was taken along the
length of the conveyor and down the centre of the trajectory stream which was then
extracted for comparison with the trajectory models. Figure 8.51 shows the trajectory
model results. For the low-speed conveying condition, Vb = 1 m/s, an inset image has
been included showing a close up of the bottom of the stream. For the low-speed
conveying condition, the C.E.M.A. (1966; 1979; 1994; 1997), M.H.E.A. (1986) and
Booth (1934) methods show the best agreement with the simulation data although the
DEM stream is slightly wider. The high-speed conveying conditions see several
trajectory models producing nearly identical streams and have been grouped together.
The EDEM simulation data fits extremely well with the trajectory models for C.E.M.A.
(2005), Goodyear (1975), Korzen (no air drag) (1989), Golka et al. (no divergent
coefficients) (2007) and Booth (1934).
203
Figure 8.49 Upper trajectory boundary for the high mass flow rates for the experimental data and trajectory models
204

8.7 DISCUSSION
This chapter has presented a number of the more widely used and/or readily available
trajectory prediction methods published in the literature. They range in complexity from
the basic, Goodyear (1975) and Dunlop (1982), to the complex, Booth (1934), Golka
(1992; 1993) and Korzen (1989). Some methods include a multitude of parameters such
as C.E.M.A. (1966; 1979; 1994; 1997; 2005) and M.H.E.A. (1986) while others
incorporate parameters which no others address, such as divergent coefficients (Golka,
1992; Golka, 1993; Golka et al., 2007) and air drag (Korzen, 1989).
Comparisons for the low-speed conveying conditions show that each trajectory method
produces a slightly different profile. Some of this variation is due to the discharge angle
determined from each method, as detailed in Chapter 7. Once high-speed conveying is
achieved, all trajectories leave at the point of tangency between the belt and head pulley,
and there is a tendency for some of the methods to predict nearly identical trajectory
profiles, especially as the belt speed increases.
A major observation of the trajectory methods is that the Korzen (1989) method predicts
a trajectory profile which is noticeably below the other curves, due to the inclusion of
air drag. In considering air drag, one can expect the trajectory stream to drop away
much quicker than the other methods, but is it truly representative? In the context of this
research it has proved not to be, but in saying that, the experimental test work was
performed indoors in a very still environment and the particles tested experimentally did
205
Figure 8.51 High mass flow rate trajectory streams for the trajectory models and EDEM simulations
206
not contain a sufficient quantity of fines for which air drag is more relevant, so the true
impact could not be gauged. In an outdoor situation, cross-winds are an environmental
effect which definitely needs to be considered and this can be achieved using the
Korzen method.
The experimental test program consisted of producing a range of trajectories for

polyethylene pellets and corn. These results uncovered a number of issues which have
not been reported on previously. They include: the dynamic effect of the material on the
conveyor belt before discharge as it passes through the transition zone, where the lateral
spread of material creates curved wings at the extremities of the trajectory stream; and
the effect that particle slip has on the particle discharge velocity as it leaves the head
pulley and the resulting prediction of trajectories based incorrectly on belt speed rather
than particle discharge velocity.
The discrete element modelling section of this research has also uncovered a number of
interesting aspects. Firstly, as detailed in the sensitivity analysis in Section 6.3.9, it is
not critical in this particular application that all parameters be 100% accurate in the
initial setup of the simulations (i.e. shear modulus and Poissons ratio). In saying this, it
is still best to have indicative values which represent the materials in question, as
closely as possible.
Investigations were performed for both spherical particles, where the equivalent volume
diameter is determined based of the average of true particle measurements and also
shaped particles, where more realistic models of the test materials have been generated
with the aid of CAD software. The result from these investigations is that there is very
little difference between the trajectories generated for the spherical particles and the
shaped particles. This implies that for trajectory investigations, spherical particles can
produce as valid a result as the shaped particles in much less simulation time and also
less setup time, when considering the generation of shaped particles.
Comparisons of the various trajectory methods yielded some interesting results also.
Due to the wings discovered in the experimental testing, only the upper trajectory
boundary could be compared with the analytical trajectory models. An additional reason
for this is the fact that the analytical models are two-dimensional and as such, the top
207
most point of the experimental trajectories should be a direct match for the outputs from
the analytical models, whereas the centre most point of the lower experimental
trajectory boundary cannot be visualised. The comparisons showed that the Booth
(1934) method was a very close match for the experimental upper trajectory boundaries
for all belt speeds tested.
The experimental trajectory profiles could be directly compared to the EDEM outputs
and these comparisons showed a good agreement for the upper trajectory boundary but
there was some variation in the lower trajectory boundary.
Comparing the analytical trajectory methods to the EDEM outputs required two-
dimensionalising the EDEM trajectory streams via bins. This allowed a thin
longitudinal slice of the EDEM output to be compared with the analytical methods. This
time, both the lower and upper trajectory boundaries could be compared and it was
found for both polyethylene pellets and corn, that the EDEM trajectory stream matched
the Booth (1934) method nearly identically for all belt speeds tested.
Overall, to accurately determine the profile of a conveyor trajectory, the logical choice
is to determine this experimentally but in an effort to save the cost of establishing either
pilot plants or downtime on operational facilities, the option of either using the Booth
(1934) method or EDEM are viable alternatives. This accurate prediction is especially
vital in view of the design and positioning of conveyor transfer hoods, which will be
discussed in Chapter 9. This final statement must be interpreted in the context of the
products investigated as part of this research.
208
CHAPTER 9 CONVEYOR TRANSFER HOOD ANALYSIS
Chapter
9
CONVEYOR TRANSFER HOOD ANALYSIS
209
9.1 INTRODUCTION
The upper section of a conveyor transfer can be comprised of a number of different
elements, including; impact plates, rock boxes and transfer hoods. The positioning of
any of these elements is critical for the smooth flow of material through the remainder
of the transfer. The impact point of the incoming stream will be determined as a result
of the conveyor trajectory, which was detailed in Chapter 8. This chapter will detail the
experimental investigation of particle flow through a transfer hood and the
corresponding analytical modelling and discrete element simulation of the particle flow,
finally providing comparisons between the methods.
9.2 THE FLOW OF POLYETHYLENE PELLETS THROUGH A

TRANSFER HOOD
9.2.1 Experimental Particle Flow Investigation

One of the key features of the conveyor transfer research facility is that the transfer
enclosure and hood, see Figure 9.1, and spoon have been constructed of acrylic. This
provides the ability to record a variety of material flow characteristics with both high-
speed video and digital still cameras. The Redlake X3 MotionPro high-speed digital
video camera has been used to capture the particle flow through the hood.
Figure 9.1 Detail of the conveyor transfer hood
210
Two conveyor belt speeds, Vb, have been investigated, Vb = 2 m/s and Vb = 3 m/s.
These both result in high-speed conveying conditions, with the material discharging
from the point of tangency between the conveyor belt and the head pulley. Initially, a
conveyor belt speed of Vb = 1 m/s was also to be investigated but the geometry of the
conveyor resulted in slow-speed conditions. The product achieved a substantial angle of
wrap around the head pulley before discharge and as a consequence the material had
very little horizontal displacement, meaning the use of a transfer hood was not required.
To investigate the influence of mass flow rate on the velocity of the particle stream, a
low mass flow rate was selected as well as a mass flow rate which allowed the
conveyors to operate at full capacity, based on edge distance calculations (C.E.M.A.,
2005). Table 9.1 summarises the feed rates used. Initially, a low feed rate of 2 tonnes
per hour (tph) was used for the 3 m/s belt speed, however, this was found to be too low
when analysing the data via Image Pro Plus, hence the increase to 10 tph for the
subsequent hood geometry and analysis.
Table 9.1 Product feed rates used in experimental tests
Belt Speed, Vb Low Feed Rate High Feed Rate

(m/s) (tph) (tph)
2 2 31
3 (Position A) 2 38
3 (Position B) 10 38
One hood position was used for the Vb = 2 m/s hood geometry. For the Vb = 3 m/s case,
two hood positions were used, the initial hood position was aligned so that the incoming
flow would cause a substantial angle of incidence with the hood, while the second case
aligned the hood to minimise the angle of incidence. Figure 9.2 shows a snapshot of the
steady-state flow through each of the transfer hood geometries for both the low and high
feed rates.
A number of key observations can be made from Figure 9.2:
- for each of the high feed rates, the angle of incidence is less than the low feed rate
equivalent due to the depth of the material stream,
- Figure 9.2c highlights a substantial amount of spray of particles after impact due
to the high angle of incidence, whereas in Figure 9.2d there is no spray,
211
- Figure 9.2b, 9.2d and 9.2e each show the product stream spreading onto the hood
wings,
- the angle of incidence in Figure 9.2f is near what would be considered ideal when
compared to that of Figure 9.2d,
Figure 9.2 Material flow through the conveyor hood

(a) Vb = 2 m/s and ms = 2 tph, (b) Vb = 2 m/s and ms = 31 tph,
(c) Vb = 3 m/s Pos A ms = 2 tph, (d) Vb = 3 m/s Pos A ms = 38 tph,
(e) Vb = 3 m/s Pos B ms = 10 tph, (f) Vb = 3 m/s Pos B ms = 38 tph
212
- the position of the hood for Vb = 3 m/s position A (Figure 9.2c and 9.2d) did not
allow for any measurement at the exit point of the hood, i.e. = 0.
The particle velocity was determined using the software package Image Pro Plus (IPP).
Calibration of the linear distance was first performed to ensure IPP analysed the video
footage correctly. The linear calibration was performed by selecting a known
measurable distance on a frame of the video and entering the true length. The time step,
see equation 9.1, was determined from the number of frames per second (FPS) being
recorded by the Redlake X3 MotionPro high-speed digital video camera.
1
Time step = 9.1
FPS 1
Utilising the manual tracking feature, particles are tracked by selecting the particle
centroid at each time step at each five degree increment, as shown in Figure 9.3. The
results for each particle are tabulated within IPP and then exported for further analysis.
In most instances, the number of particle tracked at each angular position was between
10 and 20, dependant on clarity of the video footage. The average velocity was then
determined for these particles, as well as the minimum and maximum velocity at each
angular position.
The averaged particle velocities, at each angular position for the six cases shown in
Figure 9.2, are presented in Figure 9.4. The following observations can be made from
the average particle velocity results:
- for Vb = 2 m/s, both the low and high feed rate start with an approximate velocity
of 2 m/s before increasing to a velocity of 2.75 to 2.9 m/s at the hood exit,
- the Vb = 3 m/s position A tests with the high angle of incidence showed a
pronounced drop in velocity soon after impact before steadily increasing to
approximately 3 m/s at hood exit,
- the Vb = 3 m/s position B tests showed a more consistent average velocity through
the hood, even more so for the high feed rate test. There was still, however, a
slight rise in overall particle velocity towards the hood exit.
213
Figure 9.3 Particle tracking using Image Pro Plus
Figure 9.4 Average particle velocity at each angular position around transfer hood
9.2.2 Analytical Method Analysis

Predicting the flow through a conveyor hood is possible with the use of two analytical
based methods: the inverted chute-flow model (Roberts, 2003) and the model for
cohesive material flow (Korzen, 1988). Each method requires parameters such as, initial
particle velocity, hood impact angle and the initial and average height and width of the
particle stream. If the transfer hood had not already been constructed, then estimates of
these values would have been used in order to generate a solution. However, with this
research comes the added benefit that these values can be extracted directly from the
experimental test program to better compare the analytical methods with the
experimental results, refer to Figure 9.5a and Figure 9.5b for examples of measuring the
214
stream height and width. The number of decimal places displayed in these figures is a
direct result of the software settings.
Figure 9.5a Material stream height Figure 9.5b Material stream width
through the hood through the hood
9.2.2.1 Analytical Method of Roberts

The inverted chute-flow method of Roberts (2003) is a widely accepted approach to
predicting the stream velocity through a transfer hood for granular cohesionless
materials, the force diagram is presented in Figure 9.6.
Figure 9.6 Force diagram for the inverted curved chute
215
A coefficient of equivalent friction, e, is used, which incorporates the particle wall

friction, the stream cross-section and the internal shear of the bulk solid, see equation
9.2, and is assumed to be an averaged constant for all angular positions analysed
through the hood, as the stream thickness is comparatively low. The ratio of the
pressure acting on the sides of the chute to the pressure acting on the bottom of the
chute, Kv, is generally assumed to be a value between 0.4 and 0.6 according to Roberts
(1999; 2003). There were no means of directly measuring these pressures and as such,
an estimate of 0.4 has been used based on the fact that the height of the material stream
is substantially less than the width of the stream.
The particle velocity at any given angular position through the hood can then be found
using equation 9.3, by first determining the constant of integration, K, by substitution of
the initial conditions, v = v0 and = 0. The initial velocity used in this analysis can be
assumed to be the belt speed of the feeding conveyor if the transfer hood is located close
to the belt, or can be approximated from trajectory models. For the experimental work
in this test program the actual velocity of the product stream was measured just before
impact with the hood and has been applied directly.
K v v0 H 0
e = w 1 + 9.2
vB
2 gR
v=
4 e + 1
2 ( 2e2 1) sin + 3e cos + Ke 2 e 9.3
Applying the experimental impact angle as the starting point for the analysis for each
belt speed and hood position, the results of the Roberts analytical analysis are presented
in Figure 9.7. The data presented shows a significant increase in stream velocity through
the hood for both the low and high feed rates for the 2 m/s belt speed, whereas for both
3 m/s belt speed cases there is a relatively small increase in stream velocity through the
transfer hood.
216

transfer hood using the Roberts method
9.2.2.2 Analytical Method of Korzen

Korzen (1988) investigated the dynamics of material flow on impact plates for both
cohesive and non-cohesive materials. Figure 9.8 is a representation of the stream
behaviour through this flow zone. This cohesive material model was originally intended
to have a zone of built up stationary material attached to the impact plate which the
main flow stream passed over.
Figure 9.8 Flow representation for analysis by Korzen
217
Korzen assumed this curved surface was of constant radius and the subsequent velocity
analysis in equation 9.4 used the internal friction coefficient, , due to the flow stream
shearing against the stationary material. As the polyethylene pellets used in this research
are non-cohesive and free flowing, it has been assumed that the curved shearing surface
can be replaced by a curved chute of constant radius, resulting in the coefficient of wall
friction, w, being used in the velocity analysis of equation 9.4.
The velocity of the material stream at an arbitrary angle, , can be expressed by

equation 9.4 and the constant of integration, K, can be solved using v() = vp, = p
and A() = Ap as the initial conditions.
2 gR 5 w sin + ( 4 w 2 1) cos cBgR

v ( ) = Ke4 w +
2
9.4
(1 + 16w )
2 2 A ( )
w
This method requires the cross sectional area of the material profile at each angular
position as a function of mass flow rate, stream velocity and bulk density. As previously
mentioned, the experimental testing has provided the direct measurement of the height
and width of the product stream, thus the true cross sectional area of the product stream
at each angular position through the resulting flow can be determined. It should also be
noted that for this experimental test program, the material chosen is non-cohesive,
therefore c = 0 and the right most part of equation 9.4 equals zero, negating the need to
determine the stream cross sectional area. As a result, the Korzen analytical method
takes on a similar form to that presented by the Roberts method.
belt speed and hood position, the results of the Korzen analytical analysis are presented
in Figure 9.9. The data presented shows a significant increase in stream velocity through
the hood for both the low and high feed rates for the 2 m/s belt speed in the same way as
the Roberts method, whereas for both 3 m/s belt speed cases there is nearly no change in
stream velocity through the transfer hood.
218

transfer hood using the Korzen method
9.2.3 Discrete Element Modelling of Particle Flow

DEM simulations have been performed predominantly using the Chute MavenTM
software which focuses on the simulation of particle flow through conveyor transfers.
EDEM has also been used to allow high mass flow rates to be achieved for key
experiments only. With both software packages, first a three dimensional CAD model is
imported into the software defining the relevant conveyor components, as well as the
particle and system parameters supplied in Table 5.1. The Chute MavenTM software
simulates particles as spheres only so there will be a degree of variation between the
simulations and the experimental results, in all likelihood resulting in an over-prediction
of the velocity through the transfer hood as the simulated spherical particles should be
more free-flowing. EDEM has the facility to create shaped particles by clustering
spherical particles and this shape effect will be investigated in limited detail for the
simulations performed.
9.2.3.1 Chute MavenTM Simulations

The degree to which particles roll or slide during a simulation is quantified by the
restrain parameter. The restrain of the particles is defined as 100% for fully sliding and
0% for fully rotational, other percentages refer to combinations of the two; refer back to
Section 6.2.3 for a full explanation. In an attempt to quantify a representative restrain
value, the high-speed video footage of the experimental tests was reviewed and it was
219
concluded that the percentage of particles which fully rotate on the surface of the
Polystone Ultra liner was dependant on stream thickness. In regions of substantial
stream thickness with minimal voidage, the percentage of particles able to rotate was
low, visually around 10%. However, the number of particles able to fully rotate or roll
is even lower due to the compaction of the particles. In regions of low stream thickness,
it was observed that approximately 30% of particles could roll as the stream was less
constrained. To assume that all particles are fully restrained, especially for a free
flowing material, is not ideal, thus a particle restrain of 80% was selected for the Chute
MavenTM simulations.
To confirmation this assumption, a selection of belt speeds and mass flow rates were
selected and DEM simulations performed as a sensitivity analysis, focusing on variation
of particle restrain, while leaving all other parameters constant. Three particle restrains
were investigated, 100%, 80% and 50%. On review of the outputs it was found that as
the restrain was reduced, there was an increase in the quantity of rogue particles which
would diverge from the main flow stream. However, analysis of the main steady-state
flow stream found no change in the average stream velocity at the exit of the hood for a
given belt speed regardless of mass flow rate. This finding indicates that the assumption
to use 80% particle restrain for the Chute MavenTM simulations is acceptable.
Further calibration was performed to investigate the issue of particle friction, raised in
Section 6.2.2.1. Two Chute MavenTM simulations were completed; see Table 9.2, where
only the coefficient of particle friction was modified in each test. The value for
coefficient of particle friction used in test 1 was based on the wall yield loci test on a
sheet of Polyethylene and test 2 was based on the result of an instantaneous yield loci
test. The resulting average stream velocities at each 5 angular position around the hood
are presented in Table 9.3. At the point of discharge from the transfer hood (i.e. = 0),
there is effectively no change in the average stream velocity between the two tests. This
appears to indicate that the coefficient of particle friction is not one of the primary
parameters that will affect the outcome of a simulation and as such the concern raised
over which method to use in determining the coefficient of particle friction was
somewhat unwarranted.
220
Table 9.2 Chute MavenTM DEM simulation parameters
Test Belt Speed Feed Rate Coefficient of Coefficient of % Restrain

(m/s) (tph) Particle Friction Wall Friction
1 2 5 0.222 0.282 80
2 2 5 0.966 0.282 80
Table 9.3 Average stream velocity from particle friction calibration
Angle from Horizontal ()

Test 0 5 10 15 20 25 30 35 40 45
1 3.264 3.177 3.083 2.961 2.849 2.707 2.567 2.410 2.230 2.140
2 3.261 3.164 3.076 2.965 2.838 2.704 2.577 2.414 2.238 2.161
Three Chute MavenTM DEM simulations were performed using the low product feed
rates shown in Table 9.1, to allow direct comparisons with the experimental results.
DEM simulations for the high mass flow rates, used experimentally, could not be
produced due to the limitations described in Section 6.2.7.2. Instead, the decision was
made to look for other ways to produce comparisons. The effect of mass flow rate in the
simulations was investigated by preparing additional tests for mass flow rates between
those tested experimentally, the full series of simulations being;
- 2, 5 and 10 tph for a belt speed of 2 m/s, and
- 2, 5, 10 and 15 tph for the 3 m/s belt speed hood positions
The results of these tests would hopefully present trends which could be applied to
predict what would happen at the higher feed rates. On completion of the simulations,
an example of which can be seen in Figure 9.10, the simulated product streams were
extracted and combined to produce Figure 9.11. The following observations were made;
- as product feed rate increases, so does the height of the material stream, which in
turn reduces the angle of incidence at the point of impact with the transfer hood,
- hood position A for the 3 m/s belt speed shows substantial particles diverging
from the main particle stream due mainly to the high angle of incidence, which is
backed up in Figure 9.2c where particles can be seen diverging in the
experimental hood,
221
- hood position B for the 3 m/s belt speed has a more controlled stream flow as a
result of the optimised positioning of the hood, minimising the angle of incidence
of the material in-flow.
Figure 9.10 Example output from a Chute MavenTM simulation
Figure 9.11 Extracted Chute MavenTM simulation outputs of the various product
feed rates for the different belt speeds and transfer hood positions
Further post processing of the simulation data focussed on the average stream velocity
through the transfer hood for each belt speed and hood position and the results are
presented in Figure 9.12. Matlab was used for this analysis, an example of the M-file
222
used for this can be found in Appendix C. Only particles with their centres within 3 mm
of the Polystone Ultra liner and also within 50 mm either side of the central flow axis
were selected. The reason for this was two-fold; first to ensure the only particles being
analysed were the ones in contact with the conveyor transfer and secondly, so that the
width of the stream analysed could be considered a two-dimensional slice to produce
particle velocity data which could be directly related back to the analytical models. In
all cases there was a transient behaviour of the average stream velocity at the point of
impact with the transfer hood. It is also evident that once this initial transient zone has
passed, the average particle velocity for each group of tests falls along the same line,
resulting in the average exit velocity of the stream being the same. This fact would seem
to indicate that regardless of the mass flow rate used, the average exit velocity of the
stream will be equivalent to those shown in Figure 9.12. This trend may or may not hold
true for other products and will be investigated in future research.
Figure 9.12 Chute MavenTM DEM simulation results for all mass flow rates
Comparing Figure 9.11 and Figure 9.12, there is an indication that the transition
between the transient behaviour at the impact point and the steady-state flow
approximately coincides with the imaginary point at which the lower trajectory stream
would impact with the hood if the material stream was extremely thin.
223
9.2.3.2 EDEM Simulations

The setup for the EDEM transfer hood simulations uses the same method that was
employed for the conveyor trajectories. The three-dimensional CAD model has been
updated to incorporate the transfer hood and spoon, the results from the transfer spoon
being reported in Chapter 11. The results of the mass flow rate calibration and
coefficient of rolling friction for the trajectory simulations have been applied to these
simulations, along with the same particle and system parameters. Figure 9.13 shows the
model geometry imported into EDEM for the 2 m/s belt speed.
Figure 9.13 Conveyor transfer geometry for Vb = 2 m/s imported into EDEM
When the decision was made to incorporate EDEM into this research, it was not the
intention to repeat every simulation performed with Chute MavenTM as this would have
resulted in substantial additional time. For the transfer hood, the two belt speeds
experimentally tested were again used and for the 3 m/s belt speed, position B was
chosen for the hood position due to it being the ideal geometry with respect to the
angle of incidence of the incoming stream. Spherical and shaped particle simulations
were performed for each hood geometry, as summarised in Table 9.4.
Table 9.4 Experimental geometries simulated with EDEM
Geometry Particle Mass flow

representation rates (tph)
Vb = 2 m/s Spherical and shaped 31
Vb = 3 m/s, Pos B Spherical and shaped 38
224
On completion of the simulations, the data for the particle displacement and velocity
were extracted. The data exports in a different format to that of Chute MavenTM so some
additional post processing of the data was required to allow the existing Matlab
programs to be used. The average stream velocities from the EDEM simulations have
been graphed in Figure 9.14. The results show that for both the spherical and shaped
particles, the average stream velocities are identical once the initial transient behaviour,
at the impact point, has steadied.
Figure 9.14 EDEM simulation results for transfer hood geometries
9.2.4 Method Comparisons for Polyethylene Pellets

The individual results of the methods have been presented above, but to fully appreciate
the comparison between each method for each belt speed and transfer hood geometry,
Figure 9.15 to Figure 9.17 are presented. Both the low and high feed rates have been
plotted for the experimental, Roberts and Korzen methods. For the DEM, only the low
feed rate has been plotted for Chute MavenTM, due to the limitations as explained in
Section 6.2.7.2 and for EDEM, only select high mass flow rate simulations have been
presented in Figures 9.15 and 9.17.
225
Figure 9.15 Comparison of methods for a belt speed of 2 m/s
Figure 9.16 Comparison of methods for a belt speed of 3 m/s with hood position A
There is a general trend of over-prediction of the stream velocities through the transfer
hood compared to the experimentally determined stream velocities for the Roberts and
Korzen methods as well as for the DEM simulations from Chute MavenTM and EDEM.
Furthermore, the following observations have been made focussing on the hood exit
velocities:
- for the 2 m/s belt speed case, the Roberts method produced 9.6% and 18.1% over-
predictions for the low and high feed rates respectively for the hood exit velocities
compared to the experimental results;
226
Figure 9.17 Comparison of methods for a belt speed of 3 m/s with hood position B
- for the 3 m/s belt speed with the hood in position A, the Roberts method produced
13.2% and 30.2% over-predictions for the low and high feed rates respectively at
the 5 angular position (as the hood exit was obscured), compared to the
experimental results;
- for the 3 m/s belt speed with the hood in position B, the Roberts method under-
predicted the experimental hood exit velocity by 1% for the low feed rate while it
over-predicted the experimental hood exit velocity by 9.4% for the high feed rate;
- for the 2 m/s belt speed case, the Korzen method under-predicted the experimental
hood exit velocity by 0.1% for the low feed rate while it over-predicted the
experimental hood exit velocity by 7.5% for the high feed rate;
- for the 3 m/s belt speed with the hood in position A, the Korzen method produced
4.1% and 23.6% over-predictions for the low and high feed rates respectively at
the 5 angular position (as the hood exit was obscured), compared to the
experimental results;
- for the 3 m/s belt speed with the hood in position B, the Korzen method under-
predicted the experimental hood exit velocity by 10.1% for the low feed rate while
it over-predicted the experimental hood exit velocity by 2.0% for the high feed
rate;
- the Chute MavenTM DEM simulations undertaken to compare with the
experimental low mass flow rate tests all over-predicted the experimental hood
227
stream exit velocity, by 11.2% for the 2 m/s case, 6.3% for the 3 m/s belt speed
with the hood in position A (compared at the 5 angular position) and 2.4% for the
3 m/s belt speed with the hood in position B;
- the EDEM simulations showed a trend very similar to the results from the Chute
MavenTM simulations, however, the EDEM simulations were based on the high
mass flow rates so the angular starting position around the hood was larger. Even
so, the stream velocity at the exit of the hood matches identically with those
generated from Chute MavenTM.

HOOD

Preliminary experimental trials with the iron ore quickly showed that even though the
material was pre-screened to specified size ranges, a substantial amount of dust was still
present, possibly as a result of the method of transportation from the mine site. To
overcome potential occupational health and safety issues, dust extraction was installed
as detailed in Section 3.8; however, there was still significant fugitive dust in the air
when the conveyor transfer research facility was operating. Additional fans were
employed to keep dust away from expensive equipment and personnel working around
the conveyor transfer research facility.
The scope for the experimental test work was reduced from that carried out for the
polyethylene pellets. To minimise the impact of the continuing dust issues, only optimal
transfer hood positions are to be used, reducing the number of tests required.
A belt speed of 1 m/s was not tested due to the fact that for the polyethylene pellets this
was a low-speed condition, not requiring the use of the transfer hood. The belt speed
was set for 2 m/s and the transfer hood adjusted to obtain a low angle of incidence while
ensuring the flow exiting the hood was vertical. Additionally the transfer spoon was
aligned to receive an optimum flow, presented later in Section 11.3.1.
The Redlake X3 MotionPro high-speed digital video camera was once again used to
record the particle flow through the transfer hood, including the discharge of material
228
from the conveyor head pulley. Once testing commenced, it was clear that the capture of
the flow stream was going to be difficult due to the propagation of dust coming between
the video camera and the transfer hood. Making the situation worse was the dust settling
on the inner surface of the acrylic wings of the hood, see Figure 9.18. The decision was
made that footage would only be recorded at the conveyor discharge, the point of
impact with the hood and the exit of the hood. Attempts would also be made to record
the particle flow at a mid point around the hood.
During the post processing of the transfer hood video footage, it was found that the flow
at the impact point was visible, see Figure 9.19a, along with that at the exit. However,
due to the afore mentioned dust residue coating the wings of the transfer hood, the
footage obtained at both the 30 and 15 positions around the hood was not clear enough
to define individual particles and thus no analysis was possible. Figure 9.19b shows the
typical visibility at the 30 position and although some particles can just be made out,
when viewing successive frames in Image Pro Plus, it was impossible to accurately
track them.
On completion of the testing for the 2 m/s belt speed, the conveyor transfer was
reoriented for a 3 m/s belt speed. Again, the angle of incidence was optimised before
video footage was recorded. The dust being generated for the 3 m/s belt speed was
higher than that of the previous 2 m/s tests as evident in Figure 9.20. This additional
dust made recording of the flow stream impossible and so any further experimental
testing of the iron ore was abandoned.
The velocity data which was able to be extracted from the 2 m/s belt speed tests is
presented in Figure 9.21. As a consequence of only being able to track particles at the
point of impact and exit of the transfer hood, no particle velocity trend could be
produced through the hood. This will reduce the possible comparisons which can be
made with the analytical models and discrete element method simulations.
229
Figure 9.18 Iron ore dust build upon the transfer hood wings
(a) impact point, Vb = 2 m/s (b) 30 position around hood, Vb = 2 m/s
Figure 9.19 Iron ore particle flow through the transfer hood
230
Figure 9.20 Transfer hood setup for Vb = 3 m/s showing excessive dust
Figure 9.21 Average particle velocity in the transfer hood
9.3.2 Analytical Method Analysis

The data obtained from the experimental test program for iron ore has been used once
again to generate the results of the Roberts and Korzen analytical methods. Due to the
inability to obtain information for each angular position, estimations of the stream
thickness and stream widths have been made based on experimental observation.
231
9.3.2.1 Analytical Method of Roberts

The Roberts analytical method has been applied using the same method as described in
Section 9.2.2.1. The coefficient of equivalent friction, e, has been recalculated based
on the particle properties measured for iron ore and the ratio of the pressure acting on
the sides of the chute to the pressure acting on the bottom of the chute, Kv, has again
been estimated at 0.4 based on the fact that the height of the material stream is
substantially less than the width of the stream and also because there was no means of
directly measuring these pressures.
belt speed and hood position, the results of the Roberts analytical analysis are presented
in Figure 9.22. The plotted data shows a noticeable increase in stream velocity through
the hood for both the low and high feed rates for the 2 m/s belt speed, resulting in nearly
identical exit velocities.
Figure 9.22 Predicted stream velocity of iron ore through the

transfer hood using the Roberts method
9.3.2.2 Analytical Method of Korzen

The Korzen analytical method has been applied using the same method as described in
Section 9.2.2.2. The iron ore used in this testing is non-cohesive and the same
assumption has been made that the curved shearing surface can be replaced by a curved
232
chute of constant radius, resulting in the coefficient of wall friction, w, being used in
equation 9.4.
The cross sectional area of the material profile was extremely hard to measure for this
material and as a result, averaging has been applied to the observed height and width of
the product stream. As a consequence, an averaged cross sectional area at each angular
position through the hood has been calculated.
belt speed and hood position, the results of the Korzen analytical analysis are presented
in Figure 9.23. The data presented shows a significant increase in stream velocity
through the hood for both the low and high feed rates for the 2 m/s belt speed in the
same way as the Roberts method, with both sets of curves producing nearly identical
exit velocities.
Figure 9.23 Predicted stream velocity of iron ore through the

transfer hood using the Korzen method

The DEM simulations for the iron ore section are limited due to only one belt speed
being tested experimentally. As a result, the decision was made to produce simulations
for both the low and high mass flow rates for Chute MavenTM and EDEM. Spherical
particles have been used for the Chute MavenTM and EDEM simulations and shaped
233
representations have also been used in EDEM to investigate potential influences on the
stream velocity.

The restrain parameter, discussed in length previously, has been estimated again at 80%.
Due to the dusty nature of the iron ore, visual inspection of the flowing stream proved
difficult. The previous findings showed that the velocity of the stream itself was not
altered by the restrain value used, more so the integrity of the particle stream. Visually,
the stream integrity was good, thus deciding on 80% as a good approximation.
The coefficient of particle friction for polyethylene was not found to be a major
influence on the results, so an estimate of 0.5 was used for iron ore for all simulations.
The other parameters have been taken directly from Table 5.2.
Reviewing the results from the polyethylene pellet simulations, the effect of mass flow
rate on the stream velocity was negligible and has been assumed the case with iron ore
also. Simulations were therefore completed for the low mass flow rate only, i.e. 15.3 tph
for a belt speed of 2 m/s. Matlab was used to analyse the stream velocity and the results
are shown in Figure 9.24.
Figure 9.24 Chute MavenTM DEM simulation results for all mass flow rates
234

The setup of the simulations follows the same method as for the polyethylene pellets,
with the iron ore parameters coming from Tables 5.2, 5.4 and 5.5. Spherical particles
have been generated based on the equivalent volume diameter and two particles were
modelled in Pro-Engineer to be used in the shaped particle simulations, see Figure 9.25.
In the setting up of a particle factory within EDEM, there is an option to use a binary
input which allows two modelled particles to be selected. Each of these particles
contains multiple spherical particles, nine and six particles respectively and will
dramatically increase the time needed to complete any simulations.
Figure 9.25 Shaped representations of iron ore particles
As iron ore had not previously been tested with EDEM, a sensitivity analysis was
undertaken for the coefficient of rolling friction, but only for a belt speed of 2 m/s. A
simulation geometry was created to measure particle discharge velocity from the
conveyor belt for coefficients of rolling friction ranging from 0.05 to 0.30. The analysis
method followed that used during the EDEM trajectory investigations and the results
are presented in Table 9.5. For both the spherical and shaped simulations, the results
were nearly identical, with the analysed particle discharge velocities all being within 1%
of the belt speed. It was previously noted for polyethylene pellets that the lower the
coefficient of rolling friction was for the spherical particles, the more particles diverged
from the main stream. For this reason, a coefficient of rolling friction of 0.15 was
chosen for the spherical particle conveyor transfer simulations. For the shaped particle
simulations, a smaller coefficient of rolling friction was chosen, 0.05.
235
Table 9.5 Results of rolling friction sensitivity simulations
Spherical Shaped
Particle Particle
Coefficient Velocity at % of Velocity at % of
Belt Speed of Rolling Discharge Belt Discharge Belt
(m/s) Friction (m/s) Speed (m/s) Speed
2 0.05 1.989 99.46 1.995 99.73
2 0.10 1.995 99.76 1.996 99.79
2 0.15 1.996 99.81 1.995 99.73
2 0.20 1.995 99.73 1.995 99.75
2 0.25 1.994 99.69 1.995 99.77
2 0.30 1.994 99.71 1.994 99.69
The conveyor transfer simulations were then setup using the experimental positioning of
the transfer as a guide, to provide a direct comparison. An appropriate time step was
then chosen, based on the steady-state condition of the material flow through the
conveyor transfer, and the displacement and velocity data was exported for analysis.
Using the previously developed Matlab programs, the minimum, average and maximum
stream velocity was determined for each simulation and is presented in Figure 9.26.
The results show that there is very little difference between the spherical and shaped
particle simulations for either mass flow rate, with all four simulations resulting in the
same exit velocity from the transfer hood. The only noticeable difference is that the
shaped particle simulations contact the transfer hood approximately 2.5 before the
spherical particle simulations do. This is as a result of the way the particles pack on the
conveyor belt, with the shaped particles interlocking and filling more of the voids
between particles.
The EDEM simulation data presented here is part of a larger picture, that is, an extract
from the entire transfer simulations. The outputs provided in Section 11.3.3.2 consist of
the other main component of these simulations. As will be explained in detail in Section
11.3.3.2, an unusual observation was made with respect to a combination of the
coefficient of rolling friction and the shear modulus. The results that follow are as a
direct result of further investigations into this.
236
Figure 9.26 EDEM simulation results for low and high mass flow rates
Figure 9.27 shows the output of ten high mass flow rate simulations, five each for
spherical and shaped particles, as detailed in the accompanying legend. As well as the
initially defined coefficient of rolling friction of 0.15 for spherical particles, 0.05 was
also investigated, along with various shear modulus values.
Figure 9.27 Comparison of EDEM simulation outputs for varying coefficient of

rolling friction and shear modulus
The particle volume for the spherical particles was based on the volume of the sphere
used to define the particles within the simulation. For the shaped particles, the initially
defined coefficient of rolling friction of 0.05 was investigated, along with 0.15. The
237
particle volume was determined using two methods. Firstly, the volume was based on
the volume of the particle template imported from CAD, signified by the legend entry
Vol-template in Figure 9.27. The second method determines the volume of the particle
based on the surface of the clustered particles used to generate the shaped particles. This
volume is representative of only the component of the imported template which is
occupied by solid spheres and is signified by the legend entry Vol-surface.
Summarising the results of the investigation found that regardless of the coefficient of
rolling friction used or the shear modulus value, the point of impact of the particle with
the hood remained constant and the subsequent velocities through the hood were
identical, ultimately all predicting the same particle exit velocity from the hood. The
one minor exception is the case where the template was used to determine the particle
volume, where the point of impact was a couple of degrees lower than the rest.
In the following comparisons, the originally determined average particle velocities,

shown in Figure 9.26 will continue to be used in light of the minimal differences
obtained in figure 9.27.
9.3.4 Method Comparisons for Iron ore

The results from the above sections have been combined to provide a concise
comparison of the average iron ore stream velocities through the conveyor transfer, as
shown in Figure 9.28. Both the low and high feed rates have been plotted for the
experimental, Roberts and Korzen methods. For the DEM, only the low feed rate has
been plotted for Chute MavenTM, due to the limitations as explained in Section 6.2.7.2
and for EDEM, both low and high mass flow rates for spherical and shaped particle
simulations have been presented.
All the methods used to predict the experimental stream velocity through the transfer
hood compare favourably. For the low mass flow rate, the Roberts method is an exact
match, however, over-predicts the exit velocity from the hood by a small amount
(4.7%). The Korzen method under-predicts both mass flow rates by 10.4% and 5.1%
respectively. The Chute MavenTM DEM output for the low mass flow rate initially
predicts a lower angle of impact in the transfer hood but goes on to predict the same exit
velocity found experimentally. The low mass flow rate EDEM simulations also start at a
238
slightly lower angle of impact, but have the same exit velocity. The spherical particle
high mass flow rate EDEM simulation shows a near identical angle of impact, although
this is in the transient flow zone; however, the shaped particle simulation starts at a
lower angle of impact, which was discussed as being due to the packing nature of the
shaped particles.
Figure 9.28 Comparison of methods for a belt speed of 2 m/s
9.4 DISCUSSION
Experimentally, two conveyor belt speeds have been investigated (2 m/s and 3 m/s)
using two test materials. A number of transfer hood positions were used to best
transition with the trajectory stream leaving the feed conveyor by producing as low an
angle of incidence as possible. Some less than ideal transfer hood positions were also
trialled to investigate the effect on the material stream. The result of these non-ideal
hood positions was that material impacted at a larger angle, causing more material to
spray laterally from the stream rather than following the natural vertical curve of the
hood. In saying this, there was very little change in the particle stream velocity through
the transfer hood once past the initial transient zone. For the polyethylene pellets,
experimental data was captured at 5 degree increments around the hood but for the iron
ore, there was significant dust generated which meant that only data at the point of
impact and exit from the hood could be captured.
239
The analytical methods of Roberts and Korzen provide reasonable predictions of the
particle stream velocity through the hood for both materials tested experimentally.
There was a better agreement between the analytical methods and the experimental
results for the iron ore, where the Roberts method provided an exact match for the
particle stream velocity at the exit of the transfer hood for the low mass flow rate. Some
of this variation could be put down to the fact that neither of these methods can account
for the non-spherical nature of the particles.
The DEM simulations produced by Chute MavenTM as part of this research showed that
regardless of the mass flow rate for a given belt speed and/or hood position, the particle
stream velocity at the hood exit velocity was identical. This fact means that significant
simulation time can be saved by only investigating low mass flow rates if only the
particle velocity at the exit of the transfer hood is required. But by doing so, important
interactions of the material with the transfer hood could be overlooked.
The Chute MavenTM DEM software only allows for spherical particles to be simulated.
With the inability to simulate the true particle shape, it was expected that there would be
differences between the DEM and experimental results. This proved to be less than first
expected, with some of the results quite close to the experimental particle velocities.
The EDEM simulations provided the chance to investigate the effect of shape. Spherical
simulations were completed which compared well with those from Chute MavenTM.
When the shaped particles were simulated, it was found that there was very little
difference in the flow behaviour through the hood and the particle stream exit velocities
were an exact match for the spherical equivalents. The main drawback to completing
simulations with shaped particles is the fact that simulations can take substantially more
time to complete depending on the number of spherical particles clustered together.
An advantage of the DEM simulations over the analytical models is the fact that the
transient flow region, at the point of impact with the transfer hood, can be modelled.
Another advantage is that DEM can provide a three dimensional perspective of the flow
behaviour whereas the analytical models rely solely on a two dimensional approach.
240
In the experimental testing of polyethylene pellets, the transfer hood was positioned in a
less that ideal location for one set of 3 m/s belt speed tests. As previously mentioned,
the dynamic behaviour of the particles could clearly be seen, however, for the analytical
methods, this could not be modelled. The analytical models assume that the material is
flowing smoothly around the transfer hood from the angle of impact, not accounting for
the possible variation in angle of incidence. Counter to this are the results from the
DEM simulations. They do take this angle of incidence into account and as has been
shown, can replicate the experimental observations quite well.
Summing up, for a given transfer hood position and belt speed, there are a range of
stream velocities resulting from the analysis of the experimental, analytical and DEM
methods. This has a flow on effect when the material stream enters the freefall zone
between the transfer hood and spoon, as will be discussed in the following chapter.
241
CHAPTER 10 PARTICLE FREEFALL
Chapter
10
PARTICLE FREEFALL
242
10.1 INTRODUCTION
When material flows through a conveyor transfer, it generally undergoes some degree
of freefall between the exit of the hood and the inlet of the spoon. There are two main
analytical approaches that can be taken to predict the freefall velocity, both explained in
Section 2.7. The first, when air resistance is neglected, relies on uniformly accelerated
linear motion and the second, includes the influence of air resistance, which also
requires the determination of the terminal velocity of the particles. Discrete element
simulations can also be produced to investigate particle freefall if air drag is neglected;
however, coupling with computational fluid dynamics (CFD) software is required when
air drag is to be considered.
10.2 PARTICLE FREEFALL VELOCITY OF POLYETHYLENE PELLETS
10.2.1 Experimental Measurement of Freefall and Terminal Velocity

The measurement of the particle freefall velocity through the conveyor transfer was
achieved by removing the transfer spoon and conveying material through the transfer
hood at belt speeds of 2 m/s, 3 m/s and 4 m/s with two mass flow rates, as shown in
Table 10.1.
Table 10.1 Experimental freefall tests
Belt Speed Mass flow Rate

(m/s) (tph)
TEST 1 2 1
TEST 2 2 9
TEST 3 3 1
TEST 4 3 9
TEST 5 4 9
As previously explained, the exit of the hood is aligned to ensure material exits
vertically. Figure 10.1 shows the arrangement used to record the particle freefall
velocity. A vertical scale with 100 mm increments was attached to the exit of the hood
with a total length of 800 mm.
243
Figure 10.1 Experimental particle freefall setup
The Redlake MotionPro X3 high-speed digital video camera was used to record the
particle velocity through the freefall zone and the video footage was subsequently
analysed with Image Pro Plus. The results from the five tests are presented in Figure
10.2. The experimental results are presented with the initial vertical position being
assigned as the origin. Using the equation for uniformly accelerated linear motion to
determine final velocity, as presented in equation 2.92, the initial experimental velocity
(particle velocity at the exit of the hood) was used to predict the final velocity for each
experimental drop height. These values have also been added to the results presented in
Figure 10.2 as a comparison. It can be seen that there is a good overall agreement
between the experimentally measured freefall velocities and those predicted.
An interesting observation from the results shown in Figure 10.2 is that regardless of the
belt speed, hence, the variation in initial velocity of the material, the experimental
freefall results are nearly identical for the five tests analysed, as shown in Figure 10.3.
There also appears to be some transient behaviour present soon after the material leaves
244
the transfer hood, evident by the slightly wider scatter of velocities, but becomes more
controlled the further the material freefalls.
(a) (b)
(c) (d)
(e)
Figure 10.2 Experimental and theoretical freefall results for polyethylene pellets,
(a) Vb = 2 m/s, ms = 1 tph, (b) Vb = 2 m/s, ms = 9 tph, (c) Vb = 3 m/s, ms = 1 tph,
(d) Vb = 3 m/s, ms = 9 tph, (e) Vb = 4 m/s, ms = 9 tph
245
Figure 10.3 Comparison of experimental freefall velocity results
Further analysis of the experimental data was performed using the terminal velocity
equation, equation 2.93. This provided an experimental estimate of the terminal velocity
of polyethylene pellets and is presented in Table 10.2.
Table 10.2 Estimates of the experimental terminal velocity of polyethylene pellets
Terminal Velocity
(m/s)
Test 1 6.561
Test 2 6.534
Test 3 6.616
Test 4 6.764
Test 5 6.796
Average 6.654
10.2.2 Analytically Determining Terminal Velocity

The analytical methods of determining terminal velocity are described in Section 2.7.
The sphericity of particles is not addressed by all methods, so in those cases where only
spherical particles are considered, the sphericity factor (Wypych, 1993) has been
adopted to provide consistency, the results summarised in Table 10.3.
246
Table 10.3 Predicted terminal velocity of polyethylene pellets
Terminal velocity Terminal velocity

based on a adjusted for
spherical particle sphericity
(m/s) (m/s)
Wypych (1993) 10.262 6.363
Tran-Cong et al. (2004) Reynolds Number outside correct range
Turton and Levenspiel (1986)* 10.950 6.789
Turton and Clark (1987) * 9.996 6.198
Swamee and Ojha (1991) * 11.040 6.845
Majumder and Barnwal (2004) * 10.758 6.670
Haider and Levenspiel (1989) * 10.896 6.756
Haider and Levenspiel (1989) ** - 7.126
Geldart (1990) ** - 6.646
Brown and Lawler (2003) * 10.637 6.595
Karamanev (1996) * 10.903 6.760
Average 10.680 6.675
* method based on spherical particles, method of Wypych (1993) used to adjust for particle sphericity
** method only determines terminal velocity based on particle sphericity
10.2.3 Chute MavenTM DEM Simulation of Terminal Velocity

To investigate the freefall velocity of particles generated from Chute MavenTM, various
simulations were performed, see Table 10.4. The simulations consisted of a room with
varying floor height and an injection box, from which a range of low mass flow rates
were produced. The vertical particle velocities were then extracted directly from Chute
MavenTM and combined to form Figure 10.4.
Table 10.4 Range of Chute MavenTM DEM simulations performed
Drop Height Mass flow Rate

TEST (m) (tph)
DEM 1 0.8 0.005
DEM 2 0.8 0.01
DEM 3 0.8 0.02
DEM 4 0.8 0.05
DEM 5 0.8 0.1
DEM 6 8 0.005
DEM 7 8 0.01
DEM 8 8 0.02
DEM 9 12 0.005
DEM 10 20 0.005
247
Equation 2.91 was again used to provide the theoretical particle velocity based on drop
height and is also displayed on Figure 10.4. Figure 10.5 represents the same data,
however, only shows the first 1m of drop height, which is more relevant to the
experimental testing performed. It is clear that the DEM simulation is an exact match
for the theoretical particle freefall velocity.
Figure 10.4 Particle freefall velocity obtained from Chute MavenTM simulations
Figure 10.5 Particle freefall velocity obtained from Chute MavenTM simulations
The transfer hood DEM simulations performed in Chapter 9 also included the freefall
zone and the transfer spoon. The particle velocity data from the freefall zone was
extracted from the experimental tests for;
- Vb = 2 m/s, 2 tph
- Vb = 3 m/s, position A, 2 tph
248
- Vb = 3 m/s, position B, 10 tph
The data was then plotted versus the relative drop height of the particles, where 0 m is
the outlet of the transfer hood and has a corresponding particle velocity equal to the
velocity at the exit of the transfer hood and the results are presented in Figure 10.6.
(a) Vb = 2 m/s, 2 tph
(b) Vb = 3 m/s, position A, 2 tph
(c) Vb = 3 m/s, position B, 10 tph

Figure 10.6 Chute MavenTM DEM freefall data from conveyor transfer simulations
249
Again, the theoretical particle freefall velocity was calculated from equation 2.91 and
plotted on the same graph for comparison. As with the previous freefall DEM
simulations, it can be seen that the theoretical freefall velocity matches the simulated
data exactly. In the experimental tests, there is the expectation that there will be a minor
air resistance influence present which would act on the particles and reduce their overall
velocity. This is a factor which cannot be accounted for by Chute MavenTM as there is
no facility to couple the analysis with computational fluid dynamics. Even though this is
the case, the outputs from Chute MavenTM have produced identical outputs to those
found experimentally, which seems to indicate that there is very little, if any, air
resistance affecting the flow of the particles. This also implies that the freefall
calculations within Chute MavenTM are based on uniformly accelerated linear motion.
10.3 PARTICLE FREEFALL VELOCITY OF IRON ORE

The dusty nature of the iron ore has been presented previously. This resulted in the
experimental determination of the freefall velocity of iron ore not being possible in the
conveyor transfer. An alternative method was devised, where a small square acrylic
hopper was filled with iron ore and allowed to freefall through a height of 800 mm, see
Figure 10.7. The outlet velocity of the iron ore was set to 0 m/s and the Redlake X3
MotionPro high-speed digital video camera was used to capture the particle freefall at
100 mm increments which was then analysed to produce the results shown in Figure
10.8.
Further analysis of the experimental data was carried out using the terminal velocity
equation, equation 2.93. This provided an experimental estimate for the terminal
velocity of iron ore and was found to be V = 5.278 m/s.
250
Figure 10.7 Experimental setup to measure the freefall velocity of iron ore
Figure 10.8 Experimental and theoretical freefall results for iron ore

The terminal velocity of the iron ore test material has again been approximated using
the methods described in Section 2.7. For the methods which do not include sphericity,
the sphericity factor (Wypych, 1993) has again been adopted to provide consistency.
The results of these comparisons are shown in Table 10.5.
251
Table 10.5 Predicted terminal velocity of iron ore

(m/s) (m/s)
Wypych (1993) 23.495 11.043
Turton and Clark (1987) * 22.818 10.725
Swamee and Ojha (1991) * 23.695 11.137
Geldart (1990) ** - 13.152
Brown and Lawler (2003) * 21.859 10.274
Karamanev (1996) * 23.876 11.222
Average 23.382 11.429

As reported in Section 10.2.3, the assumption has been made that the freefall velocity of
particles determined by Chute MavenTM are based on uniformly accelerated linear
motion. As this is not dependant on any particle characteristics, such as mass, there will
be no difference in freefall velocity determined for any range of materials. For this
reason, no DEM simulations have been performed for this section.
10.4 PARTICLE FREEFALL VELOCITY OF CORN

Although corn was not tested through the conveyor transfer, the experimental particle
freefall velocity and subsequent terminal velocity has been completed for completeness.
The same experimental procedure as for the iron ore freefall velocity determination has
been used. Once again, the exit velocity from the hopper was set to 0 m/s and high-
speed video footage at 100 mm increments of vertical fall was recorded.
252
Figure 10.9 Experimental and theoretical freefall results for corn
Further analysis of the experimental data was carried out using the terminal velocity
equation, equation 2.93. This provided an experimental estimate for the terminal
velocity of iron ore and was found to be V = 5.436 m/s.

The terminal velocity of corn has again been approximated using the methods described
in Section 2.7 and is summarised in Table 10.6.
Table 10.6 Predicted terminal velocity of corn

(m/s) (m/s)
Wypych (1993) 17.988 7.195
Turton and Clark (1987) * 17.123 6.849
Swamee and Ojha (1991) * 17.525 7.010
Geldart (1990) ** - 6.646
Brown and Lawler (2003) * 15.831 6.332
Karamanev (1996) * 17.708 7.083
Average 17.361 7.067
253

As reported in Section 10.2.3, the assumption has been made that the freefall velocity of
particles determined by Chute MavenTM are based on uniformly accelerated linear
motion. As this is not dependent on any particle characteristics, such as mass, there will
be no difference in freefall velocity determined for any range of materials. For this
reason, no DEM simulations have been performed for this section.
10.5 DISCUSSION
The particle freefall velocity of polyethylene pellets was analysed from experimentally
obtained video footage when flowing through the conveyor transfer, between the hood
and the spoon. The data compared very well to that predicted from the equation for
uniformly accelerated linear motion, indicating that there is minimal deviation due to air
resistance. Experimental investigation of the freefall velocity of iron ore and corn was
performed by an alternative method as iron ore generated too much dust in the transfer
zone and corn was not tested through the conveyor transfer. The results of this analysis
also showed that the experimental results followed the equation for uniformly
accelerated linear motion.
The experimental results for the three test materials were also used to estimate the
terminal velocity of each material. The results for the polyethylene pellets were very
similar, however, for iron ore and corn there was a noticeable deviation. It is believed
that the prediction of the terminal velocity is inaccurate due to the relatively short drop
heights the particle velocities were measured within and that the polyethylene pellet
results were similar by chance only. The methods explained in Section 2.7 rely heavily
of the particle density and small adjustments can alter the predictions greatly.
Discrete element modelling simulations were performed with Chute MavenTM to

determine the freefall velocity of polyethylene pellets and were found to almost
identically mirror the results from the equation for uniformly accelerated linear motion.
This finding was somewhat expected as there is no provision in Chute MavenTM to
account for air drag via coupling with computational fluid dynamics software.
254
CHAPTER 11 CONVEYOR TRANSFER SPOON ANALYSIS
Chapter
11
CONVEYOR TRANSFER SPOON ANALYSIS
255
11.1 INTRODUCTION
The lower section of a conveyor transfer needs to be able to capture the incoming flow
of material and direct it in the direction of the receiving conveyor. In this research, the
lower transfer section is made up of a transfer spoon. The material will impact with the
spoon based on the geometry and positioning of the upper section of the conveyor
transfer, in this case, the transfer hood which was detailed in Chapter 9, and also the
zone in which the material underwent a period of freefall, as described in Chapter 10.
This chapter will detail the experimental investigation of particle flow through a transfer
spoon and the corresponding analytical modelling and discrete element simulation of
the particle flow, finally providing comparisons between the methods.
11.2 FLOW OF POLYETHYLENE PELLETS THROUGH A TRANSFER

SPOON

The conveyor transfer spoon has been constructed of acrylic to provide the ability to
record a variety of material flow characteristics with both high-speed video and digital
still cameras, see Figure 11.1. The Redlake X3 MotionPro high-speed digital video
camera has been used to capture the particle flow through the spoon.
Figure 11.1 Detail of the transfer spoon
256
Figure 11.1 shows , the angle that the top surface of the transfer spoon makes with the
horizontal, which varies throughout the experimental testing, as shown in Table 11.1.
The transfer spoon is a composite of two radii, and at the exit of the spoon is a
converging section to centrally concentrate the flow before feeding onto the receiving
conveyor. The spoon is also lined with 6 mm Polystone Ultra to reduce noise, chute
wear and frictional losses. The first spoon radius has been divided into incremental
steps of 2.5 while the second radius has steps of 5 signifying the locations where the
particle velocity will be analysed. These angles are measured relative to the origin of
each respective radius as points of interest along the spoon where experimental data is
analysed. The exit angle of the spoon, with respect to the horizontal, for a given transfer
spoon geometry will be + 27 + 38. When plotting the analysed velocities, whether
for the experimental, analytical or discrete element methods, the angular positions from
the second radius follow on directly from those of the first radius even though there are
two separate origins. The point at which the change in radius occurs will be evident in
the results presented for both the analytical and discrete element methods as there is a
distinct change in the particle velocity curve profile. To allow direct comparisons
between all three analysis methods (i.e. experimental, analytical and DEM methods), a
low experimental mass flow rate was selected. Table 11.1 summarises the belt speeds
and corresponding feed rates used. Initially, a feed rate of 2 tonnes per hour (tph) was
used for the 3 m/s belt speed, however, it was found that the stream thickness through
the transfer spoon was too thin to allow accurate measurement, hence the increase to 10
tph for the subsequent 3 m/s spoon geometries.
Table 11.1 Product feed rates used in experimental tests
Belt Speed, Low Feed Rate, Initial angle of

Vb ms spoon
(m/s) (tph) with horizontal,
()
1 2 17
2 2 14.5
3 position A 2 9.5
3 position B 10 19.5
3 position C 10 6.5
257
The 1 m/s belt speed results in low-speed conveying conditions, i.e. the material wraps
around the head pulley before discharging. Due to the angle of wrap, there was no need
to divert the material using a conveyor hood, thus the material underwent freefall from
the head pulley until impacting with the spoon. The 2 m/s and 3 m/s belt speeds both
result in high-speed conveying conditions, i.e. the material discharges from the point of
tangency between the conveyor belt and the head pulley and is thus directed to the
transfer spoon by first flowing around the transfer hood.
One spoon position was used for the 1 m/s and 2 m/s belt speeds and the angle of
incidence of the incoming material stream was kept relatively small to maximise the
particle velocity through the transfer spoon. For the 3 m/s belt speed, three spoon
positions were used, partly due to the positioning of the transfer hood above and also to
investigate the influence of the angle of incidence on the material flow through the
spoon. Figure 11.2 shows a snapshot of the steady-state flow through each of the
transfer spoon geometries.
A number of key observations can be made from Figure 11.2, as summarised below.
For the 1 m/s belt speed, there was no transfer hood used to direct the flow to the
spoon, therefore the incoming stream is not vertical when impacting on the
transfer spoon, instead the angle of incidence is a function of the resulting
conveyor trajectory.
Spoon position B for the 3 m/s belt speed has been positioned to create a larger,
non-ideal, angle of incidence. This caused a slowing of the material stream,
resulting in a build-up of material at the spoon exit.
Experimental analysis was not possible for the entire spoon due to the material
flow in the converging section not being perpendicular to the video camera.
The velocity exiting the spoon was also unable to be measured as the outlet was
below the upper surface of the conveyor body.
The particle velocity was determined using the Image Pro Plus software package. Using
the manual tracking feature, numerous particles were tracked at each incremental
location around the spoon by selecting the centroid of a particle in each frame of the
video footage. The results can then be exported for further analysis.
258
Figure 11.2 Material flow through the conveyor spoon

(a) Vb = 1 m/s and ms = 2 tph, (b) Vb = 2 m/s and ms = 2 tph,
(c) Vb = 3 m/s position A ms = 2 tph, (d) Vb = 3 m/s position B ms = 10 tph,
(e) Vb = 3 m/s position C ms = 10 tph,
The velocity profiles for the low and high mass flow rates tested experimentally are
presented in Figure 11.3. These results show that there is quite substantial variation in
the particle velocities throughout each of the spoon geometries. For the low mass flow
rates, except that of spoon position B for the 3 m/s belt speed, there is an increasing
velocity trend as the material flows through the spoon. In most cases the measured
velocities are much greater than the initial belt speeds due to both the acceleration of the
product stream through the transfer hood and also the free-fall section. There is a
noticeable difference in the trend for spoon position B for the 3 m/s belt speed, where
259
the influence of additional material dramatically reduced the particle velocity through
the spoon as a build up of material occurs at the outlet converging section for the high
mass flow rate.
Figure 11.3 (a) Average experimental particle velocities for low mass flow rates
Figure 11.3 (b) Average experimental particle velocities for high mass flow rates
11.2.2 Analytical Method Analysis of Roberts

Predicting the flow through a conveyor spoon is possible using the analytical chute-flow
model (Roberts, 2003). The method of Roberts is a widely accepted approach to
predicting the stream velocity for granular cohesionless materials, and the force diagram
260
is presented in Figure 11.4. Parameters such as initial particle velocity, spoon impact
angle and the initial and average height and width of the particle stream are required in
the analysis. If the transfer spoon had not already been constructed, estimates of these
parameters would have been required. However, these values can be extracted directly
from the experimental test program.
Figure 11.4 Force diagram for the curved chute
On occasion, the material flow needs to be converged and there is provision for this
supplied by Roberts (1999; 2003). As the spoon used in this experimental testing has a
converging section before discharging to the receiving conveyor, this should also be
incorporated. The effect of incorporating the converging analysis to the analytical
method was applied for the two extremes tested experimentally and resulted in
negligible velocity reduction in the converging section of the transfer spoon; most likely
due to the relatively thin stream thicknesses not substantially altering the coefficient of
equivalent friction, see Figures 11.5a and 11.5b. For this reason, the application of the
Roberts analytical method has been performed with the assumption that the spoon is of
constant width throughout.
A coefficient of equivalent friction, e, is used, which incorporates the particle wall

friction, the stream cross-section and the internal shear of the bulk solid and is assumed
to be an averaged constant as the stream thickness is comparatively low, see equation
11.1. The ratio of the pressure acting on the sides of the chute to the pressure acting on
261
the bottom of the chute, Kv, is generally assumed to be a value between 0.4 and 0.6
according to Roberts (1999; 2003). There were no means of direct measurement for
these pressures and as such, an estimate of 0.4 has been used based on the height of the
material stream being substantially less than the corresponding width of the stream.
The stream velocity at any given angular position through the spoon can then be found
using equation 11.2, by first determining the constant of integration, K, by substitution
of the initial conditions, v = v0 and = 0. The initial velocity used in this analysis has
been taken directly from the experimental velocity analysis. The predicted velocity
profiles are presented in Figure 11.5. The starting velocity of each predicted velocity
path is based on the experimental results, allowing a direct comparison to the
corresponding experimentally tested belt speeds and spoon geometries.
K v v0 H 0
e = w 1 + 11.1
vB
2 gR
v=
4 e + 1
2 ( )
1 2 e2 sin + 3e cos + Ke 2 e 11.2
Figure 11.5 (a) Predicted average stream velocity through the spoon by Roberts
method for low mass flow rates
262
Figure 11.5 (b) Predicted average stream velocity through the spoon by Roberts
method for high mass flow rates

As with the transfer hood analyses, Chute MavenTM has been predominantly used but
EDEM has also been used for some key simulations. EDEM will also be used to
investigate the shape effect of the particles within the simulations.

The Chute MavenTM simulations performed in Chapter 9 were for the complete
conveyor transfer, however, only the transfer hood data was extracted for analysis. For
the transfer spoon the particle velocity data has now been extracted for use in this
section. As was the case with the transfer hood analysis, the restrain parameter has been
set at 80%.
Five DEM simulations were prepared using the product feed rates shown in Table 11.1,
to allow direct comparisons with both the experimental and Roberts analytical method
results. Two examples of the particle displacements from DEM simulations are
presented in Figure 11.6, the X and Y displacements are with respect to the central axis
of the head pulley of the feed conveyor. Figure 11.6a shows the particle flow through
the spoon for a belt speed of 1 m/s and it is clear that the angle of incidence is quite low,
partly as a consequence of the profile of the conveyor trajectory stream leaving the feed
conveyor, resulting in a smooth thin stream of material flowing through the spoon. In
263
contrast is the result shown in Figure 11.6b. Here, for a belt speed of 3 m/s and the
spoon in position B, it can be seen that the non-ideal spoon geometry is causing a
substantial amount of particle bounce directly after the impact point mainly due to the
large angle of incidence, however, there is still material sliding through the spoon as
expected.
(a) Vb = 1 m/s, ms = 2 tph (b) Vb = 3 m/s, position B, ms = 10 tph
Figure 11.6 Two examples of the DEM simulation outputs showing both
good and bad flow trends
The results of the five DEM simulations were analysed at every 1 angular increment
around the spoon to determine the average particle velocities, shown in Figure 11.7. The
following observations were then made.
The position of the spoon for each of the five test locations shown in Figure 11.2
result in different locations for the impact of the material stream, this is shown in
Figure 11.7 by the variation in inlet angle onto the spoon (with respect to the
horizontal).
By altering the position of the spoon, the exit angle of the material flow changes
with respect to the horizontal. As shown in Figure 11.2, position B for the 3 m/s
belt speed was set to investigate the effect of a non-ideal or poor spoon
position. The resulting average stream velocity for this simulation showed a
substantially lower velocity to the other simulations.
Also evident for this simulation is the dramatic drop in velocity towards the exit
of the spoon. This corresponds with the converging section where material
264
becomes increasingly constrained between the walls of the spoon and as a result
slows the material flow due to increasing friction and increased lateral velocity
of the particles.
The results displayed in Figure 11.7 show an over-prediction to both the experimental
and analytical analyses (shown in Figures 11.3 and 11.5), which will be detailed in
Section 11.2.4 in more detail. An alternative method of DEM simulation was devised,
where only flow the transfer spoon was simulated. To achieve this, the experimental
particle velocity at the at the point of impact with the spoon was extracted and along
with the freefall data provided in Chapter 10, an injection box was positioned at a pre-
determined vertical height above the transfer spoon. Particles would then be generated
in each simulation and freefall directly onto the spoon without any pre-existing
influences from the transfer hood. In the simulations, the particle velocity at the point of
impact with the transfer spoon should now be almost identical to that found
experimentally, allowing for a more accurate comparison of methods. It should also be
noted that no simulation was performed for the 1 m/s belt speed as the particle stream
was not falling vertically, as a result of the discharge from the head pulley. The results
of the DEM simulations, which correspond to those shown in Figure 11.7, are displayed
in Figure 11.8.
These results show a noticeable reduction in the average stream velocity through the
spoon when compared with the results in Figure 11.7. This is as a direct result of
adjusting the injection box at a suitable vertical height above the transfer spoon so that
the particle velocity at the point of impact, matches that determined experimentally. It
has also been observed that at the point of impact, and shortly after, there is a larger
region of transient behaviour before steady flow is achieved. Some of these transients
were observed experimentally but not with the analytical method predictions.
265
Figure 11.7 Simulation results for all experimental belt speeds and spoon geometries
from the Chute MavenTM transfer hood and spoon simulations
Figure 11.8 Simulation results for all experimental belt speeds and spoon geometries
from the Chute MavenTM transfer spoon simulations

As previously mentioned, EDEM has also been used to produce additional simulations,
specifically for high mass flow rates matching those tested experimentally. Only three
experimental geometries were simulated, as summarised in Table 11.2. In the case of
the 3 m/s belt speed, the spoon geometry which generated the best angle of incidence
has been simulated. These spoon simulations have been extracted from the overall
simulation of the complete conveyor transfer, in the same manner that the Chute
266
MavenTM spoon simulations were extracted to generate the results of Figure 11.7. This
means that the outputs include the influences of the trajectory, hood flow and freefall
components. No EDEM simulations were performed for just the transfer spoon. The
geometries summarised in Table 11.2 were performed for both spherical and shaped
particles, as with the transfer hood, described in Section 9.2.3.2.
Table 11.2 Product feed rates used in EDEM simulations
Belt Speed, Low Feed Rate, Initial angle of

Vb ms spoon
(m/s) (tph) with horizontal,
()
1 19 17
2 31 14.5
3 position C 37.8 19.5
Once the simulations were completed, the data for particle displacement and velocity
were exported for analysis with Matlab. An example of the M-file used to determine the
average stream velocity through the spoon can be found in Appendix D, the results are
shown in Figure 11.9. These results show that there is very little difference between the
spherical and shaped particle simulations for each belt speed tested.
Figure 11.9 EDEM simulation results for transfer spoon geometries
267
Each set of results also shows that for the high mass flow rate, there is a pronounced
drop in the average stream velocity part way through the spoon. This point does not
coincide directly with the converging section of the spoon but in all likelihood is as a
result of the flow stream being restricted as material reaches the converging section,
causing a restriction to flow and thus a slowing of the stream. Towards the end of each
set of simulation results, it can be seen that there is a minor increase in the stream
velocity just before the point of discharge from the spoon.
11.2.4 Method Comparisons

At this point, every attempt has been made to make the comparisons as valid as possible
between methods, however, there are some points that need to be emphasised.
The experimental spoon analysis is an extract from the complete transfer
analysis (i.e. discharge, hood flow, freefall and spoon flow).
The experimental results have only been obtained up to the location on the spoon
where the converging section commences. As a result, comparisons between the
experimental results and those of the Roberts (2003) and DEM methods can only
be made for this region, whereas comparisons for the entire spoon can be made
between the Roberts (2003) and DEM methods.
The Roberts (2003) method has been applied to the spoon using the initial
experimental particle velocity of the stream entering the spoon, to allow a direct
comparison with the experimental results.
The initial Chute MavenTM and the EDEM simulations of the spoon, shown in
Figure 11.7 and 11.9 respectively, are an extract from the complete transfer
analysis in the same way that the experimental results were obtained. This means
that direct comparisons between the Roberts (2003) method and the Chute
MavenTM and EDEM simulations cannot be made as the initial stream velocity
entering the spoon for the DEM simulations has been influenced by the
acceleration of the stream through both the transfer hood and freefall zone.
Chute MavenTM was further used to generate spoon only simulations, providing
comparisons to the experimental and Roberts (2003) methods.
The initial DEM simulations from Chute MavenTM and EDEM all show over-
predicted particle velocities compared to the experimental results due to
acceleration of the particles in both the hood and free-fall zones.
268
Taking the above points into account, the results of the velocity analyses have been
combined for each belt speed and subsequent transfer spoon geometry and are presented
in Figure 11.10, with the following observations being made.
On all figures, the angular position of the converging section of the transfer
spoon has been highlighted by a vertical line.
Figure 11.10a shows that for the low mass flow rate, the Roberts method initially
diverges from the experimental velocity profile before converging to a velocity
similar to the experimental data at the point where the spoon converges.
However, for the high mass flow rate, the Roberts method over-predicts the
experimental results substantially.
Figures 11.10b to 11.10e show a better fit between the Roberts method and the
experimental velocities for the low mass flow rates, but there is a degree of
divergence present at, or near, the change in radius within the spoon.
Figures 11.10c, 11.10d and 11.10e all show significant experimental velocity
fluctuations along the first few degrees of the spoon, due to the particle stream
impacting on the spoon.
The initial Chute MavenTM DEM simulations over-predict the experimental
particle velocities through the spoon in all cases. This is due to the prior
acceleration of the stream through the hood and freefall zones. Of special note is
the change in the trend for the DEM analysis of Figures 11.10d and 11.10e. The
velocity drops substantially in Figure 11.10d, mainly due to the poor positioning
of the spoon, causing a build up of material at the spoon exit.
Chute MavenTM simulations of only the spoon gave much better comparisons to
the low mass flow rate experimental results and Roberts method.
In Figures 11.10d and 11.10e there is a noticeable difference between the
experimental particle velocity trend and that produced using the Roberts method.
This is most likely due to the Roberts method not being able to account for
transient behaviour within the spoon. Section 11.2.2 discussed the fact that the
Roberts method could also be adapted to account for converging spoon sections,
however, investigations showed negligible differences.
269
(a) Vb = 1 m/s
(b) Vb = 2 m/s
(c) Vb = 3 m/s, position A
(d) Vb = 3 m/s, position B

Figure 11.10 Comparison of the three analysis methods for each belt speed and
spoon geometry
270
(e) Vb = 3 m/s, position C

Figure 11.10 (cont.) Comparison of the three analysis methods for each belt speed
and spoon geometry

SPOON

The scope of the experimental test work followed directly from that completed for the
transfer hood investigations. Due to the large degree of dust being generated, the
transfer spoon was aligned to the optimal position for each belt speed case,
minimising the number of tests required. A belt speed of 2 m/s was initially tested with
both low and high mass flow rates so that when combined with the transfer hood results,
the complete flow through the conveyor transfer could be quantified. As with the
transfer hood tests, the dust being generated within the spoon meant that very little
useful information could be gathered, even with the dust extraction in operation, see
Figure 11.11. Using the Redlake X3 MotionPro high-speed digital video camera,
footage was captured at the point of impact on the spoon but was impossible throughout
the spoon for the low mass flow rate. For the high mass flow rate, even more dust was
generated but video footage was obtained at point N, which corresponded to an angular
position 49 from the horizontal. The limited results are shown in Figure 11.12.
It was unfortunate that no clearer image could be obtained at other locations through the
transfer spoon as with the current limited results, there was no way to visualise the trend
of the stream velocity through the spoon. It was hoped that the results obtained in the
following sections could predict the experimental behaviour and give an insight into
those trends.
271
Figure 11.11 Experimental testing of iron ore showing dust generation
Figure 11.12 Average particle velocity in the transfer spoon
After completing the testing for the 2 m/s belt speed, the conveyor transfer was adjusted
to allow for testing of a belt speed of 3 m/s. As expected, there was even more dust
being generated during testing and as a result, no video footage could be captured and
no further testing was carried out. This resulted in the comparison of the average stream
velocity through the transfer spoon being limited to a belt speed of 2 m/s only.
272
11.3.2 Analytical Method Analysis of Roberts

The Roberts analytical method was applied in an identical manner to the analysis of
polyethylene pellets, outlined in Section 11.2.2. The results of the analysis were only for
the 2 m/s belt speed, to provide a direct comparison with the experimental results, see
Figure 11.13.

The experimental test program yielded limited results due to the dust generation. This
means that there were minimal comparisons which could be made using DEM. This
said, the decision was made that more focus would be put upon the DEM component of
the analysis for iron ore, especially in the case of the EDEM simulations.

Two Chute MavenTM simulations were completed for iron ore, the first being extracted
from the complete conveyor transfer simulation, which included the influence of the
hood flow and particle freefall and the second from a simulation of the flow through the
transfer spoon only, see Figure 11.14. Due to the limitations previously found with
Chute MavenTM and the number of particles able to be successfully simulated, only the
low mass flow rate was compared. The results showed very little difference between the
spoon data extracted from the entire transfer simulation when compared with the results
extracted from the simulation when only the transfer spoon was considered. This seems
to imply that the entire transfer DEM simulation is generating an average particle
velocity at the point of impact with the spoon very similar to that determined
experimentally.

EDEM simulations were only completed for the 2 m/s belt speed case for iron ore,
allowing comparisons with the experimental results and the analytical method of
Roberts. Both low and high mass flow rates were simulated to maximise the possible
comparisons. The transfer spoon simulation outputs provided here are based on the
entire transfer EDEM simulations started in Section 9.3.3.2 for the transfer hood with a
bin being implemented to extract the data for the transfer spoon only.
273
Figure 11.13 Predicted stream velocity through the spoon by Roberts method
Figure 11.14 Simulation results from the Chute MavenTM transfer hood and spoon
simulations and the spoon only simulations
The coefficient of rolling friction for the spherical particles is 0.15 and for the shaped
particles is 0.05, as previously defined for the EDEM transfer hood simulations. The
simulation results are shown in Figure 11.15. The comparisons with the experimental
results and analytical methods are provided later.
A review of the data in Figure 11.15 showed an unexpected result, where the average
particle velocity for the high mass flow rate using spherical particles, dropped
dramatically part way through the spoon. None of the other simulations showed this
274
trend. It would seem more realistic for the shaped particle simulations to have a reduced
velocity through the spoon as the simulated particles would be more inclined to slide
rather than rotate or roll, i.e. the coefficient of friction between the shaped particles and
the Polystone liner in the spoon would be more conducive to seeing a reduction in
particle velocity. This unexpected result warranted further investigation, so a series of
simulations were generated to investigate the effect of varying both the coefficient of
rolling friction and the shear modulus of the iron ore. These additional simulations were
limited to the high mass flow rate cases as the increased number of particles was more
likely to show any variation which may have occurred. Ten simulations were completed
in total, five for spherical particles and five for shaped particles. All but one simulation
was determined using the particle volume based on the surface area of the sphere or
spheres being used to represent the particles for a given simulation. The other
simulations had the volume of particles being determined by the templates imported
into EDEM, used to define the particle shape. The EDEM outputs obtained from the ten
simulations are presented in Figure 11.16 with the results showing three simulations
where the average particle velocity drops noticeably through the spoon.
Figure 11.15 EDEM simulation results for low and high mass flow rates
These three simulations all had a coefficient of rolling friction of 0.15, which was the
value chosen from the initial sensitivity analysis. The remaining seven simulations all
showed a relatively consistent average particle velocity through the spoon, with a slight
275
drop in velocity near the exit of the spoon. From these results, the following early
observations were made;
as the coefficient of rolling resistance was increased for spherical particles, more
particles experienced reductions in velocity through the spoon,
there seemed to be no change in average particle velocity through the spoon for
shaped particles regardless of the coefficient of rolling friction used,
for the higher coefficient of rolling friction used in the spherical particle simulations,
there was a noticeable difference to the average particle velocity through the spoon
when the shear modulus was altered,
in conjunction with the coefficient of rolling friction used for the shaped particle
simulations, the variation in shear modulus had very little influence on the average
particle velocity through the spoon.
Figure 11.16 Ten EDEM simulations for the high mass flow rate
looking at variations of coefficient of rolling friction and shear modulus
for spherical and shaped particles
Additional data was extracted from these ten simulations to see if anything more
definite could be gleaned from the results. Firstly, using a narrowed bin of 100 mm
width taken along the central axis of the spoon, data was extracted with reference to
vertical displacement and the angular velocity of each particle. This data was then
plotted for all ten simulations in Figure 11.17. In all graphs, there was an initial vertical
section where the particles are in the freefall zone, between -500 and -850 mm, after
which the particles impact with the spoon, represented by the large horizontal spike.
276
The left column represents the five spherical particle simulations and as can be seen,
Figure 11.17a and 11.17b show similar results and vary substantially from the
remaining three spherical particle graphs. The first two graphs have a coefficient of
rolling friction of 0.05 and have substantially more particles with higher angular
velocities than the latter three spherical particle simulations with a coefficient of rolling
friction of 0.15. The variation in shear modulus which has been used for these five
simulations does not seem to have a noticeable affect on the angular velocity.
The right hand column represents the five shaped particle simulations and in contrast to
the spherical particle simulations, the coefficient of rolling friction does not seem to
impact on the degree of angular velocity the particles have in each of the five
simulations. There is one minor difference noted with Figure 11.10h, where the particle
volume was based on the template rather than the surface of the generated particle.
In Figure 11.16, the two spherical simulations with the coefficient of rolling friction set
to 0.05 showed very similar results to all five shaped particle simulations, whereas in
the comparisons in Figure 11.17, they showed different results. The three spherical
particle simulations, where the coefficient of rolling friction was set to 0.15, showed
similar trends to the five shaped particle simulations, which goes against what was
originally seen in Figure 11.16. At this point still no conclusive explanation can be
provided. This resulted in a slightly higher angular velocity being seen within that
particular simulation. Like the spherical particle simulations, the shear modulus does
not seem to influence the outcome to any noticeable degree.
A further data set was extracted from the spoon simulations via another bin set, where
the horizontal displacement of each particle across the transfer spoon was extracted
along with the angular velocity of each particle. The results were then graphed for all
ten simulations and displayed in Figure 11.18. The results of these comparisons showed
the same trend as seen in Figure 11.17, with the spherical particle simulations having a
coefficient of rolling friction set at 0.05 showing a larger spread of angular velocities for
the particles. This again goes against the comparisons that were shown in Figure 11.16.
277
(a) (f)
(b) (g)
(c) (h)
(d) (i)
(e) (j)
Figure 11.17 The angular velocity (rpm) of particles with respect to the
vertical displacement of the particles through the transfer spoon
In one last attempt to determine a conclusive explanation for the average velocity
dropping through the spoon for the spherical particle simulations, screen captures were
taken of the actual steady-state particle streams resulting from the EDEM simulations.
278
These are presented in Figure 11.19 for the ten simulations completed. The main area of
interest is where the transfer spoon walls are converging toward the outlet. Across the
ten simulations, variations in the particle velocity can be seen. Figure 11.19(c), (d) and
(e) show reduced particle velocities towards the spoon exit, which correlates with the
results of the first set of comparisons in Figure 11.16.
(a) (f)
(b) (g)
(c) (h)
(d) (i)
(e) (j)
Figure 11.18 The horizontal displacement of particles across the transfer spoon with
respect to the angular velocity (rpm) of particles through the transfer spoon
279
The remaining seven simulations all show very similar velocities through this section of
the transfer spoon, also the same as what was shown in Figure 11.16.
(a) (b) (c) (d) (e)
(f) (g) (h) (i) (j)
Figure 11.19 Front view of the particle velocity through the spoon for all ten
simulations used in these comparisons
280
From the results of these comparisons (vis. Figures 11.16 to 11.19), there seems to be
two distinct sets of findings, those of figure 11.16 and 11.19 and those of Figures 11.17
and 11.18. As this is the case, no conclusive explanation for why the spherical
simulations with a 0.15 coefficient of rolling resistance, showed a substantial slowing of
particle velocity part way through the transfer spoon, when the corresponding shaped
particle simulation did not show the same slowing of particle velocity.
Even though the findings are inconclusive, there is an indication that the coefficient of
rolling resistance used for the spherical particles should have perhaps been set at 0.05 as
it was for the shaped particle simulations. Doing this is justified when the results of
Table 9.5 are considered. The 0.05 coefficient of rolling resistance for spherical
particles showed a 99.46% agreement between the belt speed and corresponding particle
discharge velocity. As such, if Figure 11.15 was adjusted to include the data from the
spherical simulation where the coefficient of rolling resistance was set to 0.05 and the
shear modulus was unaltered (i.e. 1.926 x 109), the result is shown in Figure 11.20.
There is still a minor drop in average particle velocity in the spoon for the high mass
flow rate spherical simulation compared to the shaped simulation but only equates to a
0.2 m/s difference at the largest deviation.
The comparisons made in the preceding pages are lacking one vital aspect. They do not
provide comparisons to the other methods used to determine average particle velocity
through the transfer spoon. This will be covered in Section 11.3.4 below.
11.3.4 Method Comparisons

Combining the results from the experimental, analytical Roberts model and DEM
simulations from Chute MavenTM and EDEM, the comparison graph of Figure 11.21
has been produced. It can instantly be seen that the Roberts method for both the low and
high mass flow rates matches the DEM simulations results for Chute MavenTM and
EDEM very well except for one EDEM simulation. This particular simulation result has
been investigated in detail in the previous section but without reference to any of the
other methods determining the average particle velocity through the spoon. Now, with
all data being presented on the one graph, the comparisons can be made. What should
be noted is that the experimental results for the high mass flow rate fall directly on this
281
particular EDEM simulation result. Further comparing this simulation to the

experimental result, it can be seen that there is a variation in the initial impact point on
the spoon. This is actually the case with all the EDEM simulations.
Figure 11.20 Adjusted EDEM simulation results for low and high
mass flow rates
Figure 11.21 Average particle velocities through the transfer spoon for all methods
It is unfortunate that the dust generation in the experimental test program was so
dominant as more experimental data points would have provided the validation needed
282
to prove or disprove that the coefficient of rolling resistance of 0.15 was the correct
choice for the EDEM simulations of spherical particles.
11.4 DISCUSSION
Experimentally, two products have been tested, polyethylene pellets at belt speeds of 1,
2 and 3 m/s and iron ore at a belt speed of 2 m/s. The position of the transfer spoon has
been adjusted in each case to provide an ideal impact with respect to incoming angle
of incidence; however, for the polyethylene pellet testing, the non-ideal positioning of
the transfer spoon was also investigated. The non-ideal positioning of the spoon showed
a noticeable reduction in the velocity of the particles flowing through the spoon and the
flow-on effects of this are a build-up of material at the exit, causing a bottleneck which
chokes the spoon.
The Roberts analytical method showed that there was a general agreement with the
experimental polyethylene pellet results up to the point where the radius change
occurred in the spoon profile. For the iron ore comparisons, limited results could be
achieved experimentally which made it extremely difficult to obtain comparisons with
the Roberts method. The average particle velocity at the point of impact with the spoon
was all that could be measured for the low mass flow rate, so the experimental velocity
trend could not be determined. Slightly better, was the result for the high mass flow
rate, where one additional point was measured just prior to the converging of the spoon.
This still did not provide the velocity trend but showed that the Roberts analytical
method over-predicted the experimental result substantially.
The Chute Maven discrete element modelling software only allows for spherical
particles to be simulated. With the inability to simulate true particle shape, differences
between the DEM and experimental results are expected, including differing particle-
particle and particle-wall interactions and also the fact that air drag effects cannot be
modelled.
The particle stream outputs of the Chute MavenTM DEM simulations, examples of
which are shown in Figure 11.6, showed similar visual characteristics to the
experimental results, although this is somewhat hard to see in Figure 11.2. This
indicates that the Chute MavenTM software is capable of modelling the behaviour of the
283
material flow stream reasonably well, whereas the analytical method of Roberts is not
capable of predicting such dynamic and transient behaviour.
The EDEM software has the added benefit that particles can be generated by clustering
spherical particles together. This should allow for a more realistic simulation output as
particles will not be as free flowing as simulations containing just spherical particle
estimates. This has had mixed results, with different coefficients of rolling friction used
for spherical and shaped particle simulations based on a sensitivity calibration. If the
parameters can be sufficiently determined for spherical particle simulations, the results
of simulations do not vary much from those determined from the shaped particle
simulations. An advantage here is that substantial time can be saved from simulations
using spherical particles.
The EDEM simulations showed varying results for the two test materials. For the
polyethylene pellet comparisons, simulations were performed for selected experimental
geometries only. The results showed a similar trend to those experimentally, although
there was a velocity offset present based on the fact that the data was being extracted
from a simulation of the whole conveyor transfer. The influence of the particle flow
through the transfer hood and freefall zones does not match that seen experimentally.
For the EDEM simulations of iron ore, the influence of the coefficient of rolling friction
illustrated just how critical it is to the overall particle velocity through the transfer
spoon. If this value is set too high, there is a noticeable decrease in particle velocity
through the spoon.
284
CHAPTER 12 BRINGING IT ALL TOGETHER
Chapter
12
BRINGING IT ALL TOGETHER
285
12.1 INTRODUCTION
The myriad of data obtained from the trajectory, transfer hood, freefall and transfer
spoon investigations have been presented in Chapters 8 through 11. This data was
obtained experimentally, from analytical models and also from discrete element
simulations. In an attempt to present this information as a whole, in a logical and
straightforward manner, graphs have been produced for each material and transfer
geometry incorporating each analysis method. Figures 12.2 to 12.11 detail the results
for polyethylene pellets while Figures 12.12 and 12.13 detail the results for iron ore.
12.2 POLYETHYLENE PELLET COMPARISONS

The comparisons provided in this section cover both the low and high mass flow rate
cases for the polyethylene pellet experimental, analytical modelling and discrete
element modelling investigations. Each comparison is split up in to the six key locations
within the conveyor transfer for which data could be determined (in most cases). The
individual distances or angular positions are not being considered in these comparisons,
just the specific event occurring within the conveyor transfer. It should also be noted
that not all analysis methods could be applied to each conveying case, such as the
Korzen method only being able to be applied to the transfer hood, Chute MavenTM only
being applied for the low mass flow rates and EDEM only being applied to key
experimental setups for the polyethylene pellet investigations.
For the 1 m/s belt speed cases presented in Figures 12.2 and 12.3, the results show a
similar trend for all analysis methods in that the particle velocity through the spoon is in
the vicinity of four times that of the particle velocity at discharge from the conveyor.
The substantial increase in particle velocity can be attributed to the height in which the
material must fall from the discharge point to the point of impact with the transfer
spoon, see Figure 12.1. The experimental particle velocity at the point of impact with
the spoon is taken from the high-speed video analysis. The Chute MavenTM and EDEM
curves are extracted from the analysis of the simulation data. The Roberts (2003)
method requires an initial particle velocity on the spoon to generate the velocity profile
through the spoon. This is determined using projectile motion theory to determine the
magnitude of the velocity at the point of impact with the spoon. Figure 12.1 represents
the height that the material must fall from the discharge point on the conveyor to impact
with the spoon, h1 + h2. Height h1 is determined using trigonometry, based on the
286
discharge angle, d, and the radius to the outer surface of the belt. Height h2 is
determined from measurements taken from the conveyor transfer research facility. V0 is
assumed to have a magnitude equal to the belt speed.
Figure 12.1 Representation of how to estimate the impact velocity on the spoon
In Chapter 11, the Roberts (2003) method used an initial particle velocity in the spoon
based on the experimental particle velocity at the point of impact with the spoon (for all
belt speeds). The 1 m/s belt speed has been explained above, but for belt speeds of 2 m/s
and higher, the exit velocity of particles from the hood is used to predict the particle
velocity at the end of the freefall zone (which coincides with the impact point on the
spoon). This initial particle velocity in the spoon is then used to predict the particle
velocity through the spoon. The results of this method are presented in Figures 12.4 to
12.11.
An area worthy of comment is the freefall zone, which is indicated on each figure.
Based on equation 2.92, the particle velocity should increase in this region before the
particle stream impacts with the transfer spoon. The particle velocity generally does
increase in this region; however, the experimental results do not show this as
convincingly. This is most likely due to the way in which the high-speed video footage
287
was analysed. The final particle freefall velocity coincides with the initial particle
velocity at the point of impact with the transfer spoon. As a result, the transient
behaviour of the material at the impact point masks the true final freefall velocity and
gives a somewhat reduced measure of the velocity at that point. This is also seen with
the Chute MavenTM results in Figure 12.8. There is a dramatic drop in particle velocity
at the end of the freefall section, which should not be the case. Referring to Figure 11.8,
it can clearly be seen that the graph of the simulation data for this case has a lower than
expected initial velocity in the spoon as a direct result of the transient nature of the
material at the point of impact.
The Korzen (1989) method was never designed to account for flow through a transfer
spoon so the comparisons made using this method are somewhat limited. This said, the
Korzen method does generally follow the experimental results quite well for the particle
velocity through the hood.
It is unfortunate that the original discrete element modelling performed with Chute
MavenTM was unable to generate the required high mass flow rates as this would have
aided in providing a complete comparison for the high mass flow rate cases. The
addition of the EDEM simulations for some of the high mass flow rate showed a good
agreement with the Chute MavenTM outputs for polyethylene pellets but a mixed
outcome for the iron ore comparisons.
Figure 12.2 Polyethylene pellets, Vb = 1 m/s, ms = 2 tph
288
289
Figure 12.6 Polyethylene pellets, Vb = 3 m/s, Position A, ms = 2 tph
Figure 12.7 Polyethylene pellets, Vb = 3 m/s, Position A, ms = 37.8 tph
Figure 12.8 Polyethylene pellets, Vb = 3 m/s, Position B, ms = 10 tph
290
Figure 12.9 Polyethylene pellets, Vb = 3 m/s, Position B, ms = 37.8 tph
Figure 12.10 Polyethylene pellets, Vb = 3 m/s, Position C, ms = 10 tph
Figure 12.11 Polyethylene pellets, Vb = 3 m/s, Position C, ms = 37.8 tph
291
12.3 IRON ORE COMPARISONS

As previously discussed in Section 11.3, limited experimental results were obtained for
the iron ore, especially for the spoon component. For the results which could be
obtained, a similar trend was seen for the low and high mass flow rates, as shown in
Figures 12.12 and 12.13.
The Roberts (2003) method has again been used to predict the particle flow through the
entire conveyor transfer, using the velocity at the exit of the hood to determine the
velocity after the freefall zone and then using that velocity as the initial velocity for the
flow through the spoon.
The Korzen (1989) method provides a very close agreement to the other methods for the
low and high mass flow rates but there is a minor under-prediction of the hood exit
velocity.
Focusing on the freefall section, the general trend is for an increase in velocity from the
exit of the transfer hood to the impact point of the transfer spoon, but the experimental
results for both the low and high mass flow rates show very little increase in particle
velocity. This is most likely due to the way in which the high-speed video footage was
recorded. The final particle freefall velocity coincides with the initial particle velocity at
the point of impact with the spoon. As a result, the transient behaviour of the material at
that point masks the true final freefall velocity and gives a somewhat reduced measure
of the velocity at that point. This is also true of the analysis of the DEM simulation
output from Chute MavenTM and EDEM to some extent.
The EDEM results for the low mass flow rate show nearly identical results for the
spherical and shaped particle simulations, but for the high mass flow rate, there is a
noticeable difference through the spoon. Potential reasons for this variation have been
discussed in Chapter 11.
292
Figure 12.12 Iron ore, Vb = 2 m/s, ms = 15.3 tph
Figure 12.13 Iron ore, Vb = 2 m/s, ms = 63.8 tph
12.4 DISCUSSION
From the comparisons presented in Sections 12.2 and 12.3, one consistent observation is
that for all methods there is a very good agreement for the particle velocity through the
conveyor hood. It is after this point that the majority of the variation within the methods
begins. More precise measurement of the impact velocity of the spoon is required to
avoid misrepresenting this as there is a blurring of what component of the particle
velocity can be classed as freefall and that which is the result of the impact behaviour of
the materials.
293
CHAPTER 13 CONCLUSIONS AND FURTHER WORK
Chapter
13
CONCLUSIONS AND FURTHER WORK
294
13.1 OVERVIEW
This thesis describes the research performed, investigating the particle flow mechanisms
in a belt conveyor transfer.
An experimental conveyor transfer research facility was designed and commissioned to

experimentally investigate the particle velocity of the flow stream through the conveyor
transfer hood and spoon and also to investigate and quantify conveyor trajectories.
Several test materials have been used along with various conveyor belt speeds and mass
flow rates.
Analytical models were then used as predictive tools to generate either particle velocity
trends or predictive trajectory profiles, allowing direct comparison to the experimental
findings.
Lastly, discrete element modelling was used to generate both simulated particle flows
through transfer chutes and also conveyor trajectories. These simulations were then
compared to the analytical models and experimental results.
13.2 CONCLUSIONS
The main aims of this thesis were to develop a database of conveyor transfer parameters
and to generate a series of comparisons as a means of validation for the analytical and
discrete element modelling of particle flow mechanisms within a conveyor transfer,
with respect to experimental equivalents.
13.2.1 Conveyor Discharge Angle

The conveyor discharge angle dictates the path that the conveyor trajectory takes when
material is discharged from the conveyor head pulley. The discharge angle is relevant
for low-speed conveying, where material wraps around the head pulley to some angular
position. For high-speed conveying material leaves the conveyor belt at the point of
tangency between the belt and the head pulley.
Comparisons were made between experimental, analytical and discrete element

modelling data and it was found there were a wide spread of results for the discharge
angle with the only partial match being between the experimental results and those
295
obtained from EDEM for both spherical and shaped particles. Even though this was the
case, the correlation was only in agreement from 0.5 to 1 m/s belt speeds, above this the
EDEM results diverged from the experimental.
From these findings it would seem that there needs to be more investigation into the
discharge angle with respect to the analytical models and to a lesser extent the discrete
element modelling.
13.2.2 Conveyor Trajectories

The experimental testing generated a range of trajectories for polyethylene pellets and
corn which have been compared to numerous trajectory models and discrete element
modelling simulations using EDEM. It is acknowledged that two granular free flowing
materials is a small overall sample, nevertheless, the data obtained has provided a
wealth of information.
Experimentally, checks should be made to determine if the material is discharging from

the belt conveyor at the belt speed. If this is not the case, then incorrect assumptions on
the position of the trajectory stream could result. This will have implications on the
subsequent design of any conveyor transfer which is required. Observation of the
experimental trajectories has shown that as the trajectory moves further from the head
pulley, there is diverging of the stream with respect to its profile. Another observation
which does not seem to have been documented previously is the fact that the extremities
of the trajectory stream curve downward as a result of the dynamic behaviour of the
product stream as is passes through the conveyor transition zone, creating a convex
profile to the lower trajectory boundary. This has implications if transfer chutes are to
be designed from two-dimensional analytical models which do not have the ability to
account for this.
The analytical methods vary in complexity and it has been shown that each trajectory
method produces an individual profile for low-speed conveying. High-speed conveying
sees a merging of some trajectory methods as they predict very similar paths. There are
methods which do not include particle characteristics (vis. C.E.M.A. (1966; 1979; 1994;
1997; 2005)) and as such, if the same conveyor geometry is used for conveying two
differing materials, the result will be trajectory profiles identical to each other. Other
296
methods include numerous particle characteristics, such as Korzen (1989). The

comparisons presented in this research have shown that the inclusion of air drag results
in a noticeable under-prediction of the material trajectory.
Discrete element modelling of the conveyor trajectories showed the trajectories for both
spherical and shaped particle simulations were very similar. This is most likely due to
the particle streams not being affected by any external influences on leaving the head
pulley of the conveyor. The DEM simulations have produced the same dynamic
behaviour seen in the experimental testing (i.e. diverging trajectory stream and convex
curving of the lower trajectory boundary).
The comparisons generated between the experimental results, analytical methods and
DEM have shown that; EDEM is capable of producing very accurate predictions of the
experimentally determined conveyor trajectories, the Booth (1934) method predicts the
experimental results accurately and EDEM and Booth generate identical trajectory
predictions for all belt speed cases.
13.2.3 Conveyor Transfer Hoods

The experimental testing of the transfer hood used in this research provided interesting
results. The ability to adjust the positioning of the hood allowed investigation of the
influence of the angle of incidence of the incoming flow stream and showed that a
higher angle of incidence generated more lateral spread of material at impact. This was
not generally a problem for the granular non-cohesive materials used throughout this
testing program but has the potential to generate substantial dust generation, chute wear
and product degradation. It was observed, however, that the velocity of the flow stream
through the chute was not substantially affected.
The high-speed video analysis of the polyethylene pellet flow stream allowed for the
velocity profile of the flow stream to be mapped throughout the hood to allow for
further comparisons to be made with the analytical and DEM methods. There was some
difficulty in capturing the corresponding iron ore flow stream due to the dust that was
generated. This meant that the detailed comparisons possible with the polyethylene
pellets could not be achieved to the same level.
297
The analytical models of Roberts (2001) and Korzen (1989) to determine chute flow
have no facility to account for the non-sphericity of particles. There may be ways to
incorporate this by modifying the coefficient of friction values used in each method;
however, this was not investigated as part of this research. Both methods resulted in a
slight over-prediction for the experimental comparison with polyethylene pellets. A
very close result was achieved with respect to the iron ore experimental data, but it must
be emphasised that with the limited experimental data obtained, there is no way to
compare the trend of the velocity profile through the transfer hood, instead only
comparing the particle velocities at the inlet and exit of the hood.
DEM simulations were performed using Chute MavenTM and EDEM. At this point in
time, from the data obtained, there are still some inaccuracies in the prediction of the
particle velocity through the transfer hood with respect to the experimental results. The
DEM software has the potential to closely predict the experimental flow of particles
through transfer chutes, however, the results indicate that there is still the need for
further calibration of the particle characteristics to better equate the results. The effect of
particle shape was investigated using EDEM, however, no discernible difference was
observed between the average particle velocity for spherical and shaped particle
simulations. A benefit of the DEM simulations is the transient behaviour of the particles
at the impact point can be observed, whereas the analytical models cannot.
13.2.4 Particle Freefall

The particle free fall of material occurred between the transfer hood and spoon. There
are two analytical approaches which can be used; that using the equation of motion for
constant acceleration or the terminal velocity equation. It has been found from this
research that the velocity of polyethylene pellets through the freefall zone of the
conveyor transfer compare very well to the equation of motion for constant acceleration.
This was also true for corn and iron ore. When considering the terminal velocity
equation, there were more noticeable differences. This is most likely due to the way in
which the terminal velocity of each product was determined. Ideally to measure the
terminal velocity, a large drop height is required over which numerous velocity
measurements are obtained to accurately determine the terminal velocity. This research
relied on a much smaller drop height to estimate this and as a result is the most likely
cause of the inaccuracies.
298
13.2.5 Conveyor Transfer Spoons

The experimental positioning of the transfer spoon was dictated by the position of the
transfer hood above. The effect of misalignment of the transfer spoon was also
investigated with the result being an increased angle of incidence causing elevated
lateral spreading of material at impact along with reduction in particle velocity,
generating stagnation of material at the spoon discharge and subsequent spillage.
The design of the converging style transfer spoon in this research did not allow for a
complete analysis of the particle velocity trend. This was something which did not
become evident until the test program was well underway. As a result, the comparisons
are only valid up to the point where the converging section of the spoon begins.
The analytical chute flow method showed a general agreement with the experimental
polyethylene results, if not a slight over-prediction. Less comparison was achievable
with the iron ore testing due to the level of dust present.
Chute MavenTM and EDEM were used to generate simulations of the particle flow
through the transfer spoon. Both were able to generate the dynamic behaviour of the
flow stream seen experimentally. The DEM simulations produced over-predictions of
the particle velocity through the conveyor transfer; however, the velocity trend matched
that seen experimentally.
As discussed in Chapter 12, the accurate measurement of the particle velocity at the
point of impact with the transfer spoon needs more investigation. The assumption has
been made that the velocity measured at the impact point from the experimental and
also DEM simulations, is the final velocity from the freefall zone. This has shown to be
not entirely correct as the transient nature of the impact zone results in a reduced
velocity.
13.2.6 Discrete Element Modelling

Throughout this research, DEM has been used to generate simulations for conveyor
discharge angles, conveyor trajectories, chute flow through hoods and spoons and also
particle freefall. Before any simulations for the above could be performed, there was a
need for calibration of the DEM software. This is vital to allow for accurate and valid
299
simulations to be produced. Numerous calibrations were performed throughout this

research and it was found that some parameters are of more importance than others.
Parameters such as static friction, rolling friction, density and coefficient of restitution
are parameters which can dramatically alter the behaviour of simulations whereas
parameters such as shear modulus and Poissons ratio are more forgiving. By
reducing the shear modulus by a factor of 100, in essence softening the particle, there is
negligible difference to the simulated output, however, can dramatically reduce
simulation time.
13.3 FURTHER WORK
13.3.1 Experimental
There are a number of general issues which have arisen through out the experimental
test program and are explained in the following sections.
13.3.1.1 Transfer Hood

The current method of determining the particle velocity through the transfer hood has
been to mount the Redlake X3 MotionPro high-speed digital video camera
perpendicular to the particle flow and measuring at 5 increments. In doing this, the
particles in contact with the hood wings are the particles from which the average
velocity is determined. There is perhaps a minor influence of wall friction between the
particles and the hood wing that could influence the velocity. To allow investigation of
this, it is suggested that a new rotating mounting frame be installed below the hood,
aligned with the central axis of the conveyor on which the Redlake X3 MotionPro high-
speed digital video camera can be positioned. The camera could then capture the full
width of the particle flow through the transfer hood for each angular position as before,
to obtain a direct comparison.
The conveyor transfer hood used in the experimental trials has been described in detail
in Chapter 3. The hood has a constant width throughout its radius and as such does not
allow for the stream to concentrate before being directed towards the transfer spoon. It
is recommended that a converging transfer hood be designed and constructed to
investigate the influence on the particle velocity through the transfer hood.
300
In doing this, it is foreseen that issues will arise regarding the way in which the Redlake
X3 MotionPro high-speed digital video camera is used to record the particle flow. With
the converging hood, the wing components will no longer be perpendicular to the video
camera and as such, the resulting video footage will be skewed. There are two main
ways in which this issue could be addressed; (i) the video camera is positioned so that it
is always perpendicular to the transfer hood wings for the region which is being
recorded, or (ii) the same method as mentioned above with the new mounting frame
below the transfer hood.
13.3.1.2 Transfer Spoon

The design and subsequent positioning of the transfer spoon needs more attention in the
future. The current design results in the exit of the spoon being below the upper most
surface of the receiving conveyor and as such the exit velocity of the particle stream
could not be determined from the Redlake X3 MotionPro high-speed digital video
camera. Following from the suggestions for a new mounting frame for the video camera
for the transfer hood, a similar mounting frame could be developed for the transfer
spoon to allow video footage to be obtained for the entire width of the particle flow.
This would have two benefits; (i) the entire flow through the transfer spoon could be
recorded and (ii) an investigation into whether there are any wall friction effects
influencing the particle velocity in the current setup.
Additionally, the design of the transfer spoon is not ideal from the point of view that the
converging section towards the exit of the spoon influences the flow dramatically. This
is especially true if the spoon is not aligned in an optimum position, resulting in choking
in the converging section. To overcome this, it is recommended that a new transfer
spoon be designed which has a constant width throughout. Two designs are perhaps
needed, one which can work with the existing constant width transfer hood and another
which would be narrower to work in conjunction with the converging transfer hood,
recommended in Section 13.3.1.1.
There would seem to be a need for further investigation into the impact point within the
transfer spoon as described in Section 13.2.5, with respect to which velocity should be
assumed as the initial velocity in the spoon.
301
13.3.1.3 Conveyor Trajectories

The experimental conveyor trajectory investigations were relatively straight forward to
set up and execute. The addition of the interception hopper was a great help in capturing
and redirecting material onto the receiving conveyor. As the belt speed was increased,
there was a noticeable spreading of the trajectory stream, to the point that for a belt
speed of 6 m/s and higher, there was a quantifiable amount of material being lost from
the system and spilling onto the floor. The inclusion of additional acrylic enclosures
around the lower half of the receiving conveyor would go a long way to reducing any
such spillage. Care would also need to be taken that these enclosures did not intrude on
the adjustment of the interception hopper to capture the trajectory stream resulting from
different belt speeds.
More test materials should be tested to continue establishing a database from which a
greater range of comparisons can be made. The possibility of testing fine powders
should also be investigated if suitable dust extraction can be implemented. This will be
of specific interest with respect to possible air drag effects which can then be compared
with the Korzen (1989) method.
13.3.1.4 Test Materials

Test materials have been selected on the basis of either their; relatively spherical nature,
uniform size, ease of modelling or relevance to industry. In saying this, only three
materials were tested and to truly obtain conclusive results, more materials over a wider
range of particle characteristics should be chosen to further validate the analytical
models and discrete element modelling via experimental testing.
Ultimately, the goal is to continue experimental testing with more materials to generate
a substantial database of test data which will aid in the continued validation of
analytical and discrete element modelling methods.
13.3.1.5 Dust Generation

Within the next 18 months, the conveyor transfer research facility will be relocated to
the SMART (Simulation, Modelling and Analysis for Research and Teaching)
Infrastructure Building which is currently under construction at the University of
Wollongong. There are strict conditions in place for any experimental work being
302
performed in this new building, especially in relation to dust generation. A more

substantial dust extraction system is required to keep the transfer zone as clear as
possible for high-speed filming. The dust extraction system should also be adaptable to
allow for use during trajectory testing. Care must be taken to ensure that the dust
extraction is not too high as to interfere with the natural flow of material through either
the transfer chute or the conveyor trajectories. The positioning of any dust extraction
points should also be considered to ensure that there is no obstruction to the required
visualisation of the material flow streams.
13.3.2 Analytical Modelling

The analytical modelling of the transfer hood and spoon flow did show slight over-
predictions to that found experimentally. It may be worth investigating whether other
friction principles other that Coulomb friction can be used in these models. Perhaps the
inclusion of a coefficient of rolling friction could be implemented to aid in the reduction

of the predicted particle velocity through the transfer hood and spoon.
The analytical models for trajectory prediction show a wide range of profiles but there is
one method which has predicted the experimental trajectory very well (Booth, 1934).
There is still scope for further investigation into the discharge angle which is required in
each of these methods.
13.3.3 Discrete Element Modelling

There is an ongoing need for accurate calibration of particle and system parameters to
verify that the correct particle behaviours are being observed in bench-scale situations.
Only then can large-scale simulations be completed with confidence. The development
of a range of bench-scale experiments is currently underway at the University of
Wollongong and will provide that validation.
The linear-spring contact model has been used in this research but there is the facility to
allow other contact models to be used in EDEM. The effect of these has not yet been
investigated and the influence of these on the DEM outputs is not yet known.
303
REFERENCES
REFERENCES
304
REFERENCES
3DCAM. (2009). "Acrylic Properties - Physical, Mechanical, Electrical, Thermal,

Optical, Processing".
[Online, accessed 27 February].
http://www.3d-cam.com/materials/acrylic.asp
ACTek. (n.d.). "ACTek website".

[Online, accessed 26th September 2007].
http://www.actek.com/
Alspaugh, M., Dewicki, G. and Quesenberry, E. (2002). "Computer Simulation Solves

Conveyor Problems". Coal Age, No. January, pp. 28 - 31.
Alspaugh, M. A. (2004). "Latest Developments in Belt Conveyor Technology".

[Online, accessed 4th November, 2005].
http://www.overlandconveyor.com/pdf/Latest%20Developments%20in%20Belt%20Co
nveyor%20Technology.pdf
Anderson, S. and Britton, P. (2007). "Tackling Industrial Materials Handling Issues

with Discrete Element Modelling". Australian Bulk Handling Review, Vol. 12, No. 3,
March/April, pp. 94 - 98.
Andrews, D. (2008). "A Comparison of Belt Conveyor Trajectory Predictions", B.E.

(Mech. Eng.) Thesis, University of Wollongong, Wollongong, New South Wales,
Australia. pp. 71.
Archer, R. R., Cook, N. H., Crandall, S. H., Dahl, N. C., Lardner, T. J., McClintock, F.
A., Rabinowicz, E. and Reichenbach, G. S. (1972). "An Introduction to the Mechanics
of Solids". 2nd Ed. New York, McGraw-Hill. pp. 628.
Arnold, P. C. and Hill, G. L. (1990). "Predicting the Discharge Trajectory from Belt
Conveyors". Bulk Solids Handling, Vol. 10, No. 4, October, pp. 379 - 382.
Asmar, B. N., Langston, P. A. and Matchett, A. J. (2004). "Discrete Element Modelling

of the Vibratory Unconfined Compression Tester (VUCT) for Cohesive Solids".
Granular Matter, Vol. 5, No. 4, pp. 213 - 215.
Asmar, B. N., Langston, P. A., Matchett, A. J. and Walters, J. K. (2002). "Validation

Tests on a Distinct Element Model of Vibrating Cohesive Particle Systems". Computers
and Chemical Engineering, Vol. 26, No. 6, pp. 785 - 802.
Bakshi, S. R., Balani, K., Laha, T., Tercero, J. and Agarwal, A. (2007). "The
Nanomechanical and Nanoscratch Properties of MWNT-Reinforced Ultrahigh-
Molecular-Weight Polyethylene Coatings". JOM, Vol. 59, No. 7, pp. 50 - 53.
Balevicius, R., Kacianauskas, R., Mroz, Z. and Sielamowicz, I. (2008). "Discrete-

Particle Investigation of Friction Effect in Filling and Unsteady/Steady Discharge in
Three-Dimensional Wedge-Shaped Hopper". Powder Technology, Vol. 187, No. 2, pp.
159 - 174.
305
REFERENCES
Benjamin, C. W. (1999). "Rapid Design Approach to the Zero-Spillage Transfer

Chute". Australian Bulk Handling Review, Vol. 4, No. 2, June/July, pp. 52 - 58.
Bertrand, F., Leclaire, L.-A. and Levecque, G. (2005). "DEM Based Models for the
Mixing of Granular Materials". Chemical Engineering Science, Vol. 60, No. 8-9, pp.
2517 - 2531.
Booth, E. P. O. (1934). "Trajectories from Conveyors - Method of Calculating Them

Corrected". Engineering and Mining Journal, Vol. 135, No. 12, December, pp. 552 -
554.
Brown, P. P. and Lawler, D. F. (2003). "Sphere Drag and Settling Velocity Revisited".
Journal of Environmental Engineering, Vol. 129, No. 3 (March), pp. 222 - 231.
Bujak, B. and Bottlinger, M. (2008). "Three-Dimensional Measurement of Particle

Shape". Particle and Particle Systems Characterization, Vol. 25, No. 4, pp. 293 - 297.
Burnett, A. J. (2000). "Chutes for Stacker Reclaimers - Generation of a Design

Method". From Powder to Bulk: International Conference on Powder and Bulk Solids
Handling, London, UK, 13 - 15 June, Professional Engineering Publishing Limited, pp.
463 - 476.
C.E.M.A. (1966). "Belt Conveyors for Bulk Materials". 1st Ed, Conveyor Equipment
Manufacturers Association. pp. 332.
C.E.M.A. (1979). "Belt Conveyors for Bulk Materials". 2nd Ed, Conveyor Equipment
C.E.M.A. (1994). "Belt Conveyors for Bulk Materials". 4th Ed, Conveyor Equipment
Campbell, C. S. and Brennen, C. E. (1985). "Computer Simulation of Granular Shear

Flows". Journal of Fluid Mechanics, Vol. 151, February, pp. 167 - 188.
Castro Vazquez, J. C., Johanson, K. D. and Cristescu, N. D. (2006). "Particle-Structure

Interactions for Wall Friction Studies Using Discrete Element Models". 5th World
Congress on Particle Technology, Orlando, Florida, USA, 23rd - 27th April.
Chandramohan, R. and Powell, M. S. (2005). "Measurement of Particle Interaction

Properties for Incorporation in the Discrete Element Method Simulation". Minerals
Engineering, Vol. 18, No. 12, pp. 1142 - 1151.
Chu, K. W. and Yu, A. B. (2007). "Numerical Simulation of Complex Particle-Fluid

Flows". Powder Technology, Vol. 179, No. 1, pp. 104-114.
306
REFERENCES
Cleary, P. (1998). "Discrete Element Modelling of Industrial Granular Flow

Applications". Task Quarterly, Vol. 2, No. 3, pp. 385 - 415.
Cleary, P. (2001). "Modelling Comminution Devices Using DEM". International Journal

for Numerical and Analytical Methods in Geomechanics, Vol. 25, No. 1, pp. 83 - 105.
Cleary, P. W. (2000). "DEM Simulation of Industrial Particle Flows: Case Studies of

Dragline Excavators, Mixing in Tumblers and Centrifugal Mills". Powder Technology,
Vol. 109, No. 1, pp. 83 - 104.
Cleary, P. W. and Hoyer, D. (2000). "Centrifugal Mill Charge Motion and Power
Draw: Comparison of DEM Predictions with Experiment". International Journal of
Mineral Processing, Vol. 59, No. 2, pp. 131 - 148.
Cleary, P. W., Laurent, B. and Bridgewater, J. (2002). "DEM Prediction of Flow

Patterns and Mixing Rates in a Ploughshare Mixer". 4th World Congress on Particle
Technology, Sydney, Australia, 21st - 25th July.
Cleary, P. W., Metcalfe, G. and Liffman, K. (1998). "How Well do Discrete Granular
Flow Models Capture the Essentials of Mixing Processes ?" Applied Mathematical
Modelling, Vol. 22, pp. 995 - 1008.
Cleary, P. W. and Sawley, M. L. (1999). "Three-Dimensional Modelling of Industrial

Granular Flows". 2nd International Conference on CFD in the Minerals and Process
Industries, CSIRO, Melbourne, Australia, 6th - 8th December, pp. 95 - 100.
Cleary, P. W. and Sawley, M. L. (2002). "DEM Modelling of Industrial Granular

Flows: 3D Case Studies and the Effect of Particle Shape on Hopper Discharge".
Applied Mathematical Modelling, Vol. 26, pp. 89 - 111.
Cleary, P. W., Sinnott, M. D. and Morrison, R. D. (2006). "DEM Prediction of Particle

Flows and Breakage in Comminution Processes". 5th World Congress on Particle
Technology, Orlando, Florida, USA, 23rd - 27th April.
Coetzee, C. J. and Els, D. N. J. (2009). "Calibration of Discrete Element Parameters

and the Modelling of Silo Discharge and Bucket Filling". Computers and Electronics in
Agriculture, Vol. 65, No. 2, pp. 198 - 212.
Colijn, H. and Conners, P. J. (1972). "Belt Conveyor Transfer Points". Transactions of

the Society of Mining Engineers, AIME, Vol. 252, June, pp. 204 - 210.
Cundall, P. A. (1988). "Computer Simulations of Dense Sphere Assemblies".

Micromechanics of Granular Materials. M. Satake and J. T. Jenkins (Ed.). Amsterdam,
Elsevier, pp. 113 - 123.
Cundall, P. A. and Strack, O. D. L. (1979). "A Discrete Numerical Model for Granular
Assemblies". Geotechnique, Vol. 29, No. 1, pp. 47 - 65.
E-DEM Software Help File. (2008). Software, DEM Solutions.
307
REFERENCES
DEM-Solutions. (n.d.). "DEM-Solutions website".

[Online, accessed 26th September 2007].
http://www.dem-solutions.com
Dewicki, G. (2003). "Bulk Material Handling and Processing Numerical Techniques

and Simulation of Granular Material". Bulk Solids Handling, Vol. 23, No. 2, pp. 1 - 4.
Dewicki, G. and Mustoe, G. (2002). "Bulk Material Belt Conveyor Transfer Point
Simulation of Material Flow Using DEM". 3rd International Conference on DEMs,
Sante Fe, New Mexico, USA, 23rd - 25th September.
Di Maio, F. P. and Di Renzo, A. (2004). "Analytical Solution for the Problem of

Frictional-Elastic Collisions of Spherical Particles Using the Linear Model". Chemical
Engineering Science, Vol. 59, No. 16, pp. 3461 - 3475.
Di Renzo, A. and Di Maio, F. P. (2004). "Comparison of Contact-Force Models for the

Simulation of Collisions in DEM-Based Granular Flow Codes". Chemical Engineering
Science, Vol. 59, No. 3, pp. 525 - 541.
Di Renzo, A. and Di Maio, F. P. (2005). "Modelling Particle Collisions in DEM

Simulations with an Improved Integral Non-Linear Model". Powders and Grains
Volume 2, Taylor and Francis Group, London, pp. 1241 - 1245.
Dobry, R. and Ng, T.-T. (1992). "Discrete Modelling of Stress-Strain Behaviour of

Granular Media at Small and Large Strains". Engineering Computations, Vol. 9, pp.
129 - 143.
Douberly, E. B. (2003). "Fire-Protection Guidelines for Handling and Storing PBR

Coal". Power, Vol. 147, No. 9.
http://www.powermag.com/archive_article.asp?a=20x0e4I0810132Y1G3ax7b_1&y=20
03&m=october
Drake, T. G. and Walton, O. R. (1995). "Comparison of Experimental and Simulated

Grain Flows". Journal of Applied Mechanics, Vol. 62, No. 1, pp. 131 - 135.
Dunlop (1982). "Dunlop Industrial Conveyor Manual".
Dutt, M., Hancock, B., Bentham, C. and Elliott, J. (2005). "An Implementation of
Granular Dynamics for Simulating Frictional Elastic Particles Based on the DL_POLY
Code". Computer Physics Communications, Vol. 166, No. 1, pp. 26 - 44.
Dziugys, A. and Peters, B. (2001). "An Approach to Simulate the Motion of Spherical
and Non-Spherical Fuel Particles in Combustion Chambers". Granular Matter, Vol. 3,
No. 4, pp. 231 - 265.
CES EduPack. (2009). Software, Version 5.1.0, Granta Design Limited, Cambridge,
United Kingdom.
308
REFERENCES
Ekstrom, G. A., Leiljedahl, J. B. and Peart, R. M. (1966). "Thermal Expansion and

Tensile Properties of Corn Kernels and Their Relationship to Cracking During Drying".
Transactions of the ASAE, Vol. 9, No. 4, pp. 556 - 561.
Favier, J. F., Abbaspour-Fard, M. H., Kremmer, M. and Raji, A. O. (1999). "Shape

Representation of Axisymmetrical, Non-Spherical Particles in Discrete Element
Simulation Using Multi-Element Model Particles". Engineering Computations, Vol. 16,
No. 4, pp. 467 - 480.
Fraige, F. Y. and Langston, P. A. (2006). "Horizontal Pneumatic Conveying: a 3D

Distinct Element Model". Granular Matter, Vol. 8, No. 2, pp. 67 - 80.
Franz, M. (2007). "Design of Belt Conveyor Transfer Stations". Bulk Solids Handling,
Vol. 27, No. 3, pp. 168-173.
Fu, X., Dutt, M., Bentham, A. C., Hancock, B. C., Cameron, R. E. and Elliott, J. A.
(2006). "Investigation of Particle Packing in Model Pharmaceutical Powders Using X-
Ray Microtomography and Discrete Element Method". Powder Technology, Vol. 167,
No. 3, pp. 134 - 140.
Gan, M., Gopinathan, N., Jia, X. and Williams, R. A. (2004). "Predicting the Packing
Characteristics of Particles of Arbitrary Shapes". KONA, Vol. 22, pp. 82 - 93.
Gantt, J. A., Cameron, I. T., Litster, J. D. and Gatzke, E. P. (2006). "Determination of

Coalescence Kernels for High-Shear Granulation Using DEM Simulations". Powder
Technology, Vol. 170, No. 2, pp. 53 - 63.
Garg, O. P. (1973). "in Situ Physicomechanical Properties of Permafrost Using

Geophysical Techniques". Permafrost: North American Contribution to the Second
International Conference, Yakutsk, U.S.S.R., 13 - 28 July, National Academy of
Sciences, Washington D.C., U.S.A., pp. 783.
Geldart, D. (1990). "Estimation of Basic Particle Properties for Use in Fluid-Particle

Process Calculations". Powder Technology, Vol. 60, No. 1, pp. 1 - 13.
Goda, T. J. and Ebert, F. (2005). "Three-Dimensional Discrete Element Simulations In

Hoppers And Silos". Powder Technology, Vol. 158, No. 1, pp. 58 - 68.
Golka, K. (1992). "Discharge Trajectories of Bulk Solids". 4th International Conference

on Bulk Materials Storage, Handling and Transportation, Wollongong, NSW, Australia,
6th - 8th July, pp. 497 - 503.
Golka, K. (1993). "Prediction of the Discharge Trajectories of Bulk Materials". Bulk

Solids Handling, Vol. 13, No. 4, November, pp. 763 - 766.
Golka, K., Bolliger, G. and Vasili, C. (2007). "Belt Conveyors Principles for
Calculation and Design". Lugarno, N.S.W., Australia, K. Golka, G. Bolliger, C. Vasili.
pp. 288.
309
REFERENCES
Golz, P. and Favier, J. (2006). "DEM-CFD Modeling of Solid-Fluid Flows,". 5th World
Congress on Particle Technology, Orlando, Florida, USA, 23rd - 27th April.
Goodyear (1975). "Goodyear Handbook of Conveyor & Elevator Belting", pp. Section
11.
Govender, I., Powell, M. S. and Nurick, G. N. (2001). "3D Particle Tracking in a Mill:
A Rigorous Technique for Verifying DEM Predictions". Minerals Engineering, Vol. 14,
No. 10, pp. 1329 - 1340.
Grgic, D., Amitrano, D. and Homand, F. (2008). "Static Fatigue of Iron Ore in
Underground Iron Mines (Lorraine, France)". Post-Mining Symposium 2008, Nancy,
France, 6 - 7 February, pp. 14.
Grima, A. P. (2007). "Variable - Geometry Conveyor Transfer Research Facility", B.E.

(Mech. Eng.) Thesis, University of Wollongong, Wollongong, New South Wales,
Australia. pp. 150.
Grima, A. P. and Wypych, P. W. (2009). "Investigations into Calibration for Discrete

Element Modelling of Granular Materials". 6th International Conference for Conveying
and Handling of Particulate Solids (CHoPS 2009) and 10th International Conference on
Bulk Materials Storage, Handling and Transportation (ICBMH), Brisbane, Australia,
Engineers Australia. pp. 69 - 76.
Grger, T. and Katterfeld, A. (2006a). "On the Numerical Calibration of Discrete

Element Models for the Simulation of Bulk Solids". 5th World Congress on Particle
Grger, T. and Katterfeld, A. (2006b). "On the Numerical Calibration of Discrete

Element Models for the Simulation of Bulk Solids". 16th European Symposium on
Computer Aided Process Engineering and 9th Symposium on Process Systems
Engineering, Garmisch-Partenkirchen, Germany, July 9 - 13, pp. 533 - 538.
Grger, T. and Katterfeld, A. (2007). "Applications of the DEM in Materials Handling

Part 3 Transfer Stations". Bulk Solids Handling, Vol. 27, No. 3, pp. 158-166.
Haff, P. K. and Anderson, R. S. (1993). "Grain Scale Simulations of Loose Sedimentary

Beds: The Example of Grain-Bed Impacts in Aeolian Saltation". Sedimentology, Vol.
40, No. 2, pp. 175 - 198.
Haider, A. and Levenspiel, O. (1989). "Drag Coefficient and Terminal Velocity of

Spherical and Nonspherical Particles". Powder Technology, Vol. 58, No. 1, pp. 63 - 70.
Halford, W. (2009). "Bulk Materials Handling Flow Properties Instruction Manual".

Bulk Materials Engineering Australia (BMEA), pp. 50.
Han, T., Levy, A. and Kalman, H. (2003). "DEM Simulation for Attrition of Salt During
Dilute-Phase Pneumatic Conveying". Powder Technology, Vol. 129, No. 1-3, pp. 92 -
100.
310
REFERENCES
Harris, H. G. and Sabnis, G. (1999). "Structural Modeling and Experimental Techniques

". 2nd Ed. Boca Raton, Florida, CRC Press. pp. 789.
Hemph, R., van Wachem, B. and Almstedt, A.-E. (2006). "DEM Modeling of Hopper
Flows: Comparison and Validation of Models and Parameters". 5th World Congress on
Particle Technology, Orlando, Florida, USA, 23rd - 27th April.
Hogue, C. (1998). "Shape Representation and Contact Detection for Discrete Element
Simulations of Arbitrary Geometries". Engineering Computations, Vol. 15, No. 3, pp.
374 - 390.
Hogue, C. and Newland, D. (1994). "Efficient Computer Simulation of Moving

Granular Particles". Powder Technology, Vol. 78, pp. 51 - 66.
Hustrulid, A. I. (1997). "A Computational Methodology for Modeling Large Scale

Sublevel Caving with a 3D Discrete Element Method", PhD Thesis, Colorado School of
Mines, Golden, Colorado, USA. pp. 196.
Hustrulid, G. L. (2005). "Chute Maven Material Flow Modeling - Software for

Modeling Conveyor Transfer Points". Hustrulid Technologies Incorporated, pp. 33.
Ilic, D. D., McBride, W. and Katterfeld, A. (2007). "Validation of Continuum Methods

Utilising Discrete Element Simulations as Applied to a Slewing Stacker Transfer
Chute". 9th International Conference on Bulk Materials, Storage, Handling and
Transportation, Proceedings on USB, Newcastle, Australia, 9th-11th October, pp. 11.
Iwashita, K. and Oda, M. (1998). "Rolling Resistance at Contacts in Simulation of

Shear Band Development by DEM". Journal of Engineering Mechanics, Vol. 124, No.
3, pp. 285 - 292.
Jensen, R. P., Bosscher, P. J., Plesha, M. E. and Edil, T. B. (1999). "DEM Simulation of
Granular Media Structural Interface: Effects of Surface Roughness and Particle
Shape". International Journal for Numerical and Analytical Methods in Geomechanics,
Vol. 23, pp. 531 - 547.
Ji, S., Shen, H. H. and Orlando, A. (2006). "Scaling Effect in DEM Simulations of
Direct Shear Test". 5th World Congress on Particle Technology, Orlando, Florida,
USA, 23rd - 27th April.
Jia, C.-C., Sun, D.-W. and Cao, C.-W. (2000). "Mathematical Simulation of Stresses
Within a Corn Kernel During Drying". Drying Technology, Vol. 18, No. 4 - 5, pp. 887 -
906.
Jin, Z. M., Heng, S. M., Ng, H. W. and Auger, D. D. (1999). "An Axisymmetric Contact
Model of Ultra High Molecular Weight Polyethylene Cups Against Metallic Femoral
Heads for Artificial Hip Joint Replacements". Proceedings of the Institution of
Mechanical Engineers. Part H Journal of Engineering in Medicine, Vol. 213, No. 4, pp.
317 - 327.
311
REFERENCES
Jones, C. (1998). "Solve Common Coal-Handling Problems". Power, Vol. 142, No. 6,
November/December, pp. 31 - 32.
Kamaras, C. (2007). "Experimental Validation of DEM Simulation Modelling of

Particle Flows Through Conveyor Transfers", B.E. (Mech. Eng.) Thesis, University of
Wollongong, Wollongong, New South Wales, Australia. pp. 106.
Karamanev, D. G. (1996). "Equations for Calculation of the Terminal Velocity and

Drag Coefficient of Solid Spheres and Gas Bubbles". Chemical Engineering
Communications, Vol. 147, pp. 75 - 84.
Karuppiah, K. S. K., Bruck, A. L., Sundararajan, S., Wang, J., Lin, Z., Xu, Z.-H. and Li,
X. (2008). "Friction and Wear Behavior of Ultra-High Molecular Weight Polyethylene
as a Function of Polymer Crystallinity". Acta Biomaterialia, Vol. 4, No. 5, pp. 1401 -
1410.
Katterfeld, A. and Grger, T. (2006). "Verified Discrete Element Simulations of Bulk

Solids Handling Equipment". 5th World Congress on Particle Technology, Orlando,
Florida, USA, 23rd - 27th April.
Ketterhagen, W., Curtis, J. and Wassgren, C. (2006). "Modeling Granular Segregation

During Hopper Discharge via Discrete Element Methods". 5th World Congress on
Particle Technology, Orlando, Florida, USA, 23rd - 27th April.
Kondic, L. (1999). "Dynamics of Spherical Particles on a Surface: Collision-Induced

Sliding and Other Effects". Physical Review E, Vol. 60, No. 1, pp. 751 - 770.
Korzen, Z. (1988). "The Dynamics of Bulk Solids Flow on Impact Plates of Belt
Conveyor Systems". Bulk Solids Handling, Vol. 8, No. 6, December, pp. 689 - 697.
Korzen, Z. (1989). "Mechanics of Belt Conveyor Discharge Process as Affected by Air

Drag". Bulk Solids Handling, Vol. 9, No. 3, August, pp. 289 - 297.
Krause, F. and Katterfeld, A. (2004). "Usage of the Discrete Element Method in

Conveyor Technologies". 8th International Conference on Bulk Materials Storage,
Handling and Transportation, Wollongong, NSW, Australia, 5th - 8th July, Institution
of Engineers, Australia, pp. 289 - 293.
Kruggel-Emden, H., Rickelt, S., Wirtz, S. and Scherer, V. (2008). "A Study on the
Validity of the Multi-Sphere Discrete Element Method". Powder Technology, Vol. 188,
No. 2, pp. 153 - 165.
Kruggel-Emden, H., Simsek, E., Wirtz, S. and Scherer, V. (2006). "A New Approach for
Modeling the Normal Contact Within Granular Assemblies Through the Discrete
Element Method". 5th World Congress on Particle Technology, Orlando, Florida, USA,
23rd - 27th April.
Kruse, D. (2007). "Discussions with David Kruse, ACTek, Conducted 1-2 August 2007
", Beltcon 14 Conference, Johannesburg, South Africa.
312
REFERENCES
Kuo, H. P., Knight, P. C., Parker, D. J., Tsuji, Y., Adams, M. J. and Seville, J. P. K.
(2002). "The Influence of DEM Simulation Parameters on the Particle Behaviour in a
V-Mixer". Chemical Engineering Science, Vol. 57, No. 17, pp. 3621 - 3638.
Kwan, C. C., Mio, H., Yang, W., Ding, Y. and Ghadiri, M. (2006). "Analysis of Single
Ball Mill by Distinct Element Method". 5th World Congress on Particle Technology,
Orlando, Florida, USA, 23rd - 27th April.
Langston, P. A., Al-Awamleh, M. A., Fraige, F. Y. and Asmar, B. N. (2004). "Distinct

Element Modelling of Non-Spherical Frictionless Particle Flow". Chemical Engineering
Science, Vol. 59, No. 2, pp. 425 - 435.
Langston, P. A., Tuzun, U. and Heyes, D. M. (1995). "Discrete Element Simulation of

Granular Flow in 2D and 3D Hoppers: Dependence of Discharge Rate and Wall Stress
on Particle Interactions". Chemical Engineering Science, Vol. 50, No. 6, pp. 967 - 987.
Latham, J.-P., Munjiza, A., Garcia, X., Xiang, J. and Guises, R. (2008). "Three-
Dimensional Particle Shape Acquisition and Use of Shape Library for DEM and
FEM/DEM Simulation". Minerals Engineering, Vol. 21, No. 11, pp. 797 - 805.
Li, J., Langston, P. A., Webb, C. and Dyakowski, T. (2004). "Flow of Sphero-disc
Particles in Rectangular Hoppers A DEM and Experimental Comparison in 3D".
Chemical Engineering Science, Vol. 54, No. 24, pp. 5917 - 5929.
Li, L. and Holt, R. M. (2005). "Approaching Real Grain Shape in the Simulation of
Sandstone Using DEM". Powders and Grains Volume 2, Taylor and Francis Group,
London, pp. 1369 - 1373.
Li, S.-Q., Zhang, X., Zhao, X.-L. and Yao, Q. (2006). "Granular Surface Flow in a
Two-Dimensional Rotating Drum: Particle Dynamics Simulation". 5th World Congress
on Particle Technology, Orlando, Florida, USA, 23rd - 27th April.
Li, Y., Xu, Y. and Thornton, C. (2005). "A Comparison of Discrete Element Simulations
and Experiments for Sandpiles Composed of Spherical Particles". Powder Technology,
Vol. 160, No. 3, pp. 219 - 228.
Lin, X. and Ng, T.-T. (1995). "Short Communication: Contact Detection Algorithms for
Three-Dimensional Ellipsoids in Discrete Element Modelling". International Journal for
Numerical and Analytical Methods in Geomechanics, Vol. 19, pp. 653 - 659.
Lin, X. and Ng, T.-T. (1997). "A Three-Dimensional Discrete Element Model Using
Arrays of Ellipsoids". Geotechnique, Vol. 47, No. 2, pp. 319 - 329.
Litchfield, J. B. and Okos, M. R. (1988). "Prediction of Corn Kernel Stress and

Breakage Induced by Drying, Tempering and Cooling". Transactions of the ASAE, Vol.
31, No. 2, pp. 585 - 594.
Liu, L. F., Zhang, Z. P. and Yu, A. B. (1999). "Dynamic Simulation of the Centripetal
Packing of Mono-Sized Spheres". Physica A, Vol. 268, pp. 433 - 453.
313
REFERENCES
Lonie, K. (1989). "The Design of Conveyor Transfer Chutes". 3rd International

Conference on Bulk Materials Storage, Handling and Transportation, Newcastle,
Australia, 27th 29th June, pp. 240 - 244.
Loughran, J. G., Leonardi, C. R. and Anderson, S. I. (2004). "DEM of Bulk Flow

Through Complex Transitions". 8th International Conference on Bulk Materials
Storage, Handling and Transportation, Wollongong, NSW, Australia, 5th - 8th July,
Institution of Engineers, Australia, pp. 204 - 208.
Luding, S. (2006). "Comparison of Ring Shear Cell Simulations in 2D/3D with

Experiments". 5th World Congress on Particle Technology, Orlando, Florida, USA,
23rd - 27th April.
M&J.Engineering. (1996). "Design Operation of the WEBA Chute".

[Online, accessed 27th November 2006].
http://www.mjeng.co.za/design_operation.html
M.H.E.A. (1977). "Recommended Practice for Troughed Belt Conveyors", Mechanical

Handling Engineers Association. pp. 123.
M.H.E.A. (1986). "Recommended Practice for Troughed Belt Conveyors", Mechanical

Handling Engineers Association. pp. 199.
Majumder, A. K. and Barnwal, J. P. (2004). "A Computational Method to Predict

Particles Free Terminal Settling Velocity". Institution of Engineers (India) - Mining
Engineering Journal, Vol. 85, August, pp. 17 - 19.
Marcus, R. D., Leung, L. S., Klinzing, G. E. and Rizk, F. (1990). "Pneumatic

Conveying of Solids", Chapman and Hall. pp. 575.
Markauskas, D. (2006). "Discrete Element Modelling of Complex Axisymmetrical

Particle Flow". Mechanika, Vol. 62, No. 6, pp. 32 - 38.
Masson, S. and Martinez, J. (2000). "Effect of Particle Mechanical Properties on Silo

Flow and Stresses from Distinct Element Simulations". Powder Technology, Vol. 109,
No. 1-3, pp. 164 - 178.
Matbase. (n.d.). "Material Properties of PVC Soft, Commodity Polymers".

http://www.matbase.com/material/polymers/commodity/soft-pvc/properties
McBride, A., Govender, I., Powell, M. and Cloete, T. (2004). "Contributions to the
Experimental Validation of the Discrete Element Method Applied to Tumbling Mills".
Engineering Computations, Vol. 21, No. 2/3/4, pp. 119-136.
McBride, B. (1997). "Efficient Transfer Chutes A Case Study". BELTCON 9

International Materials Handling Conference, Midrand, South Africa, 22nd - 24th
October, pp. 1 - 7.
314
REFERENCES
McBride, B. (2000). "Curved Transfer Chute - Results of Site Trials and Optimisation".
From Powder to Bulk: International Conference on Powder and Bulk Solids Handling,
London, UK, 13 - 15 June, Professional Engineering Publishing Limited, pp. 411 - 418.
McDowell, G. R. and Harireche, O. (2002). "Discrete Element Modelling of Yielding

and Normal Compression of Sand". Geotechnique, Vol. 52, No. 4, pp. 299 - 304.
McGlinchey, D., Xiang, J., Cowell, A., Frye, L. and Peukert, W. (2004). "The
Prediction of Contact Forces During Dense Phase Pneumatic Conveying". 8th
International Conference on Bulk Materials Storage, Handling and Transportation,
Wollongong, NSW, Australia, 5th - 8th July, Institution of Engineers, Australia, pp. 342
- 346.
Meriam, J. L. and Kraige, L. G. (1987). "Engineering Mechanics - Statics". 2nd Ed,

John Wiley and Sons. pp. 454.
Mirghasemi, A. A., Rothenburg, L. and Matyas, E. L. (2002). "Influence of Particle

Shape on Engineering Properties of Assemblies of Two-Dimensional Polygon-Shaped
Particles". Geotechnique, Vol. 52, No. 3, pp. 209 - 217.
Mishra, B. K. and Murty, C. V. R. (2001). "On the Determination of Contact

Parameters for Realistic DEM Simulations of Ball Mills". Powder Technology, Vol.
115, No. 3, pp. 290 - 297.
Mohsenin, N. N. (1968). "Physical Properties of Plant and Animal Materials Part 1 of

Volume 1 Structure, Physical Characteristics and Rheological Properties". 2nd
Preliminary Ed, University Park, Pennsylvania: Department of Agricultural
Engineering, the Pennsylvania State. pp. 308.
Morrison, R. D. and Cleary, P. W. (2004). "Using DEM to Model Ore Breakage Within
a Pilot Scale SAG Mill". Minerals Engineering, Vol. 17, pp. 1117 - 1124.
Mustoe, G. G. W., Miyata, M. and Nakagawa, M. (2000). "Discrete Element Methods

for Mechanical Analysis of Systems of General Shaped Bodies". 5th International
Conference on Computational Structures Technology, Leuven, Belgium.
Niedzika, I. and Szymanek, M. (2005). "Influence of Storage Conditions of Sweet

Corn Cobs on Mechanical Properties of Kernel". Technica Agraria, Vol. 4, No. 1, pp. 3
- 11.
Nolan, G. T. and Kavanagh, P. E. (1995). "Random Packing of Nonspherical Particles".

Powder Technology, Vol. 84, No. 3, pp. 199 - 205.
Nordell, L. K. (2003). "Modern Ore Transfer Chute & Belt Feeder Designs Developed
from Discrete Element Modeling (DEM)". BELTCON 12, South Africa.
Page, J. (1991). "Examples of Good and Bad Chute Design". Chute Design Conference,
South Africa.
[Online, accessed 4th November].
315
REFERENCES
http://www.ckit.co.za/Secure/Conveyor/Papers/Bionic%20Research%201/F%20BRI1%
20Paper05.htm
Pandey, P., Song, Y., Kayihan, F. and Turton, R. (2006). "Simulation of Particle
Movement in a Pan Coating Device Using Discrete Element Modeling and its
Comparison with Video-Imaging Experiments". Powder Technology, Vol. 161, No. 2,
pp. 79 - 88.
Pettyjohn, E. S. and Christiansen, E. B. (1948). "Effect of Particle Shape on Free-

Settling Rates of Isometric Particles". Chemical Engineering Progress, Vol. 44, No. 2,
pp. 157 - 173.
Pinson, D., Reed, J., Wright, B. and Yu, A. B. (2004). "Application of Discrete Particle
Simulation to Flow in a Transfer Chute,". 8th International Conference on Bulk
Materials Storage, Handling and Transportation, Wollongong, NSW, Australia, 5th - 8th
July, Institution of Engineers, Australia, pp. 294 - 298.
Pitcher, D. M. (1981). "Loading, Discharging and Cleaning of Belt Conveyors".

BELTCON 1, South Africa.
[Online, accessed 2nd May, 2006].
http://www.saimh.co.za/beltcon/Beltcon1/paper19.html
Potapov, A. V. and Campbell, C. S. (1998). "A Fast Model for the Simulation of Non-
Round Particles". Granular Matter, Vol. 1, No. 1, pp. 9 - 14.
Pournin, L., Weber, M., Tsukahara, M., Ferrez, J.-A., Ramaioli, M. and Liebling, T. M.
(2005). "Three-Dimensional Distinct Element Simulation of Spherocylinder
Crystallization". Granular Matter, Vol. 7, No. 2-3, pp. 119 - 126.
Qiu, X. and Kruse, D. (1997). "Analysis of Flow of Ore Materials in a Conveyor

Transfer Chute Using Discrete Element Method". Symposium on the Mechanics of
Particulate Materials, Evanston, Illinois, USA, ASCE, New York, pp. 395 - 404.
Read, G. (1985). "Air Supported Belt Conveyors". Bulk Solids Handling, Vol. 5, No. 5,
pp. 1071 - 1076.
Roberts, A. W. (1999). "Feeders and Transfers Recent Developments". Bulkex 1999,

Bulk Materials Handling Conference, Homebush Bay, Sydney, Australia, 29th June -
1st July, pp. 1.1 - 1.27.
Roberts, A. W. (2001). "Chute Design Considerations for Feeding and Transfer".

BELTCON 11, Johannesburg, South Africa.
http://www.saimh.co.za/beltcon/Beltcon11/Beltcon1103.htm
Roberts, A. W. (2003). "Chute Performance and Design for Rapid Flow Conditions".
Chemical Engineering and Technology, Vol. 26, No. 2, pp. 163 - 170.
Roberts, A. W., Wiche, S. J., Ilic, D. D. and Plint, S. R. (2004). "Flow Dynamics and
Wear Considerations in Transfer Chute Design". 8th International Conference on Bulk
316
REFERENCES
Materials Storage, Handling and Transportation, Wollongong, NSW, Australia, 5th - 8th
July, Institution of Engineers, Australia, pp. 330 - 334.
Rochling. (n.d.). "Polystone G Natural".

http://www.roechling.com/en/high-performance-
plastics/thermoplastics/materials/polystone.html
Roth, M. J. (2008). "Design Issues in Bulk Materials Handling", B.E. (Mech. Eng.)
Thesis, University of Wollongong, Wollongong, New South Wales, Australia. pp. 93.
Rozentals, J. (1983). "Rational Design of Conveyor Chutes". BELTCON 2,

Johannesburg, South Africa.
Rozentals, J. (1991). "Flow of Bulk Solids in Chute Design". Chute Design Conference,
South Africa.
[Online, accessed 2nd March, 2006].
http://www.ckit.co.za/Secure/Conveyor/Papers/Bionic%20Research%201/C%20BRI1%
20Paper02.htm
SAA (1991). "Bin Flow Properties of Coal". AS 3880 - 1991, Sydney, Standards
Association of Australia. pp. 20.
Sadd, M. H., Tai, Q. and Shukla, A. (1993). "Contact Law Effects on Wave Propagation
in Particulate Materials Using Distinct Element Modeling". International Journal of
Non-Linear Mechanics, Vol. 28, No. 2, pp. 251 - 265.
Sawley, M. L. and Cleary, P. W. (1999). "A Parallel Discrete Element Method for
Industrial Granular Flow Simulations". EPFL Supercomputing Review, No. 11, pp. 23
- 28.
Shelef, L. and Mohsenin, N. N. (1969). "Effect of Moisture Content on Mechanical

properties of Shelled Corn". Cereal Chemistry, Vol. 46, No. 3, pp. 242 - 253.
Shigley, J. E. (1986). "Mechanical Engineering Design: First Metric Edition",

McGraw-Hill. pp. 699.
Shiu, W.-J., Donze, F. V. and Magnier, S. A. (2006). "Numerical Study of Rockfalls on

Covered Galleries by the Discrete Element Method". The Electronic Journal of
Geotechnical Engineers, Vol. 11, No. D.
Singh, S. S., Finner, M. F., Rohatgi, P. K., Buelow, F. H. and Schaller, M. (1991).
"Structure and Mechanical Properties of Corn Kernels: A Hybrid Composite Material".
Journal of Materials Science, Vol. 26, No. 1, pp. 274 - 284.
Sitharam, T. G. (1999). "Micromechanical Modeling of Granular Materials: Effect of

Confining Pressure on Mechanical Behavior". Mechanics of Materials, Vol. 31, No. 10,
pp. 653 - 665.
317
REFERENCES
Snow, C. L. (1991). "Degradation Control in Transfer Chutes". Chute Design

Conference, South Africa.
[Online, accessed 29th September, 2005].
http://www.ckit.co.za/Secure/Conveyor/Papers/Bionic%20Research%201/K%20BRI1%
20Paper10.htm
Song, Y., Turton, R. and Kayihan, F. (2006a). "Contact Algorithms for Tablet-Shaped
Particles and their Application to DEM Simulation". 5th World Congress on Particle
Song, Y., Turton, R. and Kayihan, F. (2006b). "Contact Detection Algorithms for DEM
Simulations of Tablet-Shaped Particles". Powder Technology, Vol. 161, No. 1, pp. 32 -
40.
Straub, M., McNamara, S. and Herrmann, H. J. (2007). "Plug Conveying in a

Horizontal Tube". Granular Matter, Vol. 9, No. 1-2, pp. 35-48.
Stuart-Dick, D. and Royal, T. A. (1991). "Streamlined Design of Chutes to Handle Bulk

Solids". Chute Design Conference, South Africa.
http://www.ckit.co.za/Secure/Conveyor/Papers/Bionic%20Research%201/I%20BRI1%
20Paper08.htm
Swamee, P. K. and Ojha, C. S. (1991). "Drag Coefficient and Fall Velocity of

Nonspherical Particles". Journal of Hydraulic Engineering, Vol. 117, No. 5, pp. 660 -
667.
Thain, L. and Broumels, J. (1983). "Conveyor Belts on Air Beat the Idler Problem:
Aerobelt a Low Friction Bulk Conveying Concept". BELTCON 2, Johannesburg, South
Africa.
Theuerkauf, J., Dhodapkar, S., Manjunath, K., Jacob, K. and Steinmetz, T. (2003).
"Applying the Discrete Element Method in Process Engineering". Chemical
Engineering and Technology, Vol. 26, No. 2, pp. 157 - 162.
Tillack, M. S. and Zhang, J. D. (1999) "Particle Dynamic Simulation of Free Surface

Granular Flows". Technical Report UCSD-ENG-080, Center for Energy Research,
University of California, San Diego, USA.
[Online, accessed 3rd May 2006]
http://aries.ucsd.edu/LIB/REPORT/UCSD-ENG/UCSD-ENG-080.pdf
Tran-Cong, S., Gay, M. and Michaelides, E. E. (2004). "Drag Coefficients of

Irregularly Shaped Particles". Powder Technology, Vol. 139, No. 1, pp. 21 - 32.
Tsuji, Y., Kawaguchi, T. and Tanaka, T. (1993). "Discrete Particle Simulation of Two-
Dimensional Fluidised Bed". Powder Technology, Vol. 77, No. 1, pp. 79 - 87.
318
REFERENCES
Tsuji, Y., Tanaka, T. and Ishida, T. (1992). "Lagrangian Numerical Simulation of Plug
Flow of Cohesionless Particles in a Horizontal Pipe". Powder Technology, Vol. 71, No.
2, pp. 239 - 250.
Turton, R. and Clark, N. N. (1987). "An Explicit Relationship to Predict Spherical

Particle Terminal Velocity". Powder Technology, Vol. 53, No. 2, pp. 127 - 129.
Turton, R. and Levenspiel, O. (1986). "A Short Note on the Drag Correlation for
Spheres". Powder Technology, Vol. 47, No. 1, pp. 83 - 86.
van Liedekerke, P., Tijskens, E., Dintwa, E., Anthonis, J. and Ramon, H. (2006). "A
Discrete Element Model for Simulation of a Spinning Disc Fertilizer Spreader I. Single
Particle Simulations". Powder Technology, Vol. 170, No. 2, pp. 71 - 85.
Vu-Quoc, L., Zhang, X. and Walton, O. R. (2000). "A 3-D Discrete-Element Method for
Dry Granular Flows of Ellipsoidal Particles". Computer Methods of Applied
Mechanics and Engineering, Vol. 187, No. 3-4, pp. 483 - 528.
Wadell, H. (1935). "Volume, Shape and Roundness of Quartz particles". Journal of

Geology, Vol. 43, pp. 250 - 280.
Walde, M. (2003). "Coal Feed Systems Give Plants a Balanced Diet". Power, Vol. 147,
No. 4.
http://www.powermag.com/archive_article.asp?a=200Ad40FE7d22093yL903O_1&y=2
003&m=may
Walton, O. R. (1982a). "Explicit Particle Dynamics Model for Granular Materials". 4th
International Conference on Numerical Methods in Geomechanics, Edmonton, Alberta,
Canada, 31st May - 4th June, A. A. Balkema/Rotterdam, pp. 1261 - 1268.
Walton, O. R. (1982b). "Particle-Dynamics Calculations of Shear Flow". US/Japan

Seminar on New Models and Constitutive Relations in the Mechanics of Granular
Materials, Ithica, New York, USA, 23rd - 27th August, Elsevier, Amsterdam, pp. 327 -
338.
Walton, O. R. (1993). "Numerical Simulation of Inclined Chute Flows of Monodisperse,

Inelastic, Frictional Spheres". Mechanics of Materials, Vol. 16, No. 1-2, pp. 239 - 247.
Walton, O. R. and Braun, R. L. (1986). "Viscosity, Granular-Temperature, and Stress

Calculations for Shearing Assemblies of Inelastic, Frictional Disks". Journal of
Rheology, Vol. 30, No. 5, pp. 949 - 980.
Walton, O. R., Braun, R. L., Mallon, R. G. and Cervelli, D. M. (1988). "Particle-

Dynamics Calculations of Gravity Flow of inelastic, Frictional Spheres".
Micromechanics of Granular Materials. M. Satake and J. T. Jenkins (Ed.). Amsterdam,
Elsevier, pp. 153 - 161.
319
REFERENCES
Wang, C.-Y. and Liang, V.-C. (1997). "A Packing Generation Scheme for the Granular
Assemblies with Planar Elliptical Particles". International Journal for Numerical and
Analytical Methods in Geomechanics, Vol. 21, No. 5, pp. 347 - 358.
Weihan, W. and Fengnian, H. (1992). "Mechanical and Dielectric Assessment of

Ultrahigh Molecular Weight Polyethylene Insulation for Cryogenic Applications". IEE
Transactions on Electrical Insulation, Vol. 27, No. 3, pp. 504 - 512.
Winkelman, J. (n.d.). "New Coal Handling Transfer Chute Technologies that

Significantly Improve the Operation by Increasing Efficiencies, Reducing Maintenance
and Increasing Operating Margins".
[Online, accessed 24th Nov, 2006].
http://www.parramattagroup.com/files/CFMTS_White_Paper.doc
Wypych, P. W. (1993). "Mech 475 Pneumatic Conveying of Bulk Solids Lecture Notes",
Department of Mechanical Engineering, University of Wollongong.
Xiang, J. and McGlinchey, D. (2003). "Numerical Simulation of Particle Motion in

Dense Phase Pneumatic Conveying". Proceedings of the 4th International Conference
for Conveying and Handling of Particulate Solids Volume 2, Budapest, Hungary, 27th -
30th May, pp. 11.1 - 11.6.
Yang, R. Y., Zou, R. P. and Yu, A. B. (2000). "Computer Simulation of the Packing of
Fine Particles". Physical Review E, Vol. 62, No. 3, pp. 3900 - 3908.
Zhang, J., Hu, Z., Ge, W., Zhang, Y., Li, T. and Li, J. (2004). "Application of the
Discrete Approach to the Simulation of Size Segregation in Granular Chute Flow".
Industrial and Engineering Chemistry Research, Vol. 43, No. 18, pp. 5521 - 5528.
Zhang, X. and Vu-Quoc, L. (2000). "Simulation of Chute Flow of Soybeans Using an

Improved Tangential Force-Displacement Model". Mechanics of Materials, Vol. 32, pp.
115 - 129.
Zhao, D., Nezami, E. G., Hashash, Y. M. A. and Ghaboussi, J. (2006). "Three-

Dimensional Discrete Element Simulation for Granular Materials". Engineering
Computations: International Journal for Computer-Aided Engineering and Software,
Vol. 23, No. 7, pp. 749-770.
Zhou, Y. C., Wright, B. D., Yang, R. Y., Xu, B. H. and Yu, A. B. (1999). "Rolling
Friction in the Dynamic Simulation of Sandpile Formation". Physica A, Vol. 269, pp.
536 - 553.
Zhu, H. P. and Yu, A. B. (2003). "The Effects of Wall and Rolling Resistance on the
Couple Stress of Granular Materials in Vertical Flow". Physica A, Vol. 325, pp. 347 -
360.
Zhu, H. P. and Yu, A. B. (2004). "Steady-State Granular Flow in a Three-Dimensional

Cylindrical Hopper with Flat Bottom: Microscopic Analysis". Journal of Physics D:
Applied Physics, Vol. 37, pp. 1497 - 1508.
320
REFERENCES
321
BIBLIOGRAPHY
BIBLIOGRAPHY
322
BIBLIOGRAPHY
Adams, G. G. (1997). "Imperfectly Constrained Planar Impacts - A Coefficient-of-

Restitution Model". International Journal of Impact Engineering, Vol. 19, No. 8, pp. 693
- 701.
Alger, G. R. and Simons, D. B. (1968). "Fall Velocity of Irregular Shaped Particles".

Journal of the Hydraulic Division Proceedings of the American Society of Civil
Engineers, Vol. 94, No. 3, pp. 721 - 738.
Arnold, P. C. (1993). "Transfer Chutes Engineered for Reliable Performance". National

Conference on Bulk Materials Handling, Yeppoon, Queensland, Australia, 22 - 25
September, Institution of Engineers Australia, pp. 165 - 173.
Conveyors". 3rd International Conference on Bulk Materials Storage, Handling and
Transportation, Newcastle, Australia, 27 - 29 June, pp. 131 - 135.
Arnold, P. C. and Hill, G. L. (1990). "Feeding Coal to and from Belt Conveyors".
International Coal Engineering Conference, Sydney, Australia, 19 - 21 June, Institution
of Engineers Australia, pp. 47 - 54.
Conveyors". Chute Design Conference.
[Online, accessed 11 November 2005].
http://www.ckit.co.za/Secure/Conveyor/Papers/Bionic%20Research%201/D%20BRI1%
20Paper03.htm
Arsenijevic, Z. L., Grbavcic, Z. B., Garic-Grulovic, R. V. and Zdanski, F. K. (1999).

"Determination of Non-Spherical Particle Terminal Velocity Using Particulate
Expansion Data". Powder Technology, Vol. 103, No. 3, pp. 265 - 273.
Benjamin, C. W., Berleigh, A. C. and Nemeth, J. (1999). "Transfer Chute Design A

New Approach Using 3D Parametric Modelling". Bulkex 1999 Bulk Materials
Handling Conference, Homebush Bay, Sydney, Australia, 29th June - 1st July, pp. 8.1 -
8.9.
Benjamin, C. W. and Nemeth, J. (2001). "Transfer Chute Design for Modern Materials
Handling Operations". Bulk Solids Handling, Vol. 21, No. 1, pp. 31 - 34.
Bi, W., Delanney, R., Richard, P., Taberlet, N. and Valance, A. (2005). "2 and 3D
Confined Granular Chute Flows - Experimental and Numerical Results". Journal of
Physics: Condensed Matter, Vol. 17, No. 24, pp. S2457-S2480.
Bierwisch, C., Kraft, T., Riedel, H. and Moseler, M. (2009). "Three-Dimensional

Discrete Element Models for the Granular Statics and Dynamics of Powders in Cavity
Filling". Journal of the Mechanics and Physics of Solids, Vol. 57, No. 1, pp. 10 - 31.
Cagli, A. S., Deveci, B. N., Okutan, C. H., Sirkeci, D. A. A. and Teoman, E. Y. (2007).
"Flow Property Measurement Using the Jenike Shear Cell for 7 Different Bulk Solids".
European Congress of Chemical Engineering (ECCE-6), Copenhagen, Denmark, 16 -
20 September.
323
BIBLIOGRAPHY
Cambalik, J. J., Checkley Jr, D. M. and Kamykowski, D. (1998). "A New Method to
Measure the Terminal Velocity of Small Particles: A Demonstration Using Ascending
Eggs of the Atlantic Menhaden (Brevoortia Tyrannus)". Limnology and Oceanography,
Vol. 43, No. 7, pp. 1722 - 1727.
Charlton, W. H., Chiarella, C. and Roberts, A. W. (1975). "Gravity Flow of Granular

Materials in Chutes: Optimising Flow Properties". Journal of Agricultural Engineering
Research, Vol. 20, pp. 39 - 45.
Charlton, W. H. and Roberts, A. W. (1970). "Chute Profile for Maximum Exit Velocity
in Granular Flow of Gravity Material". Journal of Agricultural Engineering Research,
Vol. 15, No. 3, pp. 292 - 294.
Cheng, N.-S. (2009). "Comparison of Formulas for Drag Coefficient and Settling
Velocity of Spherical Particles". Powder Technology, Vol. 189, No. 3, pp. 395 - 398.
Cleary, P. W. (2008). "The Effect of Particle Shape on Simple Shear Flows". Powder
Technology, Vol. 179, No. 3, pp. 144-163.
Concha, F. and Almendra, E. R. (1979). "Settling Velocities of Particulate Systems, 1.

Settling Velocities of Individual Spherical Particles". International Journal of Mineral
Processing, Vol. 5, No. 4, pp. 349 - 367.
Cooper, W. M. and Pollett, W. F. O. (1969). "The Shear Modulus of Flowing

Polyethylene Melts". Journal of Applied Polymer Science, Vol. 13, No. 11, pp. 2313 -
2324.
Coskun, M. B., Yalcin, I. and Ozarslan, C. (2006). "Physical Properties of Sweet Corn
Seed (Zea Mays Saccharata Sturt.)". Journal of Food Engineering, Vol. 74, No. 4, pp.
523 - 528.
Cross, R. (2002). "Measurements of the Horizontal Coefficient of Restitution for a

Superball and a Tennis Ball". American Journal of Physics, Vol. 70, No. 5, pp. 482 -
489.
Gabitto, J. and Tsouris, C. (2008). "Drag Coefficient and Settling Velocity for Particles
of Cylindrical Shape". Powder Technology, Vol. 183, No. 2, pp. 314 - 322.
Garg, O. P. (1982). "Recently Developed Blasting Techniques in Frozen Iron Ore at

Schefferville, Quebec". Proceedings of the 4th Canadian Permafrost Conference,
Calgary, Alberta, Canada, 2 - 6 March, 1981, Associate Committee on Geotechnical
Research, National Research Council of Canada, pp. 594.
Gul, R. M. and McGarry, F. J. (2004). "Processing of Ultra-High Molecular Weight

Polyethylene by Hot Isostatic Pressing, and the Effect of Processing Parameters on its
Microstructure". Polymer Engineering and Science, Vol. 44, No. 10, pp. 1848 - 1857.
Hahn, B. and Valentine, D. T. (2007). "Essential MATLAB for Engineers and

Scientists". 3rd Ed, Butterworth-Heinemann. pp. 448.
324
BIBLIOGRAPHY
Hartman, M., Trnka, O. and Svoboda, K. (1994). "Free Settling of Nonspherical

Particles". Industrial and Engineering Chemistry Research, Vol. 33, No. 8, pp. 1979 -
1983.
Herrmann, H. J. (1999). "The Importance of Computer Simulations of Granular Flow".

Computing in Science and Engineering, No. January - February, pp. 72 - 73.
Holzer, A. and Sommerfeld, M. (2008). "New Simple Correlation Formula for the Drag
Coefficient of Non-Spherical Particles". Powder Technology, Vol. 184, No. 3, pp. 361 -
365.
Hunt, B. R., Lipsman, R. L., Rosenberg, J. M., Coombes, K. R., Osborn, J. E. and
Stuck, G. J. (2001). "A Guide to MATLAB for Beginners and Experienced Users". New
York, Cambridge University Press. pp. 327.
Huque, S. T. and Walsh, M. A. (2004). "Material Trajectories: Distinct Element

Modelling (DEM) vs. C.E.M.A". 8th International Conference on Bulk Materials
Storage, Handling and Transportation, Wollongong, NSW, Australia, 5th - 8th July,
Institution of Engineers, Australia, pp. 322 - 329.
Huque, S. T. and Wypych, P. W. (2004). "An Overview of Belt Conveyor Transfers".

8th International Conference on Bulk Materials Storage, Handling and Transportation,
Wollongong, NSW, Australia, 5th - 8th July, Institution of Engineers, Australia, pp. 283
- 288.
Hustrulid, A. I. (1998). "Transfer Station Analysis". Bulk Materials Handling by

Conveyor Belt II, Orlando, Florida, USA, pp. 33 - 53.
Hustrulid, A. I. and Mustoe, G. G. W. (1996). "Engineering Analysis of Transfer Points

Using Discrete Element Analysis". Bulk Materials Handling by Conveyor Belt, pp. 9 -
13.
Kiusalaas, J. (2005). "Numerical Methods in Engineering with MATLAB". New York,

Cambridge University Press. pp. 426.
Knight, A. (2000). "Basics of MATLAB and Beyond", Chapman & Hall/CRC. pp. 216.
Li, T., Zhang, J. and Ge, W. (2004). "Simple Measurement of Restitution Coefficient of
Irregular Particles". China Particuology, Vol. 2, No. 6, pp. 274 - 275.
Liu, K. and Tanimura, S. (1997). "Numerical Analysis for Dynamic Stress

Concentration in a Rectangular Block Due to Impact". International Journal of Impact
Engineering, Vol. 19, No. 8, pp. 653 - 66.
Losenno, C. and Easson, W. J. (2002). "PIV Measurement of Free-Falling Irregular

Particles". 11th International Symposium Applications of Laser Techniques to Fluid
Mechanics, Lisbon, Portugal, 8 - 11 July, pp. 8.
Loth, E. (2008). "Drag of Non-Spherical Solid Particles of Regular and Irregular

Shape". Powder Technology, Vol. 182, No. 3, pp. 342 - 353.
325
BIBLIOGRAPHY
Low, A. G. L. and Verran, M. J. (2000). "Physical Modelling of Transfer Chutes Case

Studies and Applications in Bulk Materials Handling". ASBSH Conference at Bulkex
2000, Melbourne, Australia, 31 Oct - 2 Nov, pp. 4.1 - 4.7.
Marshall, D. W. (2007). "Analytical Equation for Estimating Terminal Velocities of

Spheroidal Particles". Journal of the Idaho Academy of Science, Vol. 43, No. 1, pp. 17
- 22.
McBride, W. and Ilic, D. (2007). "Design of Self Compensating Soft Loading Hood".
Particle and Particle Systems Characterization, Vol. 24, No. 4-5, pp. 370-374.
McIlvenna, P. and Mossad, R. (2003). "Two Dimensional Transfer Chute Analysis

Using a Continuum Method". 3rd International Conference on CFD in the Minerals and
Process Industries, CSIRO, Melbourne, Australia, 10 - 12 December, pp. 547 - 551.
Mohsenin, N. N. (1968). "Physical Properties of Plant and Animal Materials Part 2 of

Volume 1 Texture of Foods, Mechanical Damage, Aero and Hydrodynamic
Characteristics and Frictional Properties". 1st Preliminary Ed, University Park,
Pennsylvania: Department of Agricultural Engineering, the Pennsylvania State. pp. 757.
Morgan, D. C. (1991). "Chute Design in a Large Coal Handling Facility". Chute

Design Conference, South Africa.
http://www.ckit.co.za/Secure/Conveyor/Papers/Bionic%20Research%201/B%20BRI1%
20Paper01.htm
Mustoe, G. G. W. and Miyata, M. (2001). "Material Flow Analyses of Non-Circular

Shaped Granular Media using Discrete Element Methods". Journal of Engineering
Mechanics, Vol. 127, No. 10, pp. 1017 - 1026.
Nguyen, A. V., Stechemesser, H., Zobel, G. and Schulze, H. J. (1997). "An Improved
Formula for Terminal Velocity of Rigid Spheres". International Journal of Mineral
Processing, Vol. 50, No. 1 - 2, pp. 53 - 61.
Nordell, L. K. (1997). "Particle Flow Modeling: Transfer Chutes & Other

Applications". BELTCON 9, South Africa.
O'Donovan, E. (2003). "Predictable Transfers at Last! Without Trial and Error".

Australian Bulk Handling Review, Vol. 8, No. 1, pp. 51 - 52.
O'Sullivan, C. and Bray, J. D. (2005). "The Importance of Accurately Capturing

Particle Geometry in DEM Simulations". Powders and Grains Volume 2, Taylor and
Francis Group, London, pp. 1333 - 1337.
Otto, S. R. and Denier, J. P. (2005). "An Introduction to Programming and Numerical

Methods in MATLAB", Springer-Verlag London. pp. 464.
Patel, J. and Phillips, P. J. (1973). "The Young's Modulus of Polyethylene". Journal of

Polymer Science: Polymer Letters Editions, Vol. 11, No. 12, pp. 771 - 776.
326
BIBLIOGRAPHY
Ramos, C. M. and Goodwin, P. J. (1987). "Degradation of Sized Coal at Transfer

Points". BELTCON 4, South Africa.
Razavi, S. M. A., Rafe, A. and Akbari, R. (2007). "Terminal Velocity of Pistachio Nut
and its Kernel as Affected by Moisture Content and Variety". African Journal of
Agricultural Research, Vol. 2, No. 12, pp. 663 - 666.
Roberts, A. W. (1967). "The Dynamics of Granular Material Flow Through Curved

Chutes". Mechanical and Chemical Engineering Transactions, Institution of Engineers
Australia, Vol. MC3, pp. 216 - 224.
Roberts, A. W. (1969). "An Investigation of the Gravity Flow of Noncohesive Granular

Materials Through Discharge Chutes". Journal of Engineering for Industry, ASME
Transactions, Vol. 91, No. Series B, pp. 373 - 381.
Roberts, A. W. (2000). "Developments in Belt Conveying Technology and Associated

Feeding of Bulk Solids". From Powder to Bulk: International Conference on Powder and
Bulk Solids Handling, London, UK, 13 - 15 June, Professional Engineering Publishing
Limited, pp. 377 - 410.
Roberts, A. W. and Scott, O. J. (1981). "Flow of Bulk Solids Through Transfer Chutes
of Variable Geometry and Profile". Bulk Solids Handling, Vol. 1, No. 4, December, pp.
715 - 727.
Rojek, J., Zarate, F., de Saracibar, C. A., Gilbourne, C. and Verdot, P. (2005). "Discrete
Element Modelling and Simulation of Sand Mould Manufacture for the Lost Foam
Process". International Journal for Numerical Methods in Engineering, Vol. 62, No. 11,
pp. 1421 - 1441.
Rozentals, J. J. and Stahura, R. P. (1992). "Chute Problems Causes and Solutions".

Chute Design Conference, South Africa.
http://www.ckit.co.za/Secure/Conveyor/Papers/Bionic%20Research%202/A%20BRI2%
20Paper01.htm
SAA (1992). Fixed Platforms, Walkways, Stairways and Ladders - Design,

Construction and Installation. AS 1657 - 2002, Standards Association of Australia,
Sydney.
SAI (2000). Conveyors - Safety Requirements. AS 1755 - 2000, Standards Australia

International Limited, Sydney.
Savage, S. B. (1979). "Gravity Flow of Cohesionless Granular Materials in Chutes and

Channels". Journal of Fluid mechanics, Vol. 92 Part 1, pp. 53 - 96.
Savage, S. B. and Lun, C. K. K. (1988). "Particle Size Segregation in Inclined Chute

Flow of Dry Cohesionless Granular Solids". Journal of Fluid mechanics, Vol. 189, pp.
311 - 335.
327
BIBLIOGRAPHY
Schulze, D. (2008). "Powders and Bulk Solids Behavior, Characterisation, Storage and
Flow", Springer-Verlag. pp. 511.
Seifried, R., Schiehlen, W. and Eberhard, P. (2005). "Numerical and Experimental

Evaluation of the Coefficient of Restitution for Repeated Impacts". International Journal
of Impact Engineering, Vol. 32, No. 1 - 4, pp. 508 - 524.
Soodak, H. and Tiersten, M. S. (1996). "Perturbation Analysis of Rolling Friction on a

Turntable". American Journal of Physics, Vol. 64, No. 9, pp. 1130 - 1139.
Sturos, J. A. (1972). "Determining the Terminal Velocity of Wood and Bark Chips".
U.S. Forest Service - Research Note NC-131, pp. 1 - 4.
Sundstrom, P. and Benjamin, C. W. (1993). "Innovations in Transfer Chute Design".

National Conference on Bulk Materials Handling, Yeppoon, Queensland, Australia, 22
25 September, IEAust, pp. 191 - 195.
Tykhoniuk, R., Tomas, J. and Luding, S. (2003). "Shear Dynamics Simulations of High-
Disperse Cohesive Powder". Particulate Systems Analysis (PSA) 2003, Harrogate,
United Kingdom, pp. 5.
Wu, W. (2006). "Simulation of Particle Flow Using the Discrete Element Method", B.E.
Thesis, The University of Queensland, Australia. pp. 170.
Xu, J. Q., Zou, R. P. and Yu, A. B. (2006). "Quantification of the Mechanisms

Governing the Packing of Iron Ore Fines". Powder Technology, Vol. 169, No. 2, pp. 99
- 107.
Yu, A. B. (2003). "Discrete Particle Simulation of Granular Flow". 3rd International

Conference on CFD in the Minerals Process Industries, CSIRO, Melbourne, Australia,
10 - 12 December.
Zhou, Y. C., Xu, B. H., Yu, A. B. and Zulli, P. (2002). "An Experimental and
Numerical Study of the Angle of Repose of Course Spheres". Powder Handling &
Processing, Vol. 125, No. 1, pp. 45 - 54.
328
APPENDIX A PUBLICATION LIST
APPENDIX
A
PUBLICATION LIST
329
The following is a list of the publications published and/or presented as a result of the
research carried out during this PhD.
A1 CONFERENCE PAPERS
1. Hastie, D. B. and Wypych, P. W. (2006). Conveyor Trajectory Prediction

Methods A Review. Proceedings of BULKEX 2006 Conference on CD,
Melbourne, Australia, 25 27 September.
2. Hastie, D. B. and Wypych, P. W. (2007). The Prediction of Conveyor

Trajectories. Proceedings of BELTCON 14 International Material Handling
Conference, East Rand, Johannesburg, South Africa, 1 2 August.
3. Hastie, D. B. and Wypych, P. W. (2007). Conveyor Trajectory Discharge

Angles. Proceedings of the 9th International Conference on Bulk Materials
Storage, handling and Transportation (ICBMH07), Newcastle, Australia, 9 11
October.
4. Hastie, D. B., Wypych, P. W. and Arnold, P. C. (2007). The Profile of Conveyor

Trajectories. Proceedings of the 9th International Conference on Bulk Materials
Storage, handling and Transportation (ICBMH07), Newcastle, Australia, 9 11
October.
5. Hastie, D. B., Grima, A. P. and Wypych, P. W. (2007). Validation of Particle

Flow Through a Conveyor Transfer Hood Via Particle Velocity Analysis.
Proceedings of the 4th International Symposium - Reliable Flow of Particulate
Solids (RelPowFlo) IV, Tromso, Norway, 10 12 June.
6. Hastie, D. B., Grima, A. P. and Wypych, P. W. (2008). Validation of Particle

Flow Through a Conveyor Transfer Spoon Via Particle Velocity Analysis.
Proceedings of Bulk Europe 2008, Prague, Czech Republic, 11 12 September.
7. Hastie, D. B., Grima, A. P. and Wypych, P. W. (2008). Modelling and Design of

Complete Conveyor Transfers, Conveyors in Mining 2008, Perth, Australia, 11
12 November.
8. Hastie, D. B., Grima, A. P. and Wypych, P. W. (2008). Validated Computer

Simulation Modelling for Complete Conveyor Transfer Design, Innovation in
Bulk Materials Handling and Processing 2008, Sydney, Australia, 26 27
November.
9. Hastie, D. B., Grima, A. P. and Wypych, P. W. (2008). Validated Dynamic

Simulation Modelling of Particulates, 1st China International Symposium on
particles Technology (CISPT 2008), Shanghai, China, 8 11 December.
10. Hastie, D. B. and Wypych, P. W. (2009). Evaluation of Belt Conveyor

Trajectories, 6th International Conference for Conveying and Handling of
Particulate Solids (CHoPS 2009), Brisbane, Australia, 4 7 August, Proceedings
on USB drive, pp. 338 - 344.
330
11. Hastie, D. B. and Wypych, P. W. (2009). Comparison between Experimental and

Predicted Conveyor Belt Trajectories. Proceedings of BELTCON 15 International
Material Handling Conference, East Rand, Johannesburg, South Africa, 2 3
September.
A2 JOURNAL PAPERS
12. Hastie, D. B., Wypych, P. W. and Arnold, P. C. (2010). Influences on the

Prediction of Conveyor Trajectory Profiles. Particulate Science and Technology,
Vol. 28, No. 2, pp. 132 - 145.
13. Hastie, D. B. and Wypych, P. W. (2010). Validation of Particle Flow Through

Conveyor Transfer Hoods Via Continuum and Discrete Element Methods.
Mechanics of Materials, Vol. 42, No. 4, pp. 383 - 394.
14. Hastie, D. B. and Wypych, P. W. (2010). The Validity of Conveyor Belt

Trajectories when Comparing Predicted to Experimentally Obtained Results.
Submitted for peer review to Bulk Solids & Powder Science & Technology, 16
March, 2010.
A3 OTHER PUBLICATIONS
15. Hastie, D. B. and Wypych, P. W. (2007). The Prediction of Conveyor Trajectories

A Review, Australian Bulk Handling Review, Vol. 12, No. 2, pp. 56 - 65.
16. Wypych, P. W., Hastie, D. B. and Grima, A. P. (2008). Validated Computer

Simulation Modelling for Complete Conveyor Transfer Design, Australian Bulk
Handling Review, Vol. 13, No. 6, pp. 36-38.
17. Hastie, D. B. (2009). Modelling and Designing Complete Conveyor Transfers,

BULKEX Academy Presentation, BULKEX 2009 Bulk Materials and Powder
Handling Exhibition, Brisbane, Australia, 4 6 August.
A4 POSTER PRESENTATIONS
18. Hastie, D. B. and Wypych, P. W. (2007). Validated DEM Simulation of Particle

Flows and Mechanisms, International Fine Particle Research Institute (IFPRI)
Annual General Meeting, Poster Display, Perth, Australia, 11 July.

Flows and Mechanisms, Higher Degree Research Conference, University of
Wollongong, Poster Display, Wollongong, Australia, 11 September.

Flows and Mechanisms, 6th International Conference for Conveying and Handling
of Particulate Solids (CHoPS 2009), Brisbane, Australia, 4 7 August.
331
APPENDIX B DETAILED CAD DRAWINGS OF THE TRANSFER HOOD AND SPOON
APPENDIX
B
DETAILED CAD DRAWINGS OF THE
TRANSFER HOOD AND SPOON
332
Figure B1 Hood and spoon elevation view
333
Figure B2 Hood and spoon isometric and sectional view
334
Figure B3 Transfer hood
335
Figure B4 Transfer hood Polystone Ultra liner
336
Figure B5 Transfer spoon
337
Figure B6 Transfer spoon Polystone Ultra liner
338
APPENDIX C MATLAB TRANSFER HOOD M-file
APPENDIX
C
MATLAB TRANSFER HOOD M-file
339
%file name Hood_Velocity_Analysis.m

%Evaluates the velocity of the material flow through the hood from
Chute
%Maven and EDEM
%Written by Andrew Grima 2007
%Modified by David Hastie March 2008 using Matlab R2008a
clf
clc
clear all
% Data file format, note must have numbers to 2 decimal places

% Col 1 = id
% Col 2 = x
% Col 3 = y
% Col 4 = z
% Col 5 = Vx
% Col 6 = Vy
% Col 7 = Vz
% -----INPUT DATA-----
% NOTE: The axis arrangement of the DXF files which are imported into
% Chute Maven and EDEM have X as the x-axis, Y as the z-axis and Z as
% the y-axis
load vb2_2tph_hood.txt
data=vb2_2tph_hood;
xi=-446.21; % mm, X coordinate of the exit of the hood (at
% horizontal)
zi=-364.86; % mm, Y coordinate of the exit of the hood (at
% horizontal)
rad=644; % mm, radius of the inside surface of the

% Polystone Ultra
tol=3; % mm, the tolerance around each angular position
% where particles are selected
width=50; % mm, the width either side of the central axis
ndiv=350; % number of divisions between 0 and 70 degrees
ind=70/ndiv; % deg, angular step around hood
deg=0:ind:70; % deg, max=70 deg for hood
xdis=rad*cosd(deg);
ydis=rad*sind(deg);
dx=rad-xdis;
if xi<0 % Determines the coordinates of the hood at each

% degree increment
coordx=xi+dx;
end
if xi>0
coordx=xi-dx;
end
coordy=zi+ydis;
% Plot the experimental data points on a graph, along with the

% representation of the hoof from the calculations above
figure(1) % check to make sure hood is in right position

plot(coordx,coordy,'LineWidth',1.5);grid on;zoom on;
340
xlabel('X Displacement (mm)');

ylabel('Y Displacement (mm)');
title('Transfer Hood Schematic','Color','b','FontSize',12);
hold on
scatter(data(:,2),data(:,4),2,'filled');grid on;zoom on;axis ([-600
200 -500 300]);axis square;
k=size(data,1);
velhood = zeros(1,20000);
count1=0;
for j =1:1:ndiv+1; % This code searches through the selected
% coordinates and returns a velocity value
for i = 1:1:k;
if ((abs(data(i,2)-coordx(1,j))<=tol) && (abs(data(i,4)-
coordy(1,j))<=tol) && abs(data(i,3))<=width )
count1=count1+1;
velhood(j,count1)=sqrt(data(i,5)^2+data(i,6)^2+data(i,7)^2)/1000;
end
end
end
magy=size(velhood,1); % determines the average velocity for

% each divx- stage 1
kk=size(velhood,2);
magx=zeros(magy,1);
for i=1:1:magy
for j= 1:1:kk
if(velhood(i,j)>0)
magx(i,1)=magx(i,1)+1;
end
end
end
velsum=sum(velhood,2); % determines the average velocity for

% each divx - stage 2
velav=magx.\velsum;
% The nested FOR loops below determine the x position of the

% first and last non-zero values in the velhood array.
first_value_position=zeros(magy,1);
last_value_position=zeros(magy,1);
count2=zeros(magy,1);
for i=1:1:magy
count2(i)=0;
count3(i)=0;
for j= 1:1:kk
if(velhood(i,j)==0)
count2(i)=count2(i)+1;
elseif(velhood(i,j)~=0)
first_value_position(i)=count2(i)+1;
end
last_value_position(i)=count3(i)+first_value_position(i)-1;
end
end
341
% The first and last positions found in the above nested FOR
% loops are used to determine the minimum and maximum velocity values
% for each angular position around the hood
velmin=zeros(magy,1);
velmax=zeros(magy,1);
for i=1:1:magy
velmin(i)=velhood(i,first_value_position(i));
velmax(i)=velhood(i,first_value_position(i));
for j=first_value_position(i):1:last_value_position(i)
if(velhood(i,j)<velmin(i))
velmin(i)=velhood(i,j);
elseif(velhood(i,j)>velmax(i))
velmax(i)=velhood(i,j);
end
end
end
% Calculate the standard deviation of the minimum, average and maximum

% velocities for each angular position.
stdevmin = std(velmin);
stdevav = std(velav);
stdevmax = std(velmax);
% Plot the average velocity on a graph versus the angular position

% around the hood. Also on the same graph (using the HOLD function)
% plot the maximum and minimum values
figure(2)
plotx=0:ind:(magy-1)*ind;
xplot=plotx';
scatter(plotx,velav,'.r');grid on;zoom on;axis ([0 50 0 4]);
hold on;
scatter(plotx,velmax,'.b');grid on;zoom on;axis ([0 50 0 4]);

scatter(plotx,velmin,'.g');grid on;zoom on;axis ([0 50 0 4]);
xlabel('Angle (deg)');
ylabel('Velocity (m/s)');
title('Velocity of Particles Through Transfer
Hood','Color','b','FontSize',12);
hold off;
342
APPENDIX D MATLAB TRANSFER SPOON M-file
APPENDIX
D
MATLAB TRANSFER SPOON M-file
343
% file name Spoon_Velocity_Analysis.m

% Evaluates the velocity of the material flow through the spoon from
% Chute Maven and EDEM
% Written by David Hastie March 2008 using Matlab R2008a
clf
clc
clear all
% Data file format, note must have numbers to 2 decimal places

% Col 1 = id
% Col 2 = x
% Col 3 = y
% Col 4 = z
% Col 5 = Vx
% Col 6 = Vy
% Col 7 = Vz
% ----- INPUT DATA -----

% NOTE: The axis arrangement of the DXF files which are imported into
% Chute Maven and EDEM have X as the x-axis, Y as the z-axis and Z as
% the y-axis
load vb2_2tph_spoon.txt
data=vb2_2tph_spoon;
xi=-382.07; % mm, X coordinate of the exit of the spoon (at
% horizontal)
zi=-611.44; % mm, Y coordinate of the exit of the spoon (at
% horizontal)
theta1=14.5; % deg, angle that top of spoon makes with the
horizontal
% ----- CONSTANTS -----

phi1=27; % deg, inclusive angle of spoon radius 1 segment
phi2=38; % deg, inclusive angle of spoon radius 2 segment
theta2=theta1+phi1; % deg, angle radius 2 makes with horizontal
rad1=1144; % mm, radius 1 of the inside surface of the

% Polystone Ultra
rad2=654; % mm, radius 2 of the inside surface of the
% Polystone Ultra
start1=2; % deg, angular position from top of spoon for
% first measurement
tol=3; % mm, the tolerance around each angular position

where
% the particles are selected to be analysed
width=50; % mm, the width either side of the central axis
% ----- CALCULATIONS -----
rad1originx=xi-rad1*cosd(theta1);
rad1originy=zi+rad1*sind(theta1);
number_of_positions=phi1+phi2+1;
% Calculate the X and Y coordinates for all angular positions as

% declared in the CONSTANTS section above
count1a=1;
344
for m = (theta1):1:(theta1+phi1);
coordx(count1a)=rad1originx+rad1*cosd(m);
coordy(count1a)=rad1originy-rad1*sind(m);
count1a=count1a+1;
end
count1b=count1a;
rad2originx=rad1originx+(rad1-rad2)*cosd(theta1+phi1);
rad2originy=rad1originy-(rad1-rad2)*sind(theta1+phi1);
for n = (theta2+1):1:(theta2+phi2);
coordx(count1b)=rad2originx+rad2*cosd(n);
coordy(count1b)=rad2originy-rad2*sind(n);
count1b=count1b+1;
end
% Flip the X and Y coordinates so that the analysis can start from the
% exit of the spoon and progress up to the impact point
coordx = fliplr(coordx);
coordy = fliplr(coordy);
xcoord = coordx';
ycoord = coordy';
% Plot the experimental data points on a graph, along with the

% representation of the spoon from the calculations above
figure(1) % check to make sure spoon is in right position

plot(coordx,coordy,'LineWidth',1.5);grid on;zoom on;
xlabel('X Displacement (mm)');

ylabel('Y Displacement (mm)');
title('Transfer Spoon Schematic','Color','b','FontSize',12);
hold on
scatter(data(:,2),data(:,4),2,'filled');grid on;zoom on;axis ([-1200 -
200 -1400 -400]);axis square;
% end of figure 1
k=size(data,1);
velspoon = zeros(1,10000);
count2=0;
for j =1:1:number_of_positions; % This code searches through the

selected coordinates and returns a velocity value
for i = 1:1:k;
if ((abs(data(i,2)-coordx(1,j))<=tol) && (abs(data(i,4)-
coordy(1,j))<=tol) && abs(data(i,3))<=width )
count2=count2+1;
velspoon(j,count2)=sqrt(data(i,5)^2+data(i,6)^2+data(i,7)^2)/1000;
end
end
end
velspoon=flipud(velspoon);
345
magy=size(velspoon,1); % determines the average velocity for

% each divx- stage 1
kk=size(velspoon,2);
magx=zeros(magy,1);
for i=1:1:magy
for j= 1:1:kk
if(velspoon(i,j)>0)
magx(i,1)=magx(i,1)+1;
end
end
end
velsum=sum(velspoon,2); % determines the average velocity for

% each divx - stage 2
velav=magx.\velsum;
% The nested FOR loops below determine the x position of the

% first and last non-zero values in the velspoon array.
first_value_position=zeros(magy,1);
last_value_position=zeros(magy,1);
for i=1:1:magy
count3(i)=0;
count4(i)=0;
for j= 1:1:kk
if(velspoon(i,j)==0)
elseif(velspoon(i,j)~=0)
first_value_position(i)=count3(i)+1;
end
last_value_position(i)=count4(i)+first_value_position(i)-1;
end
end
% The first and last positions found in the above nested FOR
% loops are used to determine the minimum and maximum velocity values
% for each angular position around the spoon
velmin=zeros(magy,1);
velmax=zeros(magy,1);
for i=1:1:magy
velmin(i)=velspoon(i,first_value_position(i));
velmax(i)=velspoon(i,first_value_position(i));
for j=first_value_position(i):1:last_value_position(i)
if(velspoon(i,j)<velmin(i))
velmin(i)=velspoon(i,j);
elseif(velspoon(i,j)>velmax(i))
velmax(i)=velspoon(i,j);
end
end
end
% Calculate the standard deviation of the minimum, average and maximum

% velocities for each angular position.
346
stdevmin = std(velmin);
stdevav = std(velav);
stdevmax = std(velmax);
% Plot the average velocity on a graph versus the angular position

% around the spoon. Also on the same graph (using the HOLD function)
% plot the maximum and minimum values
start_plot_angle=(number_of_positions-magy)+theta1;
disp('start plot angle');
disp(start_plot_angle);
figure(2)
plotx=start_plot_angle:1:(magy+start_plot_angle-1);
% Transpose X coordinates to allow curve fitting
xplot=plotx';
% Fit polylines to the data
p=polyfit(xplot,velav,3);
f=polyval(p,xplot);
disp('Polynomial coefficients for average velocity: - ');
disp(p);
q=polyfit(xplot,velmin,3);
g=polyval(q,xplot);
disp('Polynomial coefficients for minimum velocity: - ');
disp(q);
r=polyfit(xplot,velmax,3);
h=polyval(r,xplot);
disp('Polynomial coefficients for maximum velocity: - ');
disp(r);
scatter(xplot,velav,'.r');grid on;zoom on;axis ([0 90 0 5]);
hold on;
plot(xplot,f,'LineWidth',1.5)
plot(xplot,g,'LineWidth',1.5)
plot(xplot,h,'LineWidth',1.5)
scatter(plotx,velmax,'.b');grid on;zoom on;axis ([0 90 0 5]);

scatter(plotx,velmin,'.g');grid on;zoom on;axis ([0 90 0 5]);
xlabel('Angle (deg)');
ylabel('Velocity (m/s)');
title('Velocity of Particles Through the Transfer
Spoon','Color','b','FontSize',12);
hold off;
347

02 Whole

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

02 Whole

Uploaded by

Copyright:

Available Formats

University of Wollongong Thesis Collections

University of Wollongong Thesis Collection

Belt conveyer transfers : quantifying and

This paper is posted at Research Online.

A thesis submitted in fulfilment of the

DAVID BRYAN HASTIE, BE (Hons), ME (Hons)

SCHOOL OF MECHANICAL, MATERIALS AND

David Bryan Hastie

The design of conveyor transfers can be based on experimental investigations, although

Figure 3.1 Final conveyor transfer research facility design 66

Figure 5.1 Stereopycnometer 82

Figure 6.1 Three dimensional CAD model of a conveyor transfer 101

Figure 8.1 Low-speed, horizontal conveyor, lower path, pulley diameter,

Figure 9.1 Detail of the conveyor transfer hood 210

Figure 10.1 Experimental particle freefall setup 244

Figure 11.1 Detail of the transfer spoon 256

Figure 12.1 Representation of how to estimate the impact velocity on the

Table 4.1 Belt speed calibration chart 77

Table 5.1 Particle and bulk characteristics of polyethylene pellets 93

Table 6.1 Sensitivity analysis based on 100% particle restrain 110

Table 8.1 Conveyor parameters used for comparisons 148

Table 9.1 Product feed rates used in experimental tests 211

Table 10.1 Experimental freefall tests 243

Table 11.1 Product feed rates used in experimental tests 257

a acceleration or deceleration of material on a straight chute m/s2

CN normal damping coefficient -

R radius for Hertz-Mindlin model m

1.2 OBJECTIVES AND SCOPE OF THE RESEARCH

1.3 ORGANISATION OF CHAPTER CONTENT

Figure 2.1 Transfer chute mechanisms

2.2 ISSUES RELATING TO CONVEYOR TRANSFER DESIGN

2.2.1 Conveyor Belt Wear and Damage

2.2.2 Spillage of Material

Another possible cause of spillage is at start-up or shutdown. In these cases, material is

2.2.3 Degradation of Material

2.2.4 Material Hang-ups

2.2.5 Blockage of the Transfer Chute

2.2.6 Noise Emissions

2.2.7 High Maintenance Costs

2.3 DUST CONTROL

2.4 CONVEYOR DISCHARGE AND TRAJECTORY

2 Rc ( RPM of end pulley )

2.4.2 M.H.E.A. 1977

2.4.3 M.H.E.A. 1986

substituting V 2 ( ) = Rc g cos and = d into equation 2.7 the discharge angle, d,

To determine whether low-speed or high-speed conditions apply, equation 2.2 is again

Table 2.1 Divergent coefficients (Golka et al., 2007)

CASE 1: V1 < Vcr and V2 < Vcr

CASE 2: V1 > Vcr

CASE 3: V1 < Vcr and V2 > Vcr

please see print copy for image

Figure 2.3 Material discharge when belt and head pulley

please see print copy for image

2.4.10 Trajectory Discussion

2.5 UPPER TRANSFER CHUTE DESIGN OPTIONS

2.5.1 Hood Design

please see print copy for image

Figure 2.5 Hood design (McBride, 1997)

Figure 2.6 Inverted curved chute

The radius of curvature of the discharge trajectory is presented in equation 2.44.

The coefficient of equivalent friction, e, for a chute having a rectangular cross-section

please see print copy for image

Figure 2.7 (a) Rectangular cross-section, (b) circular cross-section and

For a chute having a rectangular cross-sectional area, C1 is given by equation 2.51.

d2 ( 22.5 + d2.046 )